Abstracts - Web viewCensus population estimates and mortality records from 1981-2011, ... network meta-analysed each draw, and combined findings from each draw using Rubin’s rules.

Society for Social Medicine, 61st Annual Scientific Meeting, University of ManchesterSeptember 5-8 2017

Tuesday 5th September 2017- University Place, Oxford Road, University of Manchester10:00 onwards Early Career Researcher (ECR) Workshop: Room 6.207, University Place10:00 onwards Mid Career Researcher (MCR) Workshop: Room 6.205, University Place10:00 onwards SSM Mentoring Workshop: Room 6.206, University PlaceWednesday 6th September 2017- University Place, Oxford Road, University of Manchester10:00-18:00 Registration11:00-11:30 Conference welcome (Lecture Theatre B)11:30-12:45 Cochrane Lecture and Discussion (Lecture Theatre B): Professor Jennie Popay “What

happened to the ginger bread man: twenty years of qualitative evidence synthesis"

12:45-13:45 Lunch (ECR/MCR meetings and events)13:10-13:45 Early Career Researcher event (1.218)13:10-13:45 Mid-Career Researcher event (1.219)13:45-15:20 Oral presentation session A and Workshop 1 "Appraisal, revalidation and professional

responsibilities as an academic in public health"15:20-16:20 Refreshments and poster session A (P1-P48)16:20-17:55 Oral presentation session B and Workshop 2 "Context is crucial in interventions, but how do

we examine it?"19:00 Conference reception: Great Hall, Manchester Town Hall, Albert Square, M2 5DBThursday 7th Sept- University Place, Oxford Road, University of Manchester08:30 onwards Registration9:00-10:35 Oral presentation session C and Workshops 3 "Extending MRC Guidance on Developing and

Evaluating Complex Health Interventions" and 4 "Reviewing evidence to support decision-making on changes in service delivery and organisation: practical approaches and challenges"

10:35-11:30 Refreshments and poster session B (P49-P98)11:25-13:00 Oral presentation session D and Workshops 5 "Novel trial designs and analytical methods

for evaluating complex public health interventions: a discussion" and 6 "Thinking creatively about communicating your research"

13:00-13:45 Lunch13:45-15:00 "How do we change the social determinants of health and health inequalities?”- A discussion

with Professor Michael Marmot and Professor Richard Wilkinson (Lecture Theatre B)

15:15-17:00 Walking tours17:00-18:00 Society for Social Medicine Annual General Meeting (Lecture Theatre B)18:30 for 19:00 start Conference dinner: Harvey Nichols Second Floor Bar and Brasserie, 21 New Cathedral Street;

Manchester; M1 1ADFriday 8th Sept- University Place, Oxford Road, University of Manchester08:30 onwards Registration9:00-10:35 Oral presentation session E and workshop on "Use of Bayesian networks to facilitate

evidence synthesis and evaluation of interventions in complex systems"10:35-11:00 Refreshments11:00-12:10 Plenary session: top ranking abstracts (Lecture Theatre B)12:15-13:30 Pemberton Lecture and Discussion: Professor Alastair Leyland "God bless the child: empty

pockets and the struggle to reduce inequalities in health" (Lecture Theatre B)

13:30-13:45 Awards, closing remarks and farewell12 noon onwards Lunch (available to take away)

SSM Annual Scientific Meeting 2017 Oral Presentations (OP)

Wednesday 6 September 2017

Children 1

OP01

CHANGES IN THE RELATIONSHIP BETWEEN ASTHMA AND ASSOCIATED RISK FACTORS IN CHILDREN AGED 8-13 OVER FIFTY YEARS: ECOLOGICAL STUDY FROM ABERDEEN, NORTH EAST SCOTLAND

MS Barnish1, N Tagiyeva1, G Devereux1, L Aucott2, S Turner1

1Child Health, University of Aberdeen, Aberdeen, UK2Medical Statistics, University of Aberdeen, Aberdeen, UKBackground Asthma is the most common chronic childhood medical condition globally. After sharp rises in prevalence over the second half of the twentieth century, falling prevalence has been found in some countries including the United Kingdom during the first decade of the twenty-first century. In order to gain insight into the hitherto unconfirmed factors underlying changing susceptibility to asthma, we used data from one of the longest-running paediatric asthma epidemiology studies in the world: the Aberdeen Schools Asthma Survey (ASAS). We hypothesised that the relationship between asthma and associated risk factors had changed between 1964 and 2014.Methods An ecological study design was used. Parents of children aged 8-13 in state schools in the City of Aberdeen, North East Scotland, were invited to participate in a questionnaire survey in May 1964, May 1989 and then at five-year intervals to 2014. Child history of asthma and eczema, parental smoking, parental asthma, sex and socioeconomic status (SES) were determined. 2 knot structural change models, with knots after 1964 and 1999, were constructed to assess changes in the relationship between child history of asthma and these risk factors over time.Results Data for analysis were available for 15,108 children aged 8-13 (75% response rate). Asthma prevalence rose from 4% in 1964 to 28% in 2004 before falling to 22% in 2009 and 19% in 2014. Parental smoking prevalence fell in a near-linear fashion from 58% in 1989 to 28% in 2014. The odds ratio (OR) for a child with asthma to have eczema increased between 1989 and 1999 by 1.031 (95 confidence interval 1.028, 1.035) and by 1.042 (1.038, 1.047) between 2004 and 2014. The OR for a child with asthma to have a parent who smoked rose by 1.032 (1.028, 1.036) between 1989 and 1999 and by 1.043 (1.038, 1.047) between 2004 and 2014. The OR for a child with asthma to have a parent with asthma, to be male and to be from the most deprived communities also rose over the study period.Conclusion As hypothesized, we found that the relationship between asthma and associated risk factors such as child eczema, sex, parental smoking, parental asthma and deprivation changed over the period 1964 to 2014. Limitations in our study include regulatory changes and falling response rates over time. The changing nature of relationships with asthma suggests that modification of environmental exposures, e.g. to second-hand smoke, can reduce population asthma susceptibility.Keywords: epidemiology, paediatrics, asthma

OP02

TRENDS IN CURE AND RELAPSE BY CLINICAL CHARACTERISTICS FOR CHILDREN DIAGNOSED WITH LEUKAEMIA AGED 0-17 YEARS IN YORKSHIRE 1990-2009: A POPULATION-BASED STUDY

L Smith1, AW Glaser2, 3, SE Kinsey2, 3, DC Greenwood1, RG Feltbower1

1Division of Epidemiology and Biostatistics, University of Leeds, Leeds, UK2Leeds Institute of Cancer and Pathology, University of Leeds, Leeds, UK3Leeds General Infirmary, Leeds Teaching Hospitals NHS Trust, Leeds, UK

Background The 10-year survival estimates for children aged 0-14 years diagnosed with leukaemia have increased from 23% during the early 1970s to 81% for 2001-2005. Statistical cure models offer an alternative approach to examining survival by simultaneously estimating the proportion of patients cured and the survival of those ‘uncured’. The proportion cured is defined as the proportion of patients as a group for whom there is no excess mortality compared to the general population. The aims of this study were to estimate the cure proportion for childhood leukaemia and examine trends by clinical prognostic risk factors. Trends in relapse free survival were also examined. Methods Children aged 0-17 diagnosed with leukaemia between 1990 and 2009 were extracted from the Yorkshire Specialist Register of Cancer in Children and Young People (n=583). Flexible parametric cure models were used to estimate cure proportions and median survival times (MSTs) of those ‘uncured’ by age at diagnosis, sex, diagnostic subtype, white cell count (WCC), and period of diagnosis. A further cure model based on relapse free survival and a competing risk model for relapse with death as a competing risk were also fitted to examine patterns of relapse. Results The standardised (adjusting for age, sex, subtype and WCC) cure proportion increased from 0.63 (95%CI: 0.55-0.70) for those diagnosed between 1990 and 94 to 0.83 (95%CI: 0.75-0.88) for those diagnosed 2005-2009. Over this same time period the MST of the uncured remained around 2 years. There were significant differences in cure proportions by age, subtype and WCC, and differences in MST by age and subtype. Models based on relapse free survival found that the proportion cured increased from 0.45 (95%CI: 0.38-0.53) to 0.78 (95%CI: 0.71-0.84) and the MST to relapse or death remained between 1.5-1.7 years. The risk of relapse decreased over time (Hazard ratio 0.18 (95%CI: 0.10-0.31) for 2005-09 compared to 1990-94). Conclusion These results demonstrate that the proportion of patients cured, defined either by overall survival or relapse free survival, has increased substantially. There was no change in the median survival time of the uncured group during this time period, however, the risk of relapse has decreased. Cure models provide an alternative and clinically informative method to assess trends in survival for cancer patients.  Keywords: cancer, survival methods, children

OP03

IS IN-UTERO EXPOSURE TO MATERNAL H1N1 INFLUENZA INFECTION AND VACCINATION ASSOCIATED WITH AN INCREASED RISK OF CHILDHOOD SEIZURES? A NORWEGIAN REGISTRY-BASED STUDY

L Oakley1, IJ Bakken2, SE Håberg2

1Department of Non-communicable Disease Epidemiology, London School of Hygiene and Tropical Medicine, London, UK2Department of Children’s Health, Norwegian Institute of Public Health, Oslo, Norway

Background Previous studies have suggested that in-utero exposure to infection is associated with an increased risk of childhood seizures, but there is a lack of evidence regarding in-utero exposure to influenza. The objective of this study was to investigate whether in-utero exposure to the H1N1 pandemic, influenza infection, or vaccination is associated with a higher risk of childhood seizures.Methods Registry-based study including all children born in Norway between 01/10/2009 and 31/12/2015 (n=254,347). Data were linked from sources including the Medical Birth Registry, the Norwegian Immunization Register, the primary care reimbursement system, and the Norwegian Patient Registry. We investigated three exposures: 1) in-utero exposure to the H1N1 pandemic (≥1 pregnancy day during the main H1N1 pandemic wave), 2) in-utero exposure to maternal influenza infection (diagnosis of influenza-like illness in primary care, and/or laboratory confirmed H1N1

infection), and 3) in-utero exposure to H1N1 vaccination. We used Cox Proportional Hazards modelling to compare the incidence of seizures (any seizure, febrile seizure, epilepsy) according to exposure status from birth until 31/12/2015. Hazard ratios were adjusted for parity, maternal age, multiplicity, sex and maternal smoking. Results 24.4% (62,032) children were exposed in-utero to the H1N1 pandemic, of whom 3.7% (2,299) were exposed in-utero to maternal influenza. Among 77,671 children with ≥1 in-utero day during the vaccination period, 34.9% (n=27,138) were exposed to vaccination. The risk of febrile seizures was slightly increased after in-utero exposure to the pandemic (aHR 1.06, 95%CI 1.00-1.12), but there was no evidence of an increased risk of epilepsy (aHR 1.08, 95%CI 0.93-1.26). There was no evidence of an overall association between in-utero exposure to maternal H1N1 infection and childhood seizures (febrile seizures aHR 1.17, 95%CI 0.92-1.49; epilepsy aHR 0.93, 95%CI 0.50-1.75). However, when stratified by trimester of exposure we observed a 40% increased risk of febrile seizures after infection during the second trimester (aHR 1.42, 95%CI 1.02-1.99). In-utero exposure to vaccination was not associated with an increased risk of childhood seizures. Discussion This large study benefits from virtually no loss to follow-up and mandatory vaccination reporting. The limitations includes our inability to validate outcome data, and the under-reporting of influenza infection.Our finding of no increased risk subsequent to in-utero exposure to H1N1 vaccination supports the safety of vaccination in pregnancy. Although we found no overall evidence that in-utero exposure to maternal H1N1 infection was associated with febrile seizures, a small increased risk of febrile seizures after second trimester exposure warrants further investigation. Keywords: pregnancy, influenza, febrile seizures

OP04

THE ASSOCIATION BETWEEN THE CHILDCARE ENERGY-BALANCE ENVIRONMENT AND UK 3-4-YEAR-OLDS’ ANTHROPOMETRIC INDICES

KR Hesketh1,2, SE Benjamin Neelon1,2, EM van Sluijs1

1CEDAR and MRC Epidemiology Unit, University of Cambridge, Cambridge, UK 2Department of Health, Behavior and Society, Johns Hopkins Bloomberg School of Public Health, Baltimore, MD, USA

Background Although parents provide the majority of childcare before they enter formal schooling, preschool-aged children now spend increasing amounts of time in out-of-home care. Childcare centres, in addition to parents, therefore have a significant responsibility for shaping children’s energy-balance behaviours. However, little is known about how the UK childcare environment influences children’s anthropometric indices.  We assessed how the amount of time spent in childcare, and how the nutrition, physical activity and overall childcare environment were associated with children’s anthropometric indicators (z-BMI score; waist-to-height ratio (WHR); sum of skinfold thickness (SST)).Methods We recruited 3-4-year-old children across socio-economic strata from 30 childcare centres in Cambridgeshire, UK. Trained personnel measured children’s height, weight, waist circumference, and subscapular and tricep skinfolds. Parents reported weekly childcare attendance patterns; we assessed the childcare environment relating to obesity (e.g. nutrition and physical activity) using the Environment and Policy Assessment and Observation system. We explored associations between childcare attendance and environment and anthropometric outcomes using two-level hierarchical regression (level 1: child; level 2: childcare centre). All models were adjusted for child ethnicity; maternal educational attainment; maternal BMI; and maternal employment. WHR and SST models were additionally adjusted for child sex and age in months (which are both taken into account when calculating z-BMI scores).Results 196 children (49% female) from 30 childcare centres provided valid data. Neither time spent in care, nor the nutrition, physical activity, or overall childcare environment were associated with children’s z-BMI score, WHR and SST. These findings remained after adjusting for child and

maternal variables; several of the latter were independently associated with the outcomes of interest. Discussion In constrast to previous international evidence, neither time spent in childcare nor the environment itself were associated with UK preschool-aged children’s adiposity-related outcomes. The childcare environment remains important to the Government’s obesity strategy, and has been central to intervention efforts to prevent or reduce early childhood obesity to date. However, family-level factors also warrant substantial attention when considering obesity prevention strategies for young children.Keywords: Childcare Family Anthropometry

Health Inequalities 1

OP05

WHICH AGES AND CAUSES OF DEATH EXPLAIN THE WIDENING LIFESPAN VARIATION GAP IN SCOTLAND?  A POPULATION BASED STUDY USING ROUTINE DATA

R Seaman1,2, A.H. Leyland2, F Popham2

1Max Planck Institute for Demographic Research , The Max Planck Society , Rostock , Germany2MRC/CSO Social and Public Health Sciences Unit, University of Glasgow , Glasgow , UK

Background Scotland’s relative lifespan variation ranking within Western Europe deteriorated after 1980.  It is not clear how Scotland’s national lifespan variation trend is associated with socioeconomic inequalities in age and cause of death. We calculate lifespan variation for deprivation quintiles over a thirty year period. We apply stepwise decomposition by age and cause of death to better understand the changing nature of mortality inequalities. Methods Census population estimates and mortality records from 1981-2011, were matched with the Carstairs score, an area level measure of relative deprivation. Life tables by year, sex and deprivation quintile were constructed. Lifespan variation was calculated using e†. The magnitude of inequalities were estimated using the slope and relative indices of inequality. 95% confidence intervals were produced using Monte-Carlo simulation. The lifespan variation gap between the most and least deprived at (1) the same time point and (2) a comparable level of life expectancy was decomposed. The sensitivity of the results to starting at age 0 were tested by repeating the analysis starting at age 35.Results Lifespan variation  for males from the most deprived quintile was 12.2 years (12.1 years-12.3 years) in 1981 and increased to 12.3 years (12.1 years-12.4 years) in 2011. For the least deprived lifespan variation decreased from 11.2 years (11.0 years-11.3 years) to 10.4 years (10.3 years-10.6 years). This caused the socioeconomic gap to widen over time in absolute and relative terms. In 2011 there was a 2.1 year (1.9 year-2.4 year) difference or a 19%(17%-21%) difference. The gap widened because of increasing differences in mortality rates across working ages from external causes. In 1981 external causes explained 55.1% of the gap and by 2011 they explained 69.5% of the gap. Deaths from circulatory disease explain less of the lifespan variation gap over time. At a shared level of life expectancy the most deprived quintile experience higher mortality rates from external causes of death despite arriving at this life expectancy thirty years later in time. Substantive conclusions were unchanged during sensitivity analysis.Conclusion The lifespan variation gap widened because of deaths across working ages from external causes. Scotland must reduce deaths across working adult ages from external causes if it is to reduce the gap and improve its ranking within Western Europe. Routinely monitoring lifespan variation inequalities is valuable for extending our understanding of the changing nature of mortality inequalities and is relevant for countries considering which public health strategies will reduce mortality inequalities.Keywords: Lifespan variation, mortality inequalities, age and cause of death

OP06

SOCIO-DEMOGRAPHIC INEQUALITIES IN CARDIOVASCULAR RISK MANAGEMENT AND EARLY DETECTION OF VASCULAR CONDITIONS BY THE NHS HEALTH CHECK: A DIFFERENCE-IN-DIFFERENCES MATCHING ANALYSIS

K Chang1, JT Lee2, E Vamos1, R Palladino1, M Soljak1, A Majeed1, C Millett1

1Department of Primary Care & Public Health, School of Public Health, Imperial College London, London, United Kingdom2Saw Swee Hock School of Public Health, National University of Singapore, Singapore, Singapore

Background England’s National Health Service (NHS) Health Check is a nationwide cardiovascular risk assessment and management programme implemented with aims to prevent cardiovascular disease (CVD), type 2 diabetes mellitus (T2DM) and chronic kidney disease, as well as to reduce inequalities in health. We aimed to compare the impact of the NHS Health Check on: i) Early detection of vascular conditions among population subgroups of age, sex, ethnicity and deprivation; and ii) The management of cardiovascular risk among high-risk population subgroups of age, sex, and deprivation.Methods We obtained retrospective electronic medical records from the Clinical Practice Research Datalink for a randomly selected sample of 138,788 patients aged 40–74 years, without known CVD or diabetes, and were registered with 462 English general practices between 2009 and 2013. We estimated programme impact for each subgroup using a difference-in-differences matching analysis that compared changes in outcome over time and between Health Check attendees and non-attendees.Results 21.4% (29,672/138,788) of the study population attended a Health Check. The programme was associated with increased detection of hypertension and T2DM among Health Check attendees. A significantly greater number of hypertension and T2DM incident cases were detected in male than female attendees (e.g. an additional 4.02%, 95% CI: 3.65% to 4.39%, and 2.08%, 1.81% to 2.35% male and female attendees were detected with hypertension respectively). A significantly greater number of T2DM incident cases were detected among attendees living in the most deprived area (1.60%, 1.23% to 1.97%) compared with those living in the least deprived area (0.79%, 0.52% to 1.06%).The programme was associated with significant reductions in 10-year CVD risk scores, total cholesterol and systolic blood pressure while statin prescribing increased among high-risk attendees. However, no major differences in programme impact on cardiovascular risk management were observed between subgroups (e.g. programme impact on 10-year CVD risk score was -1.13%, -1.48% to -0.78% in male and -1.53%, -2.36% to -0.71% in female attendees).Conclusion The NHS Health Check may have narrowed inequalities in the diagnosis of hypertension and T2DM but inequalities in the management of CVD risk remained unchanged. Proactive strategies may be required to address known inequalities in CVD outcomes.Keywords: health inequalities, prevention of cardiovascular disease, health services research

OP07

IMPACT OF AN INTEGRATED HEALTH AND WELLBEING APPROACH TO ADDRESSING MULTIPLE LIFESTYLE RISKS AND REDUCING HEALTH INEQUALITIES: A MIXED METHODS STUDY

S Visram1, F Thirlway2, N Akhter3, N Walton4, S Lewis4

1School of Medicine, Pharmacy and Health, Durham University, Stockton-on-Tees, UK2Department of Anthropology, Durham University, Durham, UK

3Wolfson Research Institute for Health and Wellbeing, Durham University, Durham, UK4Department of Geography, Durham University, Durham, UK

Background The return of public health to local government in England in 2013 created an opportunity to integrate preventive services with agencies that act on the wider determinants of health. A number of local authorities subsequently developed integrated health and wellbeing approaches, in recognition that the previous ‘silo’ approach to the provision of single-issue lifestyle services had made little impact on inequalities. These integrated services often involve targeting the most disadvantaged geographical and non-geographical communities locally. One example is the Wellbeing for Life (WFL) service in County Durham.Methods The impact of WFL was evaluated using a mixed methods study design, involving: i) ethnographic observations plus interviews and focus groups with clients (n=58), staff (n=47), volunteers (n=15) and external stakeholders (n=10); ii)  secondary analysis of intervention monitoring data at baseline (n=1461 clients), three (n=1201), six (n=380) and 12 months (n=133); and iii) a value for money assessment. Primary outcome measures were the EQ-5D™ and short Warwick Edinburgh Mental Wellbeing Scale (SWEMWBS). Secondary outcome measures were BMI, fruit and vegetable intake, physical activity, alcohol consumption, smoking and self-efficacy.Results Clients reported positive experiences of engaging with the service and described making lifestyle changes, such as reducing portion sizes or trying new activities. Many would have liked a longer intervention (usually 8-12 weeks) and staff agreed that the duration was often insufficient for those with more complex physical, mental and/or social needs. Pre/post analysis of the intervention data revealed significant improvements across all outcome measures, with the largest changes observed in clients with the least positive results at baseline. These changes were largely maintained at both follow-up periods. Furthermore, reductions in the differences in EQ-5D™ and SWEMWBS scores between clients from the 30% most deprived communities and all other clients at the six-month follow-up stage implied that inequalities had narrowed over time. The value for money assessment indicated an estimated cost per quality-adjusted life year (QALY) of £3900 and a social return on investment of around £3.59 for every £1 spent on WFL.Discussion An integrated health and wellbeing approach can be acceptable to members of the target communities, encourage them to make and maintain lifestyle changes, and potentially reduce health inequalities. The WFL service appeared to represent good value for money, although the reliance on self-report data and lack of a control group were limitations of the study design. Further research is needed to establish the effectiveness of this approach over other intervention models.Keywords: wellbeing, lifestyle, evaluation

Lifecourse Later Life

OP08

EXPLAINING TRENDS IN CORONARY HEART DISEASE MORTALITY AND SOCIOECONOMIC INEQUALITIES IN DENMARK 1991-2007: IMPACTSEC MODEL ANALYSIS USING ROUTINE DATA

AM Joensen1, T Joergensen2,3,4, S Lundbye-Christensen5, MB Johansen5, M Guzmán-Castillo6, P Bandosz6, J Hallas7, EIB Prescott8, S Capewell6, M O'Flaherty6

1Department of Cardiology, Aalborg University Hospital, Aalborg, Denmark2Research Center for Prevention and Health, The Capital Region, Glostrup, Denmark 3Department of Public Health, University of Copenghagen, Copenhagen Denmark4Department of Clinical Medicine, Aalborg University, Aalborg, Denmark5Unit of Clinical Biostatistics and Bioinformatics, Aalborg University Hospital, Aalborg, Denmark6Department of Public Health and Policy, University of Liverpool, Liverpool, United Kingdom7Clinical Pharmacology, Department of Public Health, University of Southern Denmark, Odense, Denmark

8Bispebjerg University Hospital, Capital Region of Denmark, Copenhagen, Denmark

Background Coronary heart disease (CHD) mortality has declined substantially during recent decades but is still one of the leading causes of death, morbidity and healthcare costs in Denmark. Furthermore, socioeconomic inequalities persist. Quantifying the contributions of prevention and treatment to these recent declines might help to identify the most successful health policies, particularly for reducing inequalities.Methods We used IMPACTSEC, a previously validated policy model, to apportion the recent decline in Danish CHD mortality to changes in major cardiovascular risk factors, and to increases in treatments in nine non-overlapping patient groups. Participants: All Danish adults aged 25-84 years, stratified by gender, age group and quintiles of financial income. Main outcome measure: Deaths prevented or postponed (DPP), stratified by socio-economic circumstance (SEC).Results There were 11,110 fewer CHD deaths in 2007 than would be expected if the 1991 mortality rates had persisted. This reflected a dramatic 74% fall in CHD mortality rates (from 433 to 113 deaths per 100,000). Improved treatments accounted for approximately 24% (95% confidence interval = 21%-28%). This contribution was higher in more affluent quintiles (approximately 26%) and least in the most deprived group (19%). The biggest contributions came from the treatment of congestive heart failure in the community (630 DPPs=5.7% of all DPPs) and in hospital (410 DPPs=3.7%). Risk factor improvements accounted for approximately 40% (37-44%) of the mortality fall. This contribution was higher in the central quintiles -approximately 51% (47-58%) and least in the most deprived quintile – approximately 36% (29-39%). The largest contribution came from population falls in cholesterol levels approximately 24% (22.7-25.4%) of all DPPs; and decreases in smoking, some 10% (8.4-12.2%). Overall, the IMPACTSEC model could explain two thirds of the mortality fall. The 36% gap most likely reflects deficiencies in data, notably in population blood pressure and income.Conclusion Denmark has benefited from one of biggest falls in CHD mortality in high income countries. The treatment uptake rate in Denmark was comparable with that in other countries and treatments accounted for approximately one third of the total mortality fall, much as in other, comparable populations.  The largest contributions came from population-wide, non-phamacological reductions in major risk factors, notably cholesterol and smoking. Future strategies should therefore prioritise population-wide prevention policies.Keywords: Risk factors, Coronary heart disease, Modelling study

OP09

CHILDHOOD COGNITIVE ABILITY AND STANDING BALANCE IN MID TO LATER LIFE: FINDINGS FROM THE MRC NATIONAL SURVEY OF HEALTH AND DEVELOPMENT

JM Blodgett1, D Kuh1, DHJ Davis1, R Cooper1

1MRC Unit for Lifelong Health and Ageing , UCL, London, United Kingdom

Background Physical performance indicators, such as standing balance, grip strength and walking speed, are increasingly being used as markers of healthy ageing. This is based on growing evidence that poor performance on these tests is associated with adverse health outcomes including falls, disability, hospitalization and mortality. Individual variation in the levels of performance on each of these tests may be due to diverse contributing factors across life. Previous research has suggested that neurodevelopmental pathways may be particularly important for balance performance; however the few published studies on this have only examined balance at one age. We aimed to examine the associations between childhood cognitive ability, a marker of neurodevelopment, and standing balance at three ages in mid and later life. 

Methods Up to 2785 participants from the MRC National Survey of Health and Development, a British cohort study followed since birth in 1946, were included in analyses. Standing balance was assessed at ages 53, 60-64 and 69 using the one-legged stand test with eyes closed up to a maximum of 30 seconds. Childhood cognitive ability was assessed using a set of standardized tests at age 15.   Linear regression models estimated the association between childhood cognitive ability and log-transformed balance at each age.  Adjustments were made for sex, childhood socioeconomic position, current physical activity, height and weight.  Analyses were performed using STATA v14.1.Results Men had better median balance times than women at age 53 [Men: 5 (interquartile range: 3-10); Women: 4 (3-7)], 60-64 [M: 3.7 (2.5-5.6); W: 3.3 (2.3-4.8)], and 69 [M: 3.0 (2.0-5.0); W: 2.9 (1.9-4.3)], but a decline in median balance times with age was observed in both sexes. In sex-adjusted and fully-adjusted models, higher childhood cognitive ability was associated with better balance times, although this association weakened with increasing age. A one standard deviation increase in childhood cognitive ability was associated with fully-adjusted mean differences in log-balance times (ln(s)) of 0.12 (95% CI: 0.08−0.15), 0.05 (95% CI: 0.02−0.09) and 0.04 (95% CI: 0.001−0.08) at ages 53, 60-64 and 69, respectively. Conclusion Balance performance decreased with age, and was consistently lower in women than men. Higher childhood cognitive ability was associated with better balance performance at all ages but the strength of this association weakened over time. Further research should examine how the associations of other neurodevelopmental factors with balance change with age. Interventions earlier in life may help to maintain balance performance from midlife.Keywords: Physical function, cognition, life course epidemiology

OP10

RESIDENTIAL MOBILITY DURING CHILDHOOD AND LATER RISKS OF PSYCHIATRIC MORBIDITY, VIOLENT CRIMINALITY AND PREMATURE DEATH: A NATIONAL REGISTER-BASED COHORT STUDY

RT Webb1, PLH Mok1, L Appleby1, CB Pedersen2,3

1Centre for Mental Health and Safety, The University of Manchester, Manchester, UK2Centre for Integrated Register-based Research (CIRRAU), Aarhus University, Aarhus, Denmark3National Centre for Register-based Research (NCRR), Aarhus University, Aarhus, Denmark

Background Links between childhood residential mobility and pernicious trajectories through to adult maturity are incompletely understood. We therefore carried out a comprehensive investigation of the relationships between residential relocation during upbringing across a broad range of later adverse outcomes.Methods Using interlinked registers, a national cohort of N=1,475,030 persons born in Demark during 1971-1997 was followed up from 15th birthday to a maximum age of early 40s for risks of psychiatric morbidity, substance misuse, violent criminality, and natural and unnatural mortality. All cross-municipality residential moves during each age-year between birth and 15th birthday were measured. Survival analyses (© SAS Institute Inc.) was used to estimate incidence rate ratios (IRRs) associated with relocation during each age-year of upbringing. Parental socioeconomic status was delineated according to income, educational attainment and employment status.Results Elevated risks were observed for virtually all outcomes examined, with excess risk seen among those who experienced multiple versus single relocations in a year. For attempted suicide, violent offending, substance misuse, personality disorders and unnatural deaths, we observed especially sharp spikes in risk linked with multiple relocations in a year during early/mid-adolescence. For example, diagnosis for substance misuse from age 15 was over six times more likely among adolescents who moved two or more times at age 14 than for their peers (IRR 6.35; 95% CI 5.82, 6.92). With violent offending and attempted suicide, our two primary outcomes, we

found a risk gradient with increasing age at exposure among higher as well as lower socioeconomic groups. A consistent pattern of markedly elevated risk was not seen among all psychiatric illnesses examined; thus, eating disorders and obsessive compulsive disorder were not strongly linked with residential mobility during upbringing.Conclusion As with other register-based studies, potential for residential confounding is a limitation of this investigation. Nonetheless, whatever causal mechanisms are implicated, frequent residential change is a strong marker for serious familial adversities. Heightened vigilance is therefore indicated for relocated youths. Health, educational and social services should be mindful of the potential psychosocial needs of adolescents who have recently moved homes or who have relocated more than once over a short period, and effective risk management will require close cooperation between multiple public agencies. When addressing the vulnerabilities of troubled young people, the possible harmful impact of residential relocation during childhood should be considered for individuals from affluent or middle class backgrounds, as well as those who grew up in deprived families.Keywords: childhood adversity; self-harm; violence

OP11

LIFE COURSE SOCIOECONOMIC POSITION AND THE PREVALENCE OF TYPE 2 DIABETES IN LATER LIFE. A CROSS-SECTIONAL ANALYSIS FROM THE IRISH LONGITUDINAL STUDY OF AGEING

S Leahy1, M Canney1, S Scarlett1, RA Kenny1, 2, C McCrory1

1The Irish Longitudinal Study of Ageing, Trinity Colloge Dublin, Dublin, Ireland2Mercer's Institute of Successful Ageing, St. James's Hospital, Dublin, Ireland

Background A substantial body of research has pointed to an association between socioeconomic position (SEP) and the prevalence of type 2 diabetes (T2D), whereby those in lower social classes are disproportionately affected by the disease. However few studies have examined the contribution of SEP trajectories across the life course to the development of T2D. We investigate the independent and synergistic effects of childhood and adult SEP, as well as the effect of social mobility, on T2D risk in later life.Methods Cross-sectional data from The Irish Longitudinal Study of Ageing (TILDA) (n=4998), a nationally representative probability sample of adults aged 50 and older, were analysed. Prevalent diabetes was defined using subjective (self-reported doctors diagnosis) and objective data (medications usage and glycated haemoglobin testing). SEP was classified as a three level variable (low, intermediate, high) based on fathers occupation in childhood (origin SEP) and respondents primary occupation in adulthood (destination SEP). A 5-level social mobility variable was created from cross-classification of origin and destination SEP (‘Stable High’ ‘Stable Intermediate’, ‘Stable Low’ ‘Upwardly Mobile’ and ‘Downwardly Mobile’). Logistic regression was employed to assess the relationship between SEP variables and T2D. All analyses were adjusted for age and age2, and stratified by sex.Results Mean (SD) age of the sample was 63.0(9.2)y and 46.4% were male. Prevalence of T2D was 9.5% (95%CI: 8.6%-10.6%). 53.4% of the sample were classified as ‘Low SEP’ in childhood which decreased to 33.7% in adulthood.Compared to high SEP, low SEP in both childhood (Odds Ratio (OR): 1.84, 95%CI: 1.00-3.37) and adulthood (OR: 1.78, 95%CI: 1.02-3.13) was independently associated with T2D in women. When classified according to social mobility, women classified as ‘Stable Low’ were at greatest risk of T2D (OR: 2.51, 95%CI: 1.24-5.06) compared to those classified as ‘Stable High’. No associations were noted between any SEP variables and T2D in men.Conclusion This study confirms a strong association between low socioeconomic position and T2D in women which persists from childhood through to adulthood. These findings support the critical

period hypothesis which suggests that social disadvantage experienced in early life may have long lasting health consequences – in this case an increased risk of T2D. As many risk factors for T2D result from poor health behaviours which are likely adopted in early life, interventions to reduce T2D and its causes at a population level should recognise high-risk groups at all stages of the life course.   Keywords: socioeconomic inequalities diabetes

OP12

CUMULATIVE LIFECOURSE ADVERSITY AND ADAPTATION IN LATER LIFE 

S Richardson1, A Sacker1, G Netuveli1,2, E Carr1

1Research Department of Epidemiology and Public Health, International Centre for Lifecourse Studies, University College London, London, United Kingdom2Institute for Health and Human Development, University of East London, London, United Kingdom

Background Although exposures to cumulative socioeconomic disadvantage and adverse events over the lifecourse are associated with impaired physical and psychological health, the role of adverse events has received less attention. Furthermore, it is unclear to what extent their effects on later-life functioning depend on whether their primary harm was to the self or another person and their timing in the lifecourse, for example, during ‘critical’ or ‘sensitive’ periods.Methods We used data on 5,231 respondents aged 50+ years over seven waves of the English Longitudinal Study of Ageing (2002–2015) to investigate adaptation in later life using cross-sectional CASP-12 scores, subjective life satisfaction and (CES-D) depression as outcomes. Cumulative lifecourse adversity was measured by counts of 16 types of adverse events occurring within five stages over the lifecourse (ages 0–5, 6–15, 15–30 and 31–50) using retrospective life history data. We fitted linear and logistic multilevel random intercept models in Stata 14 (for repeated observations nested within individuals) to evaluate the extent to which adverse events influence later life wellbeing and whether these associations differ according to self-versus-other orientation. Models were adjusted for labour market status, physical frailty score based on the cumulative deficit model, income, wealth and other household variables. Finally, we tested the association between cumulative adversity and trajectories of CASP-12 scores over time using a latent growth curve model.Results CASP-12 scores were reduced by -0.49 (95% CI: -0.56 to -0.42) for each additional adverse event. This effect was similar for events occurring in each life stage and similar results were found for subjective life satisfaction and depression outcomes. Self-oriented events occurring in childhood had a greater (p<0.001) negative association (-0.62, 95% CI: -0.79 to -0.45) with later life wellbeing when compared with other-oriented events (-0.14, 95% CI: -0.32 to 0.03). Conversely, other-oriented events in adulthood exerted a greater influence. Total adverse life events were not associated with trajectories of CASP-12 by age.Conclusion Adverse events occurring at all stages of the lifecourse were found to independently influence adaptation in later life. These age-dependent effects differed according to their self- or other-orientation, however. Our results support the theory of allostasis, in which previous exposure to stressors results in excessive allostatic load, susceptibility to future stressors, maladaptation and functional decline.Keywords: Lifecourse, ageing, wellbeing

Health Services Research

OP14

A SYSTEMATIC REVIEW OF COST-EFFECTIVENESS EVALUATIONS OF PSYCHOLOGICAL THERAPIES FOR SCHIZOPHRENIA AND BIPOLAR DISORDER

D Buck1, GE Shields1, J Elvidge2, K Hayhurst3, LM Davies1

1Centre for Health Economics, University of Manchester, Manchester, UK2National Institute for Health and Care Excellence, Manchester, UK3Centre for Mental Health & Safety, University of Manchester, Manchester, UK

Background Schizophrenia and bipolar disorder (BD) are among the top twenty causes of disability. Costs of treatment and to society are substantial. NICE guidelines recommend psychological/psychosocial interventions are considered in response to acute episodes/recovery promotion. This review aims to assess cost-effectiveness of psychological interventions, determine robustness of the current evidence base, and identify evidence gaps. The key research question is: which psychological interventions are cost-effective, compared to usual care/alternative interventions, in schizophrenia or BD?Methods Electronic searches of PsycINFO, MEDLINE and Embase identified economic evaluations relating incremental cost to outcomes in an Incremental Cost-Effectiveness Ratio (ICER) published in English since 2000. Studies had to include: probability of cost-effectiveness at explicitly-defined thresholds; adults with schizophrenia/BD; any psychological intervention (e.g. psychological therapy, Improving Access to Psychological Therapies, integrated/collaborative care). Comparators could be routine practice, no intervention, or alternative psychological therapies. Searches were performed in August 2015 (updated January 2017). There were two screening stages with explicit inclusion criteria applied by 2 reviewers at each stage. Pre-specified data extraction/critical appraisal were performed. Results were summarised qualitatively. The review is registered on the PROSPERO database of systematic reviews.Results Of 3,785 studies identified, 11 were included. All were integrated clinical and economic randomised controlled trials. All used cost-effectiveness and/or cost-utility analysis. The commonest intervention was CBT (6/11 studies). Measures of health benefit included QALYs (5/11), QLS 1/11), PANSS (2/11), MANSA (1/11), GAF (3/11), days with normal functioning (1/11), a working memory subscale (1/11), full vocational recovery (1/11), days with a bipolar episode (1/11). Follow-up ranged from 6-months to 5-years. 6/11 studies used provider perspectives for the primary analysis; the remainder considered societal perspectives. Interventions were cost-effective in most identified studies (9/11): ICERs ranged from dominant (intervention is cost-saving AND more effective) to £18,844 per QALY; the probability of cost-effectiveness ranged from 50% to 99.5% at chosen thresholds. The two studies deemed not cost-effective involved art/body psychotherapy and noted significant uncertainty in the data as a limitation. All studies had limitations, including missing data, sample sizes and challenges controlling for other medications/treatments received outside the trial intervention.Conclusion Although recommended in clinical guidelines, there was limited evidence about the cost-effectiveness of psychological therapy for schizophrenia/BD. Most included studies concluded psychological interventions for schizophrenia/BD are cost-effective. However policy implications are unclear due to methodological limitations and heterogeneity in populations and settings between studies. The review had some limitations including potential for English-language bias and limited time-horizon.Keywords: Economic evaluation, mental health, review

OP13

THE CHANGING PREVALENCE OF BIRTHS AFTER SUBFERTILITY AND FERTILITY TREATMENT IN ENGLAND 1991-2013: EVIDENCE FROM THE CLINICAL PRACTICE RESEARCH DATALINK

C Carson1, MA Quigley1, JJ Kurinczuk1, C Bankhead2, S Stevens2

1National Perinatal Epidemiology Unit, Nuffield Dept of Population Health, University of Oxford, Oxford, UK2Nuffield Department of Primary Care Health Sciences, University of Oxford, Oxford, UK

Background We describe the prevalence of births after subfertility and fertility treatment seen in primary care in England between 1991 and 2013, and examine the impact of changing maternal characteristics over time.Methods Data from the Clinical Practice Research Datalink (CPRD) Mother-Baby dataset were used, comprising records from >600 general practices across England, linked to index of multiple deprivation (IMD). 440,623 mothers registered for >18months prior to the birth of their child were included; 239,781 first-time mothers were analysed separately to assess changes in primary infertility. Fertility history was identified using details of diagnoses, referrals and prescriptions in GP records, and grouped as: ‘no evidence of fertility problems’ and ‘any subfertility’ (comprising ‘untreated’, ‘ovulation induction’ (OI), and ‘Assisted Reproductive Technologies’ (ART), such as IVF). Change in the prevalence of births by fertility group and population characteristics was assessed using chi-squared trend tests. Direct age-standardisation (to 1991 study population) accounted for changing population structure. Results Overall, 4.7% of mothers experienced subfertility (1991: 1.7% – 2013: 6.3%), comprising: untreated 2.6% (1991: 0.6% – 2013: 3.5%), OI 0.8% (1991: 0.6%; peaking in 1995: 1.3%; 2013: 0.5%), and ART 1.3% (1991: 0.5% - 2013: 2.4%) (all p<0.001 for trend).   Women now tend to have children later – 13.5% of mothers were >35yrs in 1991, rising to 27% by 2013.  Age-adjusted 2013 figures were 3.1%, 1.8%, 0.2% and 1.1%, for all, untreated, OI and ART respectively.6.6% of first-time mothers experienced subfertility (1991: 2.1% - 2013: 9.7%): comprising untreated 3.6% (1991: 0.8% - 2013: 5.2%), OI 1.0% (1991: 0.7% - 2013: 0.6%), and ART 2.0% (1991: 0.7% - 2013: 3.9%) (all p<0.001 for trend).  Age-adjusted 2013 figures were 7.0%, 3.9%, 0.4 %and 2.7%, for all, untreated, OI and ART respectively. Subfertility was more prevalent in more advantaged women, with growing disparity suggested (age-adjusted 2.1% in IMD1 vs 1.3% in IMD5 in 1991, increasing to 5.1% vs 1.4% by 2013).Discussion Births to women with records of fertility problems have significantly increased between 1991 and 2013.   Changes in maternal age explain much of the population-level trends, but less of the observed increase among first-time mothers.  Declining births after OI prescribed in primary care may reflect changing management of patients. Medical advice must continue to highlight the effect of age on fertility, and the implications for secondary infertility. GPs and service commissioners should be aware that time trends indicate continuing growth in demand for fertility treatment in England.Keywords: Infertility, change over time, routine data

OP15

LONG TERM OUTCOMES AND MORTALITY AMONG PATIENTS ENROLLED IN A STRUCTURED PRIMARY CARE-LED DIABETES PROGRAMME

F Riordan1, SM McHugh1, V Harkins2, PM Kearney1

1Department of Epidemiology and Public Health, University College Cork, Cork, Ireland2Midland Diabetes Structured Care Programme, Ireland

Background Limited data exists, internationally and in Ireland, on long-term outcomes among people with diabetes who are managed in primary care. The Midlands Diabetes Structured Care Programme encompasses evidence-based strategies to structure diabetes management within general practice: patient registration and recall, regular diabetes review visits, active role of the

practice nurse in ongoing management, multidisciplinary specialist access, professional education, and remuneration. Our aim was to examine clinical outcome targets, complications and mortality among patients with diabetes enrolled in the programme since its establishment in 1998.Methods Data were collected in 1999, 2003, 2008 and 2015, on outcomes (clinical parameters, complications and mortality) among patients with diabetes (≥18 years) registered with participating practices.  Data were extracted from patient notes by clinical nurse specialists using a paper-based data collection form. Cause and date of death were obtained from national death records. Using Stata, chi-square tests were used to test differences in clinical outcomes over time. Cox proportional hazards regression was used to examine the association of baseline factors and mortality.Results Patients from 1999 (n=376), were followed up in 2003 (n=229), 2008 (n=96) and 2016 (n=376).The proportion of patients with a recommended blood pressure target (<130/80mmHg) increased from 9% in 1999 to 26% in 2016 (P<0.001), as did the proportion with a total cholesterol of <4.5mmol/L (22% vs.71%, P<0.001), and triglycerides <2.0mmol/L (47% vs.81%, P<0.001). The percentage achieving optimal glycaemic control (HbA1c ≤7.0%) declined (52% vs. 34%). Between 1999–2016, 22% (n=81) of patients had ever experienced a macrovascular complication; primarily CVA (n=21, 6%), MI (n=16, 4%). In 1999, 18% (n=33) had retinopathy, increasing to 57% (n=59) by 2016.   In total, 184 (49%) had died. Between 1999-2013 mortality was higher than background rates in the general population (SMR=2.2, 95% CI 1.9, 2.6). Only 25% (n=46) had cause of death recorded in their GP record. Where cause of death was obtained from national records (n=163), primary causes were MI (n=29, 17.8%) and pneumonia (n=23, 14%). Mean age at death was 77.7±9.3 years. Mortality was significantly higher among patients who were older at baseline. Gender, diabetes type, smoking status and clinical parameters at baseline were not significant predictors of mortality.Discussion Improvements in the clinical profile of patients enrolled in the programme since its introduction suggests primary-care-led integrated diabetes management can perform favourably in the long-term. However, the high incidence of macrovascular complications, prevalence of retinopathy and mortality rate indicates the importance of effective management.Keywords: structured care, diabetes, primary care

OP16

CHALLENGES EXPERIENCED BY COMMUNITY-BASED CLINICAL NURSE SPECIALISTS IN SUPPORTING THE DELIVERY OF INTEGRATED DIABETES CARE: A QUALITATIVE STUDY

F Riordan1, SM McHugh1, PM Kearney1

1Department of Epidemiology and Public Health, University College Cork, Cork, Ireland

Background In Ireland, more Diabetes Nurse Specialists (DNSs) have been introduced into the community as part of a national programme to standardise and improve diabetes care. DNSs support the delivery of a new model of care whereby uncomplicated type 2 diabetes (T2DM) is managed in primary care, and complicated T2DM is managed between primary and secondary care. Historically diabetes care in Ireland has often been delivered in an unstructured way, lacking integration between primary and secondary care. Given this context we wanted to understand the challenges faced by community-based DNSs in delivering a standardised service.Methods We purposively sampled DNSs from community-based respondents to a national survey (n=25) according to four administrative regions of the national health service.  We conducted focus groups and interviews using a semi-structured topic guide. Interviews were digitally recorded and transcribed into NVivo V.11 software for coding and analysis. Data analysis is on-going using thematic analysis.Results Sixteen DNSs participated in 2 focus groups, and 8 interviews. Preliminary analysis suggested elements of the role presented a challenge. As DNSs require their Collaborative Practice Agreement to be signed off by each GP using their service, they were currently unable to prescribe

in the community.   Despite describing this as ‘frustrating’, DNSs suggested that prescribing could remove opportunities for relationship-building with GPs through discussion of medications.  In the community, DNSs lacked the safety net of the hospital team to check things with,  and had to work more autonomously, described as ‘daunting’. Role understanding by other staff was another challenge; DNSs felt managers did not understand how the community role should work, which created difficulty when negotiating aspects of the role, including flexible working hours. The lack of a shared record between settings meant patient information from hospital appointments was not readily accessible by DNSs at GP practices and vice versa. This made patient follow-up and case discussion difficult when DNSs were off-site. The absence of administrative support in the role, considered ‘crucial’, was also highlighted. Further interviews with community DNSs are ongoing.Conclusion Community-based DNSs faced challenges presented by aspects of their role, their relationship with other staff, their work environment, and the available organisational infrastructure and resources. Although recent policy reforms in Ireland have focused on improving the integrated management of diabetes in the community, findings from this study suggest DNSs may need to be better supported to ensure delivery of a standardised model of diabetes care.Keywords: integrated care, diabetes nurse specialist, diabetes

Ethnicity 1

OP17

ETHNIC DIFFERENCES IN ILL HEALTH AND IN SOCIOECONOMIC INEQUALITIES IN HEALTH: POPULATION STUDY USING 2011 SCOTTISH CENSUS

M Allik1, D Brown2, R Dundas2, AH Leyland2

1Urban Big Data Centre, University of Glasgow, Glasgow, UK2MRC/CSO SPHSU, University of Glasgow, Glasgow, UK

Background Much has been written about high rates of poor health and health inequalities in Scotland, increasingly it is shown how these outcomes vary by ethnicity. Scottish Government has made a policy commitment to understand and address inequalities in health among minority ethnic groups. This study contributes to this by comparing health outcomes and socioeconomic inequalities in health across ethnicitiesMethods Two self-reported health measures, poor general health and limiting long-term illness (LLTI), by 5-year age groups, ethnicity and area (Datazones; population mean=815, sd=275) from the 2011 Scottish Census were examined. Ethnicity was self-reported and grouped into 9 main categories. This paper focused on the 5 largest groups: White Scottish (N=4,445,678), White British (N=417,109), White Irish (N=54,090), Other White (N=167,530) and Asian (N=140,678) and ages 0-64. Deprivation was measured using Census based indices and SIMD. Age standardized rates of ill health per 1000 people were calculated for these groups and by deprivation quintiles. Inequalities by area deprivation were measured using the slope index of inequality (SII).Results For ages 0-64 the standardized rates are lowest for Other Whites and highest for White Scottish for both measures of ill health (LLTI rates respectively 89.1 and 134.9). Differences are greatest for younger adults, LLTI rate for Other Whites aged 15-29 is 32.4, but for White Scottish 71.7, for ages 30-44 these rates are 63.6 and 124.2 respectively. On average White Scottish had poorer health than White British and Irish, both of who are also least likely to live in the most deprived areas. For ages 0-64 inequalities in health were highest for White Scottish (for LLTI the SII=164.4, 95% CI= 163.1-165.7), but not much lower for White British (SII= 150.8, CI=146.2-155.4) and Irish (SII=145.2, CI=133.6-156.8). Inequalities were much lower among Asians (SII=74.2, CI=64.1-84.1) and among Other Whites (SII=59.3, CI=51.3-66.9). Differences in health inequalities between ethnicities were greatest for ages 30-44. Ill health and inequalities among Asians increased more rapidly for older ages and were similar to White Scottish for those 60+.

Results White Scottish have poorer health compared to other ethnicities, but are also more likely to live in deprived areas compared to White British and Irish. Deprivation affects the health of all ethnicities, but much less so for some groups. Health inequalities are highest among White Scottish, similar for British and Irish, and significantly lower for Asians and Other Whites. This suggests that cultural and/or behavioural factors may reduce the effect of material deprivation on health.Keywords: health inequalities, deprivation, ethnicity

OP18

ETHNIC VARIATIONS IN INFANT MORTALITY IN ENGLAND AND WALES 2006-2012: A NATIONAL COHORT STUDY USING ROUTINELY COLLECTED DATA

Y Li, M Quigley, J Hollowell1Policy Research Unit in Maternal Health and Care, National Perinatal Epidemiology Unit, Nuffield Department of Population Health, University of Oxford, Oxford, United Kingdom

Background Marked ethnic variations in infant mortality have been documented in England and Wales. The reasons why certain ethnic groups have higher rates remain generally unknown. We aimed to evaluate ethnic variations in infant mortality, adjusting for deprivation and mother’s migration status and to explore the contribution of preterm birth to these variations.Methods Routinely collected birth and death registration data on all singleton live births (gestational age≥22w) in England and Wales, 2006-2012 were linked to other routine data on infant’s ethnicity and gestational age. We analysed infant mortality by ethnic group using logistic regression in STATA to adjust for maternal sociodemographic characteristics (age, area deprivation, migration status and marital status/registration type) and gestational age.Results In the 4,634,932 births analysed, crude infant mortality rates were highest in Pakistani, Black Caribbean, Black African, and Bangladeshi infants (6.92, 6.00, 5.17 and 4.40 per 1000 live births, respectively vs. 2.87 in White British infants). Adjustment for maternal sociodemographic characteristics changed the results little. Further adjustment for gestation significantly attenuated the risk in Black African (OR 1.17, 95% CI 1.06-1.29) and Caribbean (OR 1.02, 95% CI 0.89-1.17) but not in Pakistani (OR 2.32, 95% CI 2.15-2.50), Bangladeshi (OR 1.47, 95% CI 1.28-1.69), and Indian infants (OR 1.24, 95% CI 1.11-1.38).The association between ethnicity and infant mortality differed significantly between term and preterm infants (p<0.001 for interaction). In term infants, all three South Asian groups had a higher risk not explained by maternal characteristics or gestation. In preterm infants, adjustment for gestational age (<28, 28-31, 32-33, 34-36 weeks) fully explained the higher risks in Black African and Caribbean infants, but not in Pakistani and Bangladeshi infants. A sensitivity analysis excluding deaths due to congenital anomalies did not fully explain the excess risk in Pakistani infants.Conclusion South Asian and Black infants have higher infant mortality compared with White British infants which is not explained by deprivation, migration status or other maternal characteristics. A higher proportion of infants born at younger gestational age appears to explain the increased risks in preterm Black African and Caribbean infants but not in the South Asian groups.This large national study provides insight into ethnic inequalities in early life to inform policy, practice, and future research. Findings suggest that strategies targeting the prevention and management of preterm birth in Black African and Caribbean mothers and babies would help to reduce ethnic inequalities in infant mortality. Different strategies may be required in South Asian groups.Keywords: infant mortality, ethnic variations, health inequalities

OP19

ROMA POPULATIONS AND HEALTH INEQUALITIES: A SYSTEMATIC SCOPING REVIEW OF MULTIPLE INTERSECTING DETERMINANTS 

LC Orton1, R Anderson de Cuevas1

1Public Health & Policy, University of Liverpool, Liverpool, UK

Background There is growing evidence that Roma populations in Europe experience significantly poorer health - including higher rates of communicable and non-communicable diseases, poorer self-rated, child and maternal health - and shorter life expectancy than majority populations. Health outcomes are often worse than for others in similar social positions, suggesting the differences are not simply the effect of poverty. Roma women are thought to be worst affected. However, data pertaining to the multiple social inequalities affecting Roma populations, their interaction and cumulative effect on health over time, have yet to be explored. The aim of this paper is to critically review the state and shape of the research evidence on the multiple intersecting determinants of health for Roma populations in Europe.Methods Studies were identified by searching four bibliographic databases (MEDLINE, Science Citation Index, Social Sciences Citation Index and Scopus), screening reference lists, consulting key informants and searching organisational websites. Any empirical studies (quantitative and qualitative) that explored more than one influencing factor in the pathway to poor relative health for Roma populations residing in the 53 countries of the WHO European region were included. Two reviewers carried out screening, data extraction and quality appraisal; disagreements were discussed by the wider team. This was followed by a narrative synthesis to explore the interactive effects of multiple axes of inequality.Results From 2,043 bibliographic records, 37 studies met our inclusion criteria.  The median number of influencing factors explored through qualitative and quantitative methodologies was 2 (range 1 - 5) and 3 (range 1 – 11) respectively. Quantitative studies focus most strongly on factors that are well-defined and relatively easily turned into health indicators using available datasets, namely: poverty (10 studies) and other indicators of socio-economic status including education (13 studies), housing (9 studies) and employment (9 studies). Very few studies unpick interactive effects; instead testing each factor’s contribution to poor health independently. On the other hand, qualitative studies tended to explore “Roma culture” (11 studies) and the impact of discrimination/racism on health behaviour (10 studies). There is a lack of gender-disaggregated data and little analysis of the impact of social policy on the health of Roma populations.Conclusion Evidence for associations between determinants and health outcomes among European Roma populations is patchy.Research needs to pay more attention to the health impact of social policies that have the potential to reinforce (as well as mitigate) the exclusion of Roma populations and the disadvantages experienced by women. Keywords: systematic review, Roma, health inequalities

OP20

SHAME: EXPERIENCES OF FOOD AND POVERTY AMONG WHITE BRITISH AND PAKISTANI LOW-INCOME WOMEN

MS Power1, N Small2, KE Pickett1

1Department of Health Sciences, University of York, York, UK2Faculty of Health Studies, University of Bradford, Bradford, UK

Background Food bank use has increased sharply in the UK since 2010, however, qualitative research with food aid providers has found very low use of food banks by members of Bradford’s Pakistani community, a large minority community in Bradford. Despite this, quantitative research

has shown a substantial prevalence of food insecurity among Pakistani households (10%), albeit a lower prevalence than in White British (WB) households (18%).The study aims to understand how low-income women understand and experience food insecurity, and how this differs between WB and Pakistani women; and to explore access to and experiences of food aid among WB and Pakistani low-income women.Methods In light of potential recruitment difficulties and language and capacity restrictions, focus groups were used. With the assistance of the Better Start Bradford Innovation Hub, we identified existing group activities in Bradford where it would be appropriate to hold focus groups. Members of these groups were invited to participate in the study. Four semi-structured focus groups (n=16) were conducted in three low-income wards in Bradford. A three-stage analysis process was used and the data was analysed thematically.Results The sample included eight Pakistani women and eight WB women living in three deprived wards in Bradford. Seven Pakistani women were married to men in employment. Only one Pakistani woman was in employment. Six WB women were married or cohabiting; one was employed; four had partners who were employed and three were solely reliant on social security.Only one Pakistani women reported struggling to afford food, compared with five WB women. However, only three of the latter had used food aid; no Pakistani women had used food aid. All women described a sense of shame around not being to provide enough food for family members, particularly children. This was felt most acutely among Pakistani women. It was explained that food insecurity was experienced but concealed among Bradford’s Pakistani community; support with food was sought not from food aid but from immediate family members and, occasionally, the wider Pakistani community.Conclusion This is small scale study of low income women in one city, however it does suggest that Pakistani and WB women in Bradford experience food insecurity differently, with the latter much more likely to use food banks. Shame around food insecurity may not only deter low-income women from accessing food aid but cause food insecurity to be concealed entirely, this is most pronounced in the Pakistani community.Keywords: Ethnicity, food insecurity, poverty, stigma, shame

Children and Mental Health

OP21

ASSESSING THE IMPACT OF CHILDHOOD SOCIO-ECONOMIC CONDITIONS ON CHILD MENTAL HEALTH: FINDINGS FROM THE WIRRAL CHILD HEALTH AND DEVELOPMENT STUDY 

C Rutherford1, J Hill2, H Sharp3, D Taylor-Robinson1

1Department of Public Health and Policy, University of Liverpool, Liverpool, England2School of Psychology & Clinical Language Sciences, University of Reading, Reading, England3Department of Psychological Sciences, University of Liverpool, Liverpool, England

Background There is increasing recognition that the early-years of a child’s life are critical periods in which social inequalities in mental health can develop. A systematic review has shown that a socio-economic gradient exists in child and adolescent (5-18 years) mental health outcomes, but few studies have assessed the early emergence of inequalities in child mental health. The aim of this study was to assess the impact of childhood SECs on child mental health outcomes in preschool children and to identify potentially modifiable early life factors.Methods Analysis of the Wirral Child Health and Development Study, a longitudinal study of the early origins of child mental health, following 1233 children from a wide range of socio-economic backgrounds on the Wirral, North West England. The mental health outcome measure was parent reported child externalising problems, as measured by the Child Behaviour Checklist at 4.5 years. Maternal education at 20-weeks gestation, a measure of SECs in pregnancy, was the main

exposure. We assessed the association of SECs with child mental health outcomes (log CBCL T-score) in sequential linear models adjusting for pre- and post-natal risk factors (e.g. parental mental health and perinatal factors). Exponentiated coefficients and 95% confidence intervals are presented as geometric means, using R (version X). Results Children of mothers in the most educated quintile scored 11.0% (95% CI 3.5-18) lower for externalising problems compared to children of mothers in the least educated quintile, after adjusting for sex and ethnicity. Male sex, prenatal maternal depression at 20 weeks, and postnatal depression at 3.5 years were independently associated with an increased risk of child mental health problems. Adjusting for prenatal maternal depression attenuated the association of maternal education with child mental health to 9% (95%CI 1-16) comparing the most educated to the least). Further adjustment for perinatal factors and post-natal mental health did little to further attenuate the association of SECs with child mental health. Conclusion In a study of early child mental health we found that social disadvantage is associated with worse child mental health outcomes at age 5. This was partially explained by adjusting for maternal mental health during pregnancy, but not postnatal measures of maternal mental health. Policies supporting maternal mental health in pregnancy are important to address the early emergence of inequalities in child mental health. Parent-reported child mental health, and cohort attrition are limitations of this study. Keywords: mental health, child, inequalities

OP22

UNDERSTANDING SOCIAL INEQUALITIES IN BEING BULLIED IN CHILDHOOD: FINDINGS FROM THE UK MILLENNIUM COHORT STUDY

M Campbell1, V Straatmann1, D Taylor-Robinson1

1Department of Public Health and Policy, University of Liverpool, Liverpool , UK

Background Childhood bullying is an important policy concern. Nearly half of victims of bullying report thoughts about suicide and self-harm with negative impacts extending across the life-course.  Being bullied in childhood is common, socially patterned, however, factors explaining social inequalities in being bullied are unclear. Using a contemporary United Kingdom (U.K.) birth cohort, we aimed to assess and explain social inequalities in the risk of being bullied.Methods Analysis of the U.K. Millennium Cohort Study using a sample of 12,706 children surveyed at four sweeps (aged nine months, three, five and seven years). The main outcome was a binary, child-reported measure of being bullied at age 7. Household income quintile at birth was the main measure of socio-economic conditions. Relative risk (RR) and 95% confidence intervals (95%CI) for being bullied were estimated using Poisson regression, by household income quintile. Sequential models adjusted for risk factors for being bullied, including individual (e.g. emotional resilience, health status including obesity), parental (e.g. maternal mental health and discipline) and peer relationship (e.g. friends) factors.  Analysis used Stata/SE with svy commands to account for the sampling design and attrition.  Our sensitivity analysis will use parent and teacher reported outcome measures.Results By age seven, 48.7% (95%CI 47.5%-49.9% [n=6,183]) of children self-reported being bullied.  There was a social gradient; 53.4% (95%CI 50.6%-56.1%) in the lowest income quintile reported having been bullied, compared to 43.9% (95%CI 41.5%-46.4%) in the highest (RR 1.21[95%CI 1.10-1.33]).  Male sex, young maternal age, higher child BMI and Strengths and Difficulties Questionnaire (SDQ) scores and worse maternal mental health were independently associated with an increased risk of being bullied, whilst having a good friend was protective. Controlling for these factors in the final model attenuated the RR for being bullied for children in the lowest income quintile, compared to the highest, rendering the association non-significant (RR 1.08[95%CI 0.98-1.19]).

Conclusion Using a nationally representative cohort, about half of seven year olds reported being bullied, with a greater risk in children from the poorest homes. Increased risk was largely explained by social differences in other risk factors for bullying including friends, maternal mental health, and individual factors including emotional resilience and BMI. Interventions addressing these risks and promoting protective factors are likely to reduce social inequalities in being bullied, improving mental health outcomes for the most vulnerable U.K. children. Future research should investigate inequalities in being bullied in adolescents. The self-reported primary outcome is the main study limitation.Keywords: Bullying; inequalities; child health

OP23

MODELLING THE POTENTIAL FOR PARENTING INTERVENTIONS TO REDUCE INEQUALITIES AND POPULATION PREVALENCE OF CHILDREN’S MENTAL HEALTH PROBLEMS: EVIDENCE FROM THE MILLENNIUM COHORT STUDY (MCS)

S Hope1, A Pearce1, M Cortina-Borja1, C Chittleborough2, C Law1

1Population, Policy and Practice Programme, UCL Great Ormond Street Institute of Child Health, London, United Kingdom2School of Public Health, The University of Adelaide, Adelaide, Australia

Background Parenting programmes aim to support parents’ skills and confidence, improving parenting and, in turn, children’s mental health. Thus they have the potential to reduce population prevalence (and inequalities) in child mental health problems. We modelled the potential population impact of scale-up of parenting interventions in a national cohort. Based on review evidence, we simulated interventions with effect sizes of 0.9SD (intensive) and 0.4SD (standard). Methods We used data from the UK Millennium Cohort Study, following 18,000 children born 2000-2002. Parenting was assessed by Child-Parent Relationship Scale (CPRS: Short-Form) score when the child was 3 years. Child mental health problems (CMHP) at 5 years were assessed by Strengths and Difficulties Questionnaire (SDQ) total score, dichotomised using an established cut-off. Socio-economic circumstances were represented by mother’s highest academic qualifications (<GCSEs A-C; GCSEs A-C and above) at 9 months.Predicted probabilities of CMHP were estimated, fitting marginal structural models to examine the mediating effect of parenting, accounting for confounding using inverse-probability-treatment-weights. Inequalities were assessed with Risk Ratios (RR [95% CIs]). A series of intervention scenarios were simulated by re-estimating predicted probabilities after modifying the CPRS score (to reflect effectiveness) for eligible children (targeting).We analysed data from 14,540 children, using Stata 13.1. Survey weights and multiple imputation addressed missing data.Results Overall prevalence of CMHP at 5 years was 10.8%, and children of mothers with low educational attainment had a greater risk of CMHP (RR=2.46[95%CI:2.24-2.70]). Focusing on three parenting intervention scenarios, we showed changes in CMHP prevalence and inequalities compared to those observed:  1. Universal (standard support to all parents): prevalence change -2.5%; inequality RR=2.56[2.30-2.85]; 2. Targeted (intensive support to families receiving means-tested benefits): prevalence change -1.4%; inequality RR=2.11[1.91-2.34]; 3. Progressive universal (intensive support for families receiving benefits and standard support for others): prevalence change -3.2%; inequality RR=2.35[2.10-2.62].Conclusion Large inequalities in CMHP were apparent by age 5 years. In simulated scenarios, inequality was reduced through an intervention that explicitly set out to provide intensive support to disadvantaged families. In contrast, reductions in overall prevalence were more likely to be achieved by universal interventions. A progressive universal approach (combining intensive support for disadvantaged families with standard support for others) led to a reduction in both population

prevalence and inequality. Nevertheless, in all intervention scenarios, inequalities in CMHP remained strong. These results suggest that parenting interventions may contribute to a reduction in CMHP inequalities, particularly when including targeted support for disadvantaged families.Keywords: inequalities, child health

OP24

ELICITING CHILDREN’S PERCEPTIONS AND EXPERIENCES OF HEALTH, DIET, PHYSICAL ACTIVITY AND WEIGHT THROUGH THE DRAW, WRITE AND TELL TECHNIQUE 

M Murphy1, W Robertson1, R Johnson1, FK Boardman1

1Warwick Medical School, University of Warwick, Coventry, UK

Background Existing research exploring determinants of childhood obesity tends to overlook the experiential, contextual and cultural basis of such determinants, and lacks methodological features that support children’s own framing of issues around obesity. This study aimed to explore children’s perceptions and experiences of health, diet, physical activity and weight, with a focus on culture-specific contextual factors that may give rise to childhood obesity, in an ethnically diverse sample.Methods Children aged 9-10 years were recruited from primary schools in Coventry, UK, with purposive sampling of schools to ensure an ethnically diverse sample. Data were collected through a ‘draw, write and tell technique’ using semi-structured one-to-one interviews at school. Images, text and transcripts were analysed using a thematic framework approach, with themes reviewed and mapped against existing ecological models of health behaviours.Results To date, six of fifteen invited schools agreed to participate, with 27 children (girls n=16) successfully recruited from three schools, all of which lived in areas of high deprivation, with 81% from minority ethnic groups. Recruitment is on-going to allow for data collection with children from ethnic groups currently under-represented in the sample. Preliminary analysis has identified seven over-arching themes: knowledge, understanding, beliefs, values, motivations, intentions and group-specific experiences. Children had a grasp of the fundamental elements of a healthy diet, physical activity and maintaining a healthy weight but understandings were simplistic and often contradictory. Children viewed health as a personal responsibility, attributed to will-power, but often described external factors affecting their own ability to be healthy e.g. social pressures to indulge and permissive parenting. Positive and negative personal characteristics were assigned to others on the basis of healthfulness and weight. Key motivations for health behaviours were avoidance of the negative social consequences of being overweight, personal success and happiness. Boys and girls differed in their perception of a healthy aesthetic, but culture-specific experiences were largely absent from the data. Conclusion The findings suggest that childhood obesity interventions could focus on the existing assets that children describe, such as friendships and supportive family environments; emphasising health benefits framed by child-based motivations e.g. personal success and happiness. The lack of culture-specific experiences in children’s accounts is inconsistent with other research in ethnically diverse child populations, which may be attributable to the methods used or lack of representation of some ethnic groups. Use of the draw, write and tell technique allowed the research to be child-led, supporting children’s own framing of issues. Keywords: Child Weight Ethnicity

Health Inequalities 2

OP25

Note: Presentation moved to Gender and Family session on Friday

EXAMINING THE SITUATED HEALTH AND WELLBEING CHALLENGES OF URBAN FOOD AID PROVISION: A QUALITATIVE PARTICIPATORY STUDY WITH HEALTHCARE PROFESSIONALS IN LONDON

C Thompson1, D Smith2, S Cummins1

1SEHR, London School of Hygiene and Tropical Medicine, London, UK2Geography and Environment, University of Southampton, Southampton, UK

Background Food insecurity is negatively associated with health.  Therefore, local food aid provision is crucial to helping protect the health and wellbeing of vulnerable groups.  At present, the politics and logistics of this provision are only partially understood and under-theorised.  In London, one of the most expensive and equal cities in the world, food insecurity is intensifying.  Long-term provision of food aid presents challenges for health maintenance and promotion that are frequently overshadowed by more immediate problems at the local level.  This qualitative study examines how London healthcare professionals encounter and seek to address these challenges.Methods A snowball approach was used to recruit 38 healthcare professionals and food aid coordinators through contacts with participating London food aid organisations.  Semi-structured interviews were conducted with all participants.  Additionally, observations at a variety of food aid distribution events and stakeholder strategy meetings were used to explore narratives around organisation and health.  A critical grounded theory approach to analysis served to identify and situate both the material (physical and mental health) and social (structural inequalities and precarity) issues.Results Health and wellbeing challenges occur at both the frontline and organisational levels.  Participants explained that the mechanisms of addressing (dietary) health via food aid referral are not straightforward.  Household interaction with food aid typically occurs during episodes of crisis, which prompt them to turn to healthcare professionals.  Food insecurity is rarely the ‘presenting problem’ and is embedded within a suite of stressors.  Frontline healthcare professionals have to deconstruct complex accounts of poverty and find ways to intervene.  Offering food aid referral is just one way of doing this.  At the organisational level, a culture of precarity makes it extremely difficult to provide consistently nutritious food aid that caters to a reasonable range of medical and religious dietary restrictions.  Fluctuating sources of funding, food donation and staffing are obstacles to the implementation of dietary guidelines and the monitoring of nutritional content.Conclusion Community food aid is undermined by precarity and a lack of structural support, making health and wellbeing concerns difficult to prioritise.  Healthcare professionals and food aid coordinators are responding to this in imaginative ways, such as pooling resources, sharing expertise and best practice, and strategizing across sectors and communities.Keywords: Food insecurity; food aid; qualitative

OP26

FAMILY INCOME DYNAMICS AND SUBSEQUENT RISK FOR INTERPERSONAL AND SELF-DIRECTED VIOLENCE: A NESTED CASE-CONTROL STUDY

PLH Mok1, S Antonsen2,3, CB Pedersen2,3, MJ Carr1, N Kapur1, J Nazroo4, RT Webb1

1Centre for Mental Health and Safety, University of Manchester, Manchester, UK2Centre for Integrated Register-based Research, Aarhus University, Aarhus, Denmark3National Centre for Register-Based Research, Aarhus University, Aarhus, Denmark4Cathie Marsh Institute for Social Research, University of Manchester, Manchester, UK

Background Childhood poverty is associated with later elevated risks for self-directed and interpersonal violence. However, changes in parental income during upbringing on the risk of these outcomes remain unclear. We aimed to investigate fluctuations in parental income levels during

childhood and subsequent risk of self-harm and violent criminality; in particular, to examine the associations by: 1) parental income in the year of birth, in early and middle childhood, and adolescence; 2) time spent in financially disadvantaged versus affluent conditions; 3) changes in parental income between the year of birth and age 15 years.Methods Using interlinked Danish national registers, we constructed a nested case-control study of all first registered episodes of self-harm (N=16,915) and all first violent crime convictions (N=19,579) during 1997-2012 at ages 15-30 years. Twenty five unaffected controls were matched to each case on sex and age using incidence density sampling. Parental income was assessed in the year of birth, and at ages 5, 10 and 15 years. Incidence rate ratios (IRRs) were estimated using conditional logistic regression in Stata 13.1.Results For both outcomes, the lower the income the higher the risk. The associations were stronger for violent offending than for self-harm, with those from the lowest income quintile showing disproportionately higher risks. The longer a child lived in poorer circumstances the higher the risk, and vice versa for time spent in more affluent conditions. For example, compared with those whose parental income was consistently at the highest quintile at all 4 age points, those whose parental income remained in the lowest quintile were at 7 and 10 times elevated risk for self-harm (IRR=6.9, 95% CI 6.4-7.4) and violent offending (IRR=10.3, 9.6-11.0), respectively.  Compared with parental income being in the highest quintile at birth and also at age 15 years, all other income trajectories between these two ages were associated with elevated risks for both outcomes. In general, however, regardless of the parental income at birth, being upwardly mobile was associated with smaller elevations in risk compared with being downwardly mobile.Conclusion Family income dynamics are strongly linked with later risks of self-harm and violent criminality. Although income is a marker for an array of familial circumstances and causal inferences could not be drawn, our findings suggest that reducing poverty at any stage during children’s development and promoting upward socio-economic mobility could ameliorate such risks, and have a marked cumulative societal benefit.     Keywords: Family income; Poverty; Violence

OP27

SOCIOECONOMIC POSITION AND MORTALITY FROM BRAIN TUMOUR – A SWEDISH POPULATION-BASED STUDY

A.R. Khanolkar1, M. Talbäck2, R. Ljung2

1GOS Institute of Child Health, University College London, London, UK 2Institute of Environmental Medicine, Karolinska Institutet, Stockholm, Sweden

Background Socioeconomic disparities in mortality from various cancers are well documented with patients from lower socioeconomic background having an increased risk for mortality. However, similar evidence on differences in mortality from tumours of the central nervous system is both limited and conflicting. We investigated associations between socioeconomic factors (education, income and marital status) and mortality after a brain tumour diagnosis.Methods The study included all patients diagnosed with a primary brain tumour in Sweden between 1993-2010 as reported to the national cancer register and were followed-up until 31st December 2015. Data on education, disposable income and marital status were obtained via linkage with national registers. We used flexible parametric models with restricted cubic splines to estimate the excess hazard ratio [EHR] (the analogue of relative survival) by socioeconomic factors for glioma, glioblastoma and meningioma. Models were adjusted for age at diagnosis, tumour location, healthcare region and country of birth, and run separately for men and women. Results 6,075 men and 7,831 women developed a brain tumour during the study period. 4,197 (69%) men and 3,370 (43%) women died by the end of follow-up. Men and women with primary education had increased mortality from glioma (EHR, 1.13, 95% CI 1.04-1.24, and 1.11, 1.00-1.24)

and glioblastoma (EHR 1.20, 1.07-1.35 and 1.14, 1.00-1.31 respectively) compared to those with university education. Men in the lowest quartile of income had 29% and 25% higher mortality from glioma and glioblastoma compared to those in the highest income quartile (EHR 1.29, 1.17-1.43 and 1.25, 1.10-1.42 respectively). Women in the lowest quartile of income had higher mortality from meningioma than those in the highest quartile (EHR 3.63, 1.76-7.52). Being single (EHR, 1.15, 1.04-1.26 and 1.21, 1.06-1.38 for men and women respectively) and widowed (EHR, 1.30, 1.08-1.58 and 1.14, 1.00-1.30 for men and women respectively) was associated with increased mortality from glioma. Similarly, being single was associated with increased mortality from meningioma in men (EHR 2.49, 1.42-4.36) and women (EHR 2.10, 1.18-3.73).  Conclusion While lower education and low income are associated with increased mortality from glioma in men, only lower education is associated with increased mortality from glioma in women. Low income was associated with increased mortality from meningioma in women only. Being single or widowed were associated with increased mortality from glioma and meningioma in both sexes. These disparities were observed despite access to a universal healthcare system. Earlier detection in individuals from higher socioeconomic groups could be a potential explanation. Keywords: Brain tumour, mortality, health inequalities

OP28

WILL GOVERNMENT TARGETS TO INCREASE PHYSICAL ACTIVITY IN CHILDREN REDUCE SOCIO-ECONOMIC INEQUALITIES IN CHILDHOOD OVERWEIGHT? A POLICY SIMULATION IN THE UK MILLENNIUM COHORT STUDY (MCS)

A Pearce1, S Hope1, L Griffiths1, M Cortina-Borja1, C Chittleborough2, C Law1

1Population, Policy and Practice, UCL GOS Institute of Child Health, London, UK2School of Public Health, University of Adelaide, Adelaide, Australia

Background The 2016 Childhood Obesity Plan included a target of 60 minutes(m) daily moderate-vigorous physical activity (dMVPA). In the UK Millennium Cohort Study (~18000 children born 2000-2002) we examined how the prevalence and social distribution of childhood overweight/obesity might change if physical activity (PA) interventions aimed at this target were rolled-out under different scenarios of eligibility, uptake and effectiveness.Methods dMVPA at 7 years(y) was captured with accelerometers, over one week, adjusted for total wear-time. At 11y, children were classified as healthy, overweight (including obese), using measured heights and weights and International Obesity Task Force cut-offs. Socio-economic circumstances (SECs) were represented by maternal education (GCSEs/None; A-Levels+).Predicted probabilities of overweight were estimated by fitting marginal structural models, adjusting for dMVPA, and accounting for baseline and intermediate confounding with inverse-probability-treatment-weights. Inequalities were assessed with risk ratios (RRs) and risk differences (RDs). Intervention scenarios were simulated by re-estimating predicted probabilities after modifying the dMVPA variable by a given amount (reflecting effectiveness), for eligible children only (for targeted interventions), and with random sampling (where uptake <100%).Analyses included 6344 children in the MCS PA study. Survey weights and multiple imputation addressed sampling design, attrition and item missingness. Sensitivity analyses using complete cases and alternative adiposity (fat mass) and SECs (income) measures produced similar conclusions.Results 28.4% children were overweight and those from lower SECs were at greater risk (RR: 1.36[95% CI: 1.25-1.49]; RD: 9.6%[9.2-9.9]). 49% achieved 60m dMVPA, although mean dMVPA was greater in low SECs groups (64.5m vs. 61.8m).  Simulations showed that, with an additional 30m dMVPA for all children, 96% would achieve 60m dMVPA. Prevalence of overweight would decline significantly (to 22.4%), relative inequality would increase (RR 1.40[1.27-1.55]), whereas absolute inequality would fall slightly (RD 8.3%[8.0-8.6]).

More realistic simulations, using effect sizes from meta-analyses and assuming 77% uptake, were less promising. A universal 4.6m increase in dMVPA would lead to negligible reductions in overweight (27.7%), with no change in inequality (RR 1.37[1.26-1.49]; RD 9.5%[9.2-9.8]). Intensive, targeted interventions achieving an additional 9.6m dMVPA in deprived neighbourhoods would not alter population-level prevalence of overweight (28.1%) or inequality (RR 1.35[1.24-1.47]; RD 9.1%[8.7-9.4]).Conclusion A simulation of the government’s ambitious PA target (in nationally-representative, objective data) achieved only a moderate reduction in population prevalence of childhood overweight and a slight narrowing of absolute (but not relative) inequalities. More realistic scenarios (modelling effects of trialled interventions) did not reduce prevalence of overweight and, even when targeted at deprived areas, inequalities remained.Keywords: child health; policies; inequalities

Lifecourse adult health

OP29

ASSOCIATIONS BETWEEN CHILD MALTREATMENT AND ADULT PHYSICAL FUNCTIONING IN A PROSPECTIVE BRITISH BIRTH COHORT

G Archer1, S Pinto Pereira1, C Power1

1Population, Policy and Practice, UCL Great Ormond Street Institute of Child Health, London, UK

Background Child maltreatment (abuse and neglect) has established associations with mental health; however, little is known about its relationship with physical functioning as few studies have been undertaken. Physical functioning in adulthood is an important outcome to consider, as it is strongly associated with an individual’s ability to work, and future disability, dependency, hospitalisations and mortality. We aimed to establish whether maltreatment was associated with physical functioning, independent from other early-life adversities.Methods Using data from the 1958 British birth cohort (n=8150), we examined associations between child neglect and physical, psychological, witnessing and sexual abuse with physical functioning at age 50. Poor physical functioning was defined by those scoring <65 on the Short-Form 36 (SF-36) Physical Functioning scale. We also examined two secondary outcomes – mental health and self-reported health at age 50. Associations between each maltreatment and outcome were assessed using logistic regression with and without adjustment for covariates, including childhood social class, birthweight, childhood health, parental chronic illness, and parental education.Results 23% of participants reported at least one type of maltreatment and 12% were identified with poor physical functioning. Neglect, psychological, and sexual abuse were associated with poor physical functioning independent of all covariates and other maltreatments: ORadj1.55 (95% CI 1.24-1.93), 1.49 (1.17-1.88) and 2.56 (1.66-3.96), respectively. Odds of poor physical functioning increased with multiple types of maltreatment; ORadj ranged from 1.49 (1.23-1.82) for a single type to 2.09 (1.53-2.87) for those reporting >3 types of maltreatment, compared to those with none. Associations of comparable magnitude were observed for mental and self-reported health outcomes; e.g. ORadj varied between 1.42-1.55 for neglect across the three adult outcomes.Conclusion Child neglect, psychological, and sexual abuse were associated with poor physical functioning at 50yrs, with accumulating risk for those with multiple types of maltreatment. To our knowledge, we are the first to demonstrate that these associations are independent of numerous early-life adversities, and comparable in magnitude to those observed for mental health and self-rated health. Our findings underscore the importance of the prevention of maltreatment and the alleviation of its ill-effects to promote healthy ageing.Keywords: child maltreatment, physical functioning

OP30

LIFETIME SEVERE AFFECTIVE SYMPTOMS AND SUBSEQUENT MENTAL STATUS: OVER 50 YEARS OF FOLLOW-UP IN THE 1946 BRITISH BIRTH COHORT STUDY

SN James1, N Sharma1, C O’Hare1, D Kuh1, M Richards1

1MRC Unit for Lifelong Health and Ageing , at University College London, London, United Kingdom

Background Affective problems are common, involving one in five individuals across their lifetime. Affective problems increase the risk for subsequent dementia and cognitive impairment, but the nature of this association is not yet clearly understood. Symptom severity is known to raise the risk, but it is unclear whether this risk is cumulative. We investigated prospective associations between lifetime severe affective symptoms and subsequent mental status and cognitive decline in a population-based sample aged 69 years, still mostly free of dementia, controlling for potential confounders. Methods Complete data from1059 study members from the Medical Research Council National Survey of Health and Development (NSHD, the 1946British birth cohort) were analysed. Prospectively-assessed affective symptoms at ages 13-15, 36, 43, 53, 60-64 and 69  were used to generate severe (case-level) binary indicators of affective symptoms at each age which were then summed to give a cumulative score and operationalised as 0 (never case-level), 1 (case-level at only one assessment), or 2 (case-level at two or more assessments). Mental status was assessed using the Addenbrooke's Cognitive Examination (ACE-III) at age 69. Measures of short-term memory, letter search speed and accuracy were assessed at ages 60-64 and 69. Linear regression models were conducted to investigate whether case-level affective symptoms were associated with lower mental status and cognition at age 69, conditional on cognition at age 60-64, and adjusting for gender, childhood cognition, childhood and midlife social class and educational attainment.Results Compared with those with no evidence of affective problems, those with case-level symptoms at only one assessment did not differ in any of the cognitive measures, before and after adjustments. However, those with case-level affective symptoms at two or more assessments had a lower ACE-III score (β=-0.88, 95%CI=-1.66,-0.10), and greater decline in short-term memory (β=-0.56, 95%CI=-1.21,-0.09), search speed (β=-11.11, 95%CI=-19.59,-2.64), and search accuracy (β=-0.71, 95%CI=-1.29,-0.13) between ages 60 and 69, before and after adjustments.Conclusion Overall, our findings show that those with recurrent severe affective problems are more likely to have poorer mental status and faster cognitive decline. One limitation is that there may be additional confounding factors that we have been unable to account for, including the potential for reverse causality. However, this is still one of the largest prospective studies to show the association between lifetime severe affective symptoms and subsequent cognitive function. Further follow-up will determine what proportion with severe recurrent symptoms develop clinical dementia.Keywords: Epidemiology; mental health; cognition.

OP31

CHARACTERISING LONGITUDINAL PATTERNS OF BACK PAIN ACROSS ADULTHOOD IN THE 1946 BRITISH BIRTH COHORT

S Muthuri1, D Kuh1, R Cooper1

1MRC Unit for Lifelong Health and Ageing, University College London, London, UK

Background Back pain is the most common musculoskeletal condition worldwide and a leading cause of years lived with disability. It affects people of all ages and is a major cause of activity limitation. Back pain is often a recurrent or long-term condition; however, most previous longitudinal

studies have short follow-up periods or limited follow-up points over a long timeframe and, have sampled populations from workplace or clinical settings. We aimed to characterise the long-term profiles of back pain over adulthood and examine whether early life factors are associated with these profiles independent of established adult risk factors.Methods Data were drawn from a nationally representative British birth cohort, the MRC National Survey of Health and Development.  For 3271 participants with at least three assessments of back pain, a longitudinal latent variable analysis was conducted on binary outcomes of back pain at ages 31, 36, 43, 53, 60-64 and 68 using Mplus v 6.12. Derived class posterior probabilities were exported to Stata v14.1 and multinomial logistic regression models with probability weighting were used to examine associations between selected childhood factors (including height, body mass index, serious illness, abdominal pain, emotional and conduct problems, father’s occupational class, mother’s education, parental divorce and parental health) and class membership. Adjustments were made for sex and adult risk factors.Results Four profiles of back pain were identified: no or occasional pain (54.6% women: 60.8% men), early-adulthood only (16.6%: 15.5%), mid-adulthood onset (18.0%: 15.7%) and persistent (10.8%: 7.9%) pain. No or occasional pain profile was treated as the referent category in subsequent analyses.In sex-adjusted models, greater height at age 7 was associated with a higher likelihood of early-adulthood only (Relative-Risk Ratio (RRR) per 1cm increase in height=1.03 [95%CI: 1.01-1.05]) and persistent pain (RRR=1.04 [95%CI: 1.02-1.07]).  Other factors associated with increased risk of persistent pain were manual father’s occupational class, severe conduct problems, parental divorce, abdominal pain and average-poor parental health. The latter two factors were also associated with an increased likelihood of mid-adulthood onset pain. In models which included mutual adjustment for other early life factors and adult risk factors, only height at age 7 remained independently associated with early-adulthood only and persistent pain.Conclusion These findings suggest that a range of factors in early life, most importantly height, are risk factors for long-term experience of back pain across adulthood; therefore interventions that aim to alleviate the burden of back pain may need to start early.Keywords: Back pain, early life factors, longitudinal latent class analysis

OP32

ASSOCIATION BETWEEN OPTIMAL GUIDELINE-INDICATED CARE AND SURVIVAL IN PATIENTS WITH ACUTE MYOCARDIAL INFARCTION AND LONG-TERM CONDITIONS: A POPULATION BASED COHORT STUDY

JA Ellis1, O Bebb1, 2, TB Dondo1, M Hall1, CP Gale1, 2

1Leeds Institute of Cardiovascular and Metabolic Medicine, University of Leeds, Leeds, UK2York Teaching Hospital, NHS Foundation Trust, York, UKBackground Acute myocardial infarction (AMI) remains the largest cause of hospitalisation and death in Europe. Long term conditions (LTC) are common in people with AMI and patients with LTCs also experience lower survival. The effect of LTCs on treatment receipt has not been investigated.Methods Myocardial Ischaemia National Audit Project (MINAP) data for 693,388 patients with ST-elevation myocardial infarction (STEMI) and non-ST-elevation myocardial infarction (NSTEMI) were used to investigate associations between LTCs (including diabetes, chronic heart failure, chronic renal failure, COPD, peripheral vascular disease, and cerebrovascular disease) and treatment receipt. Poisson and binomial models were fitted to determine the association between LTCs and receipt of care, with treatment receipt as a count and a binary optimal care vs. sub-optimal care variable (receipt of all eligible care components vs. missing one or more) as the outcome and individual as well as cumulative LTCs as exposures. Model adjustments were informed by directed

acyclic graphs. Flexible parametric survival models were fitted to investigate the interaction of LTCs and optimal care, and the impact on survival.Results Receipt of optimal care was 11.3% (n=78,376), with patients receiving on average 67% of all care opportunities (Mean 0.67, SD 0.23; Median 0.7, IQR 0.5-0.86). In those with a LTC (n=257,929), 11% (n=28,357) received optimal care. Patients with ≥ one LTC received 2.7% fewer treatments compared with no LTC (IRR 0.97, 95% CI 0.97-0.98); larger differences of 7.3% and 6.1% were evident, respectively, in patients with chronic heart failure (0.93, 0.92-0.93) and chronic renal failure (0.94, 0.93-0.94). The odds of receiving suboptimal care were not significantly different in patients with ≥ one LTC than those with no LTC (OR 1.01, 95% CI 0.89,1.13), however the odds of receiving optimal care was significantly lower in chronic heart failure (0.53, 0.46-0.61) and chronic renal failure (0.52, 0.44-0.62) compared to patients without these conditions. There were 204,667 deaths over a mean follow-up time of 2.25 years. The hazard of death in optimally treated patients with ≥ one LTC was double that of those without LTCs (HR 2.18, 95% CI 2.09-2.27) and 2.5-fold in sub-optimally treated patients with ≥ one LTC compared with no LTCs (2.60, 2.52-2.69).Conclusion Patients with LTCs received fewer treatments and were less likely to receive optimal care than those without. Treatment receipt was lowest in chronic heart failure and chronic renal failure. The worst survival was observed in patients with ≥ one LTC receiving sub-optimal care.Keywords: acute myocardial infarction; co-morbidities; survival

Screening

OP33

SYMPTOMATIC VS PRE-SYMPTOMATIC TREATMENT OF TYROSINEMIA TYPE 1 WITH NITISINONE: A SYSTEMATIC REVIEW

C Stinton1, J Geppert1, K Freeman1, A Clarke1, H Fraser1, S Johnson2, P Sutcliffe1, S Taylor-Phillips1

1Warwick Medical School, University of Warwick, Coventry, England2Warwick Library, University of Warwick, Coventry, England

Background Tyrosinemia type 1 is a rare autosomal recessive disorder of amino acid metabolism, affecting approximately 1 in 100,000 people. Without treatment, death is common in childhood. Treatment with nitisinone is associated with reductions in mortality and morbidity; some studies suggest better outcomes when treatment is initiated before the symptoms of the disorder present. An apparent benefit of earlier versus later treatment has been used to support the implementation of newborn screening for Tyrosinemia type 1, but these studies have not been synthesised or quality appraised. We conducted a systematic review to examine if individuals treated following screen detection of the disorder had better outcomes than those treated following symptomatic detection.Methods Standard systematic review methods were used. Embase, Medline, Pre-Medline, and Web of Science were searched. Participants were individuals with Tyrosinemia type 1. We compared people who received nitisinone following screen detection of the disorder (early treatment) with those who received nitisinone after symptomatic presentation (late treatment). Any reported outcomes were considered. Two reviewers independently screened and assessed records, and conducted quality appraisal (using the Quality Assessment Tool for Quantitative Studies). Data extraction was carried out by one reviewer, and checked by another. A narrative synthesis of results was carried out. Post-hoc comparisons were conducted to address confounding factors and applicability concerns.Results The titles/abstracts of 470 unique records were examined, and 50 full texts assessed. Seven articles were included in the review. Study sample sizes ranged from 17 to 148. Methodological quality of the studies was moderate to weak. There was evidence of associations between early treatment with nitisinone and lower rates of death, liver disease and transplantations,

and renal dysfunction. However, posthoc analyses suggested an association between earlier treatment and lower rates of liver transplantation but not mortality (analysis 1) or no differences in outcomes for those treated earlier versus later (analyses 2 and 3).Discussion Evidence from observational studies suggests that treatment with nitisinone initiated during the pre-symptomatic period may be beneficial to people with Tyrosinemia type 1. However, this is subject to bias and applicability concern; the apparent benefits of early treatment may not be present when these issues are addressed. There are several challenges inherent in rare diseases research, including small and heterogeneous populations, lack of appropriate comparator treatments, and limited knowledge about the disease. Our review suggests that alternative research methods or tolerance of lower levels of evidence may be required.Keywords: Systematic review, newborn screening, rare diseases

OP34

UNIVERSAL ANTENATAL CULTURE-BASED SCREENING FOR MATERNAL GROUP B STREPTOCOCCUS (GBS) CARRIAGE TO PREVENT EARLY-ONSET GBS DISEASE: A SYSTEMATIC REVIEW FOR THE UK NATIONAL SCREENING COMMITTEE (NSC)

F Seedat1, J Geppert1, C Stinton1, J Patterson1, CS Brown2, B Tan1, K Freeman1, OA Uthman1, ND McCarthy1, ER Robinson1, SA Johnson 1, H Fraser 1, A Clarke1, SA Taylor-Phillips1

1Division of Health Sciences, The University of Warwick Medical School, Coventry, United Kingdom2Birmingham Public Health Laboratory, Heartlands Hospital, Birmingham, United Kingdom3Bacteria Reference Department, National Infection Service, Public Health England, London, United Kingdom

Background GBS is the leading cause of morbidity and mortality from neonatal sepsis in the UK and patient groups are keen for screening to be implemented. Intrapartum antibiotic prophylaxis (IAP) is offered to women identified with GBS carriage or GBS risk factors to prevent mother to baby transmission and early-onset GBS disease (EOGBS, <7 days). This review on universal GBS screening for pregnant women was undertaken to assist NSC policy decision-making. Review questions were on: epidemiology of GBS, diagnostic accuracy of tests, effectiveness of IAP treatment, and effectiveness of universal GBS screening.Methods Medline, Embase, and Cochrane databases were searched. Grey literature included Public Health England, British Paediatric Surveillance Unit, Audits and Confidential Enquiries, and reference lists of included papers. Participants were pregnant women ≥35 weeks or neonates <7 days. The intervention was selective culture from recto-vaginal swabs at 35-37 weeks followed by IAP treatment for those who were culture positive. Reviewers independently screened records, extracted data, and assessed methodological quality using appropriate tools for each question, including QUADAS-2, Cochrane RoB, and RoBANS tools. Data were synthesised narratively.Results 73 studies were included from 6,287 references. EOGBS in the UK affects 0.57 per 1,000 live births with a case fatality of 5.2%. Twenty-two percent of EOGBS cases and 63% of deaths are in preterm births (many would be ineligible for screening). The natural history of GBS is not known. We estimate that universal GBS screening would be offered to approximately 718,126 pregnant term women annually.Approximately 63,347 (57.7%) women who test positive in labour and 3,282 (8%) who test negative in labour would transmit GBS to their neonates, and approximately 350 (0.5%) neonates would develop EOGBS. We estimate the positive predictive value of selective culture to detect EOGBS to be around 0.2% (350/150,806). More than 150,450 (>99%) women would be false positive and unnecessarily treated. Harms from IAP are unclear but will include antibiotic resistance and other possible health problems. There were no randomised controlled trials of the effectiveness of GBS screening and observational studies gave inconsistent results for EOGBS mortality and morbidity.

Conclusion EOGBS is an important health condition. However, tests are not accurate predictors of maternal GBS transmission, or of EOGBS. Evidence on the harms and benefits of GBS screening is limited. Universal screening is therefore not recommended.Keywords: Screening, health policy, group B Streptococcus

OP35

PRAGMATIC INTEGRATED TRIALS IN SCREENING: A BREAST SCREENING EXAMPLE INVOLVING 1.2 MILLION WOMEN

S Taylor-Phillips1, D Jenkinson1, A Clarke1, M Wallis2

1Warwick Medical School, University of Warwick, Coventry, UK2Cambridge Breast Unit, Cambridge Hospitals NHS Foundation Trust, Cambridge, UK

Background Pragmatic integrated trials use routine data and systems to automate participant selection, randomisation, outcome measurements and/or other elements to deliver large trials at low cost. Here we present an example of a large pragmatic integrated trial in breast cancer screening, and discuss the situations in which these trials are appropriate and acceptable.  Methods The intervention was a simple change to the mammography test for breast screening. Interpreting whether screening mammogram show cancer is a difficult repetitive task that can result in missed cancers and false-positive recalls, and some studies have indicated that missed cancers may increase with time on task (the vigilance decrement). In the UK two readers independently evaluate each batch of mammograms to search for signs of cancer. The intervention was to change the order in which batches of mammograms were presented for interpretation, to reduce the effects of the vigilance decrement.This was evaluated using a multicentre, double-blind, cluster randomized clinical trial at 46 breast screening centres in England for 1 year. Three hundred sixty readers participated. The primary outcome was cancer detection rate; secondary outcomes were rates of recall and disagreements between readers.Results 1 194 147 women who had screening mammograms were randomised (596 642 in the intervention group; 597 505 in the control group), and 10 484 cases (0.88%) of breast cancer were detected. There was no significant difference in cancer detection rate with 5272 cancers (0.88%) detected in the intervention group vs 5212 cancers (0.87%) detected in the control group (difference, 0.01% points; 95% CI, −0.02% to 0.04% points). There was also no difference in recall rate, with 24 681 [4.14%] in intervention and 24 894 [4.17%] in the control group (difference, −0.03% points; 95% CI, −0.10% to 0.04% points). Patterns of cancer detection and recall with time since a break indicated that performance did not decline with time on task as predicted by the vigilance decrement theory. In fact, positive predictive value increased with time on task.Discussion This pragmatic integrated trial in over 1million women cost less than £300k, and demonstrates that in certain circumstances this study design is appropriate. Considerations when planning a pragmatic integrated trial include whether consent is required at the individual or institutional level, whether the relevant outcomes are available in routine data, and the cost of the intervention.Keywords: Screening, Randomised controlled trial, pragmatic, routine data

OP36

DECISIONS ABOUT SMOKING IN PATIENTS SCREENED WITH THE EARLYCDT-LUNG TEST FOR THE EARLY DETECTION OF LUNG CANCER: A QUALITATIVE STUDY

B Young1, K Vedhara1, D Kendrick1, J Robertson2, R das Nair3

1Division of Primary Care, University of Nottingham, Nottingham, UK2Division of Medical Sciences and Graduate Entry Medicine, University of Nottingham, Nottingham, UK3Division of Psychiatry and Applied Psychology, University of Nottingham, Nottingham, UK

Background Routine screening for lung cancer in high risk groups (characterised by age and smoking history) is recommended in the USA and may be implemented elsewhere. It is unclear whether being screened for lung cancer promotes smoking cessation or conversely provides false reassurance and a ‘license to smoke’. This study aimed to understand how experiences of lung cancer screening influence individual decision making about smoking.Methods Thirty one people in Scotland, aged 51-74, took part in semi-structured interviews. They had been screened with the EarlyCDT®-Lung Test (13 positive result; 18 negative) as part of the Early Cancer Detection Test - Lung Cancer Scotland (ECLS) Study and were long-term smokers when screened. Verbatim transcripts were analysed using thematic analysis.Results Interpretations of test results was a key theme, but were often inaccurate, for example a negative result interpreted as an ‘all-clear’ from lung cancer and a positive result as meaning lung cancer will definitely develop. There was no clear pattern in decisions made about smoking in response to positive or negative test results. Emotional response to those interpretations was an overarching theme in decisions about smoking. Emotions included fear, shock, upset, worry, anxiety, guilt, relief, reassurance and indifference. Other themes included changes in perceived risk of smoking-related disease, a feeling that now is the time to stop smoking, interpersonal family influences and avoidance of thoughts about smoking. Of those who had stopped smoking, some cited screening experiences as the sole reason and some cited screening along with other coinciding factors. Cues to change were experienced at different stages of the screening process and not always immediately following a test result. Some participants indicated they underwent screening in order to try and stop smoking. Others expressed little or no desire to stop. In general, lung cancer screening was experienced as a unique opportunity, which sometimes prompted successful or unsuccessful attempts to stop smoking.Conclusion Lung cancer screening can be a ‘teachable moment’ for smoking behaviour change. Emotional responses to test results, which can be misinterpreted, were an important theme but behavioural responses varied according to the individual. Findings should be considered within the context of a group of predominantly life-long smokers undergoing a novel blood screening test, who might already have increased motivation to stop smoking. Lung cancer screening presents an opportunity to engage high risk smokers in cessation support but our findings suggest such support may need to be available flexibly to be most effective.In collaboration with the ECLS study teamKeywords: smoking; behaviour change; cancer screening

Ethnicity & children

OP37

ETHNIC VARIATION IN RISK OF SUDDEN INFANT DEATH SYNDROME AND OTHER UNEXPLAINED INFANT DEATH IN ENGLAND AND WALES 2006-2012: NATIONAL BIRTH COHORT WITH ROUTINE EXPOSURE AND OUTCOME DATA

ME Kroll1, MA Quigley1, N Dattani2, J Hollowell11Policy Research Unit in Maternal Health and Care, National Perinatal Epidemiology Unit, University of Oxford, Oxford, UK2School of Health Sciences, City, University of London, London, UK

Background About 8% of deaths under age 1 year in England and Wales are unexplained, the cause of death being recorded as sudden infant death syndrome (SIDS, 60%) or unascertained (40%). Typically, unexplained infant death (UID) occurs during sleep, perhaps triggered by unsafe sleep environments (prone/side position, soft bedding etc.). Other risk factors include preterm birth, which may increase vulnerability. Incidence of both preterm birth and (in an analysis of data from 2005) SIDS has been relatively high for Black Caribbean babies. We aimed to evaluate recent ethnic variation in UID, test the hypothesis that it might be mediated by preterm birth, and assess several other potential explanatory factors.Methods For singleton live births at gestation 22+ weeks in England and Wales during 2006-2012, the Office of National Statistics provided a file linking birth and death registrations with other routine data. Variables included ethnicity (11 groups) and six covariates: gestational age, area deprivation (IMD quintile), gender, mother’s age at delivery, mother’s country of birth (UK-born or not), parental status at birth registration (within marriage, jointly registered by both parents at same address, joint registration with different addresses, sole registration). Using Stata, we fitted a sequence of logistic regression models for UID, starting with ethnicity and adding factors in the order in which they reduced the likelihood ratio test statistic for ethnic variation.Results There were 4.6 million births and 1559 UID events (0.34 per 1000). Risk varied by ethnicity (p<0.001), being low for Black African, White non-British, and all South Asian groups, but high for Black Caribbean and two Mixed Black groups (African-White and Caribbean-White). Unadjusted odds ratios relative to White British ranged from 0.38 (95% CI 0.24-0.60) for Indian to 1.73 (1.21-2.47) for Black Caribbean. All covariates were associated with UID risk. When jointly adjusted for two factors (parental registration status and mother’s country of birth), the ethnic variation was not statistically significant (p=0.6). Other factors, including gestational age, made very little difference. Results were robust to sensitivity analysis for potential sparse-data bias.Conclusion This large population-based study found substantial ethnic disparity in risk of UID, which was not attributable to preterm birth, maternal age or area deprivation. Parental registration status and mother’s country of birth jointly accounted for the variation, suggesting mediation by cultural factors. Better understanding of infant care practices in low-risk ethnic groups might lead to more effective prevention of UID in the general population.Keywords: Sudden infant death syndrome, ethnicity, health inequalities

OP38

ETHNIC DIFFERENCES IN CHILDHOOD HEIGHT TRAJECTORIES AND THE ROLE OF EARLY LIFE FACTORS: EVIDENCE FROM THE UK MILLENNIUM COHORT STUDY

Y Lu, A Pearce, L Li1Great Ormond Street Institute of Child Health, University College London, London, United Kingdom

Background Height growth is an indicator of early life growth conditions and is associated with health in later life (including cardiovascular disease and some cancers). Cross-sectional research has shown that ethnic differences in height exist, however little is known about how children from different ethnic backgrounds grow throughout childhood. Using contemporary and UK-representative data, we investigated how childhood height trajectories differ by ethnicity and whether any differences are explained by early life factors.Methods We used data from the Millennium Cohort Study (~18,000 children born 2000-2002) and included White, South Asian (Indian, Bangladeshi and Pakistani) and Black (Black African and Black Caribbean) singletons with height measured at least 1 time point (N=15,114). Mixed effects cubic growth models were applied to height measurements at ~3, 5, 7 and 11 years to estimate ethnic differences in height trajectories with and without adjustment for early life factors: parental height, prenatal factors (birth order, maternal smoking during pregnancy and age at childbirth),

birthweight, and family socio-economic circumstances in infancy (maternal education, family income).Results Compared to their White counterparts, Black boys and girls were taller at 3 years by 2.1cm (95% CI: 1.4, 2.9) and 3.1cm (2.42, 3.73) and subsequently grew at a faster rate on average by 0.24cm/year (0.13, 0.35) and 0.31cm/year (0.18, 0.44), respectively. Consequently, height differences at age 11 were much greater. No differences were found between South Asian and White boys. South Asian girls were marginally taller at 3 years by 0.5cm (0.1, 0.9) but had a slightly slower growth rate between 3 and 11 years by -0.12 cm/year (-0.19, -0.05). Height differences at 3 years increased after adjustment of parental height and further adjustment of birthweight, possibly because ethnic minority children on average did not have taller parents and had a lower birthweight than White children. Adjustment for other early life factors did not alter these differences further.Conclusion Our study is the first to investigate ethnic differences in childhood height growth of UK children. Black children were taller and grew faster than White children, and this was not explained by early life factors explored. Despite having shorter parents, South Asians had a similar growth trajectory to White children, suggesting a greater intergenerational gain for South Asians. Future research should examine whether the height advantage seen in Black children continues beyond age 11, and the potential health implications of rapid height gain (often accompanied by rapid weight gain).Keywords: ethnicity, child growth, birth cohort

OP39

REASSESSING PATTERNS AND TIME TRENDS IN BODY MASS INDEX IN BLACK AFRICAN AND SOUTH ASIAN CHILDREN BETWEEN 2007 AND 2013: THE NATIONAL CHILD MEASUREMENT PROGRAMME

MT Hudda1, CM Nightingale1, AS Donin1, CG Owen1, AR Rudnicka1, JCK Wells2, H Rutter3, DG Cook1, PH Whincup1

1Population Health Research Institute, St. George's, University of London, London, UK2Childhood Nutrition Research Centre, UCL Great Ormond Street Institute of Child Health, London, UK3Department of Health Services Research and Policy, London School of Hygiene and Tropical Medicine, London, UK Background High body fat (BF) levels in UK children are a major problem, with particular concerns about children of South Asian and Black (presumed African) origin. However, available national data are based on body mass index (BMI), which underestimates BF in South Asian children and overestimates BF in Black children. We reassessed childhood BF patterns and time trends between 2007-08 and 2012-13 using ethnicity-specific BMI adjustments to derive adjusted BMI levels (aBMI) that were related to BF in an equivalent way across South Asian, Black and White children.Methods Analyses were based on National Child Measurement Programme (NCMP) data from 2007-08 to 2012-13 in 3,195,323 aged 4-5 years and 2,962,673 children aged 10-11 years. aBMI values were derived for South Asian and Black children in a separate study, using body composition measured by the deuterium dilution method. These aBMI values related to BF similarly in South Asians, Blacks and Whites. To examine time trends in mean aBMI and in overweight-obesity prevalence by ethnicity, multiple linear regression and logistic regression analyses were used respectively.Results In the first year (2007-08), mean aBMIs in 10-11 year-olds (boys, girls) compared with Whites (18.64, 18.98 kg/m2) were higher in South Asians (20.08, 19.94 kg/m2; both P<0.001*) and lower in Black boys but higher in Black girls (18.38, 19.21 kg/m2; both P<0.001*). Mean 5-year changes in aBMI (boys, girls) compared with Whites (0.02, 0.11 kg/m2) were significantly higher in South Asians (0.16, 0.32 kg/m2; both P<0.001*) and Black boys but not girls (0.13, 0.15 kg/m2; P=0.01, P=0.41*). Among South Asians, both means and mean changes were greater among

Bangladeshis and Pakistanis than Indians. Ethnic differences in mean aBMI levels were similar in younger children, but time trends were not as pronounced. Analyses examining overweight-obesity prevalence showed similar findings to those using mean aBMI.  Parallel analyses in unadjusted BMI (the usual NCMP data presentation) showed similar time trends but different mean BMI patterns; Black children had highest mean unadjusted BMI values while White boys and South Asian girls had the lowest.  *P-Value for difference from WhitesConclusion These analyses using aBMI data emphasize the particularly high (and rising) burden of BF among UK South Asians, compared with both White and Black children. These patterns are not adequately described by unadjusted BMI data, which tend to overestimate the BF burden in Black children and underestimate that in South Asians.  Keywords: childhood obesity, time trends, ethnic groups

OP40

ETHNIC DIFFERENCES IN EARLY GLYCAEMIC CONTROL (HbA1C) IN CHILDHOOD ONSET TYPE 1 DIABETES

A.R. Khanolkar1 2, R. Amin1, D. Taylor-Robinson3, R. Viner1, J. Warner4, E.F. Gevers5, T. Stephenson1

1GOS Institute of Child Health, University College London, London, UK2Institute of Environmental Medicine, Karolinska Institutet, Stockholm, Sweden3Department of Public Health and Policy, University of Liverpool, Liverpool, UK4Department of Child Health, Children’s Hospital for Wales, Cardiff, UK5Deptartment of Paediatric Endocrinology, Queen Mary University London, and Barts Health NHS Trust, London, UKBackground Ethnic minorities with Type 1 Diabetes (T1D) have worse glycaemic control (higher glycated haemoglobin or HbA1c) and greater risk for vascular complications. However, the impact of ethnicity on early glycaemic control when patients experience transient remission post-diagnosis is limited. We used modelling techniques to examine the independent contribution of ethnicity on HbA1c trajectories during the first six months post-diagnosis of T1D. This is important to investigate as early glycaemic control during the first year post-diagnosis is linked to subsequent future diabetes control including in adulthood.Methods Routine clinical data on 443 (50% female) children <19years of age, with T1D and attending one of three paediatric diabetes clinics in east London between Jan 2005 and Dec 2015 were included in the study.Subjects self-identified their ethnicity and were grouped into White, Mixed (any mixed ethnicity), Black, African-Somali, Bangladeshi and Asian-other (any Asian origin excluding Bangladeshi). Socioeconomic status (SES) was derived from postcode using Indices of Multiple Deprivation (IMD) 2010 for England. Linear longitudinal mixed effects models, which allow comparison of population average HbA1c levels and change over time for the different ethnic groups while controlling for potential covariates, were used to assess effects of ethnicity on longitudinal HbA1c trajectories (1,028 HbA1c datapoints) during the first six months post-diagnosis. Models were adjusted for sex, age at diagnosis and socioeconomic status. Quadratic growth curve modelling was used to identify and plot discrete HbA1c trajectories by ethnicity.Results All ethnic minorities had higher mean HbA1c at diagnosis compared to White children, with largest mean differences in Bangladeshi (9.7mmol/mol, 95% CI 5.1-14.3), Asian-Other (5.8mmol/mol, 95% CI 2.2-9.3) and Somali (5.2mmol/mol, 95% CI 0.1-10.2) children. HbA1c decreased by average of 19.5mmol/mol (-21, -18) for all children during the first one month post-diagnosis. Population averaged HbA1c decreased between diagnosis and four months, followed by a gradual increase in HbA1c levels (mean difference -30mmol/mol between diagnosis and six months). All ethnic groups had higher HbA1c levels throughout this time period compared to White children. SES was not associated with HbA1c at diagnosis nor did it significantly impact the estimates for the ethnicity – HbA1c growth trajectories. 

Conclusion Ethnic minorities have worse glycaemic control at diagnosis compared, with largest mean differences observed in Bangladeshi, Asian-other and Somali children and these higher levels track into the first six months after diagnosis. Interventions at the time of diagnosis and during the first few months are needed to reduce the observed inequalities.Keywords: ethnicity, paediatric diabetes, disease management

Thursday 7 September 2017

Lifecourse early life

OP41

INFLUENCES OF PRECONCEPTIONAL DIET ON INFANT ANTHROPOMETRY: ANALYSES FROM THE SOUTHAMPTON WOMEN’S SURVEY

SR Crozier1, HM Inskip1, KM Godfrey1, C Cooper1, SM Robinson1

1MRC Lifecourse Epidemiology Unit, University of Southampton, Southampton, UK

Background Increasing evidence suggests that the pre-conceptional period is a critical time for determining future disease risk in the offspring.  Maternal obesity before and during pregnancy has important consequences for offspring health, but there is limited information about the effect of pre-conceptional diet.Methods The Southampton Women’s Survey has measured the diet, body composition, physical activity and social circumstances of 12,583 non-pregnant women aged 20 to 34 years living in the city of Southampton, UK. 3,158 women subsequently became pregnant and delivered liveborn singleton infants within the study period.  Anthropometric measurements were taken from the babies at birth.  UK-WHO z-scores were calculated for birthweight, length and head circumference.  Principal component analysis of the dietary data collected before pregnancy identified a ‘prudent’ (healthy) dietary pattern.  Using a Directed Acyclic Graph maternal education, pre-pregnancy smoking, maternal age and parity were identified as potential confounders of the association between prudent diet score and body composition at birth.  In addition, analyses were adjusted for maternal height, and for sex, age and gestation where the outcome was not a z-score, to improve the precision of the model.  The associations were examined using linear regression models, additionally including a term for the interaction of prudent diet score with pre-pregnancy smoking.Results Whilst a higher prudent diet score was associated with a higher birthweight, length and triceps skinfold thickness, there were statistically significant interactions between pre-pregnancy smoking and prudent diet score for all anthropometric outcomes.  Amongst women who did not smoke before pregnancy there was no association between pre-pregnancy prudent diet score and birthweight (β = -0.02 (95% CI -0.07, 0.02) SD per SD increase in prudent diet score, P=0.34) or crown-heel length (β = 0.00 (-0.05, 0.04), P=0.75), whereas amongst women who did smoke before pregnancy there was a positive association between pre-pregnancy prudent diet score and birthweight (β = 0.18 (0.10, 0.25), P<0.001) and crown-heel length (β = 0.13 (0.07, 0.20), P<0.001).  Similar interactions were found for all other neonatal anthropometric measurements.Conclusion Women who have healthier diets before pregnancy have offspring who are larger and longer on average; this effect is much stronger amongst women who smoke before pregnancy than amongst those that don’t smoke.  Reduced bioavailability of key micronutrients amongst smokers points to potential mechanisms underlying this observation.  This study provides further evidence that there may be benefit in developing interventions to support women to improve health-related behaviours before they become pregnant.Keywords: Diet anthropometry smoking

OP42

GLYCAEMIC LOAD AND INDEX IN PREGNANCY ARE ASSOCIATED WITH POSTNATAL, BUT NOT PRE-PREGNANCY, DEPRESSIVE SYMPTOMS; LONGITUDINAL DATA FROM THE SOUTHAMPTON WOMEN’S SURVEY

H Inskip1,2, H Okubo3, S Crozier1, C Cooper1,2, K Godfrey1,2, S Robinson1,2, J Baird1

1MRC Lifecourse Epidemiology Unit, University of Southampton, Southampton, UK2NIHR Southampton Biomedical Research Centre, Univ of Southampton & Univ Hospital Southampton NHS Foundation Trust, Southampton, UK3Department of Health Promotion, National Institute of Public Health, Saitama, Japan

Background High glycaemic load (GL) and glycaemic index (GI) diets have been linked to poor health. In cross-sectional studies, associations between GL and GI and psychological disorders have also been identified, though the direction of association has not always been consistent. Few longitudinal data exist and none in relation to associations between GL and GI during pregnancy and postnatal depression. We aimed to assess whether higher GL and GI during pregnancy were associated with increased risk of postnatal depressive symptoms using data from the Southampton Women’s Survey (SWS).Methods The SWS is a population-based cohort study of 12,583 women aged 20-34 years assessed when not pregnant; those enrolled during the latter half of the recruitment period completed the General Health Questionnaire (GHQ-12), a short screening instrument for depressive symptoms. Women who became pregnant (n=3,158) were followed up. Six months post-partum, mothers completed the Edinburgh Postnatal Depression Scale questionnaire to determine postnatal depressive symptoms. Established cut-offs for each depression scale were used to determine presence of depressive symptoms before and after pregnancy. At 11 and 34 weeks’ gestation, diet during the preceding 3 months was assessed using an interviewer-administered food-frequency questionnaire, from which GL and GI were determined. Prevalence ratios (PRs) for postnatal depressive symptoms were obtained using Poisson regression with robust variance. Potential confounding factors were identified through a Directed Acyclic Graph, namely age, educational attainment and smoking during pregnancy. Linear regression was used to assess the relationship between pregnancy GL and GI and pre-pregnancy depressive symptoms.Results Postnatal depression data were available for 2856 women, with 2038 and 2429 of them having GL and GI data at 11 and 34 weeks’ gestation respectively. In univariate analyses, postnatal depressive symptoms were positively associated with GL at both time points and GI at 34 weeks’ gestation, but were strongest for 34 weeks’ GL: PR 1.12 per 100 GL units (95%CI: 1.04-1.20). After adjustment for confounders, 34 weeks’ GL was the only measure associated with postnatal depressive symptoms: PR 1.09 per 100 GL units (95%CI: 1.01-1.17). Notably, there was no association between pre-pregnancy depressive symptoms derived from the GHQ12 and either GL or GI.Conclusion These findings suggest that improving diet in pregnancy, particularly lowering GL in late pregnancy, might protect against postnatal depressive symptoms. Pre-pregnancy depressive symptoms were not associated with GL and GI in pregnancy indicating that reverse causation is unlikely.Keywords: mental health, diet, pregnancy

OP43

NETWORK META-ANALYSIS OF EMPIRICALLY DERIVED COMPONENT CLASSES IN PARENTING INTERVENTIONS FOR CHILD CONDUCT DISORDER

GJ Melendez-Torres1, P Leijten2, 3, W Knerr3, FEM Gardner3

1Warwick Medical School, University of Warwick, Coventry, UK2Child Development and Education, University of Amsterdam, Amsterdam, NL3Department of Social Policy and Intervention, University of Oxford, Oxford, UK

Background Child conduct disorders have important long-term public health sequelae, including higher risk of substance use and violence in adolescence and higher risk of mental health problems in adulthood. Parenting interventions informed by social learning theory have been consistently

shown to be effective for reducing child disruptive behaviour, but the diversity and variety of interventions prevents a clear understanding of why interventions are differentially effective. Intervention components rarely exist in isolation, and interventions may be best understood in terms of the combinations of components they include rather than any one component. We tested if components of social learning theory-based parenting interventions for child disruptive behaviour outcomes formed underlying groupings across trials, and analysed how these groupings were associated with differential intervention effectiveness.Methods We searched existing systematic reviews across a variety of platforms and updated searches in recent reviews. We included randomised controlled trials of parenting interventions for child conduct disorder measuring child disruptive behaviour outcomes. We developed and applied an exhaustive 19-component scheme against all trials. We estimated a latent class model for components in each trial arm and chose the best-fitting model. We subsequently took 20 draws from the probability distribution of the latent classes for each arm, network meta-analysed each draw, and combined findings from each draw using Rubin’s rules.Results We included 151 trials with 329 arms comparing relevant parenting programmes against each other, against treatment as usual or against no treatment. Across 19 components, our latent class analysis identified five intervention types (four active, one no treatment). As compared to no treatment, ‘relationship-building’ was most effective at reducing child disruptive behaviour (Cohen’s d -0.71, 95% CI [-0.90, -0.53]) but ‘parental self-management’ (-0.54, [-0.76, -0.32]), ‘breaking coercive cycles’ (-0.53, [-0.74, -0.33]) and ‘all components’ (-0.49, [-0.68, -0.29]) were effective as well. There was some evidence that ‘relationship-building’ was more effective than other active intervention types but confidence intervals for these comparisons were broad.Discussion We brought together two types of methods, latent class modelling and network meta-analysis, to examine how intervention classes are associated with differential intervention effectiveness. This is the first analysis of its kind. While all active intervention classes were better than no treatment and no active intervention class was better than any another, we did not find that using all possible components in the same intervention was associated with increased effectiveness as compared to more spare groupings of components.Keywords: child public health, parenting, systematic review

OP44

FEEDING DIFFICULTIES AND MATERNAL CONCERNS AT AGE 3 YEARS ARE ASSOCIATED WITH A DECLINE IN CHILDREN'S DIET QUALITY TO AGE 6 

M Jarman1, SM Robinson2, 3, WT Lawrence2, 3, ME Barker2, 3, KM Godfrey2, 3, C Cooper2, 3, HM Inskip2, 3, J Baird2

1Agriculture, Food and Nutrition Science, University of Alberta, Edmonton, Canada2MRC Lifecourse Epidemiology Unit , University of Southampton, Southampton, UK3NIHR Southampton Biomedical Research Centre, University of Southampton, Southampton, UK

Background Cross-sectional studies show that picky eating behaviours in young children are associated with poorer quality diets. Limited longitudinal research explores whether these effects persist until later childhood. We assessed feeding difficulties and eating behaviours at age 3 years and their influence on quality of diet at age 3 and 6.Methods The Southampton Women’s Survey is a prospective cohort study of women pre-pregnancy, those that became pregnant and their children. Complete data for this analysis were available on 1946 mother-child pairs. When children were age 3 years, mothers completed a questionnaire assessing frequency of feeding difficulties, and concerns about specific eating behaviours in three categories indicating the level of the problem. Diet was assessed using a food frequency questionnaire. Principal components analysis was used to examine dietary patterns; a z-score was derived with higher scores reflecting a better quality of diet. Dietary assessment was

repeated at age 6 years. Linear regression models were used to assess associations between feeding difficulties, concerns about eating behaviours and diet scores at age 3 and 6. Models were adjusted for maternal age, education and number of children, and interactions between feeding difficulties and maternal concern about eating behaviours were assessed.Results At age 3 more frequent feeding difficulties were associated with poorer quality diets [β -0.14 SD/category (95% CI -0.18,-0.09)]. Greater maternal concern about the eating behaviours ‘my child refused to eat the right foods’ and ‘my child has been choosy with food’ was associated with poorer quality diet at age 3 [β -0.33 SD/category (95% CI -0.43,-0.23) and β -0.20 SD/category (95% CI -0.31, -0.09), respectively] compared to children whose mothers were less concerned. These associations remained at age 6, even after adjustment for diet score at age 3, suggesting that greater maternal concern about their child ‘refusing to eat the right foods’ and ‘being choosy with food’ was associated with a worsening of children’s diet quality to age 6 [β -0.13 SD/category (95% CI -0.23,-0.04) and β -0.16 SD/category (95% CI -0.27, -0.05), respectively].Conclusion These findings suggest that the way mothers respond to their child’s eating behaviours at age 3 might influence a decline in children’s quality of diet until age 6. Heightened maternal concern may generate a negative feeding environment which exacerbates picky eating behaviours in young children. Family support during preschool years for managing eating behaviours could contribute to more optimal diets in early childhood.Keywords: Longitudinal Eating behaviours Children

Behaviours 1

OP45

THE EFFECT OF MULTIPLE RISK BEHAVIOURS IN ADOLESCENCE ON EDUCATIONAL ATTAINMENT AT AGE 16 YEARS: A UK BIRTH COHORT STUDY

CL Wright1, RR Kipping1, M Hickman1, J Heron1, R Campbell11School of Social and Community Medicine, University of Bristol, Bristol, UK

Background Single health risk behaviours are negatively associated with educational outcomes, but any relationship between multipe risk behaviours (MRB) and educational attainment has not been explored. We assessed the extent to which simultaneous engagement in MRB during adolescence is associated with examination performance at age 16 years.Methods Data from the Avon Longitudinal Study of Parents and Children (ALSPAC), a population based birth cohort study of children born in England between 1991 and 1992 were used in linear and logistic regression analyses. We estimated the association between the total number of a wide range of health risk behaviours (including tobacco smoking, hazardous alcohol use and TV viewing) ranging from 0 to 13 behaviours, and educational attainment. The outcomes were: capped General Certificate of Secondary Education (GCSE) score and odds of attaining five or more GCSEs at grades A*-C. Multiple imputation was used to account for missing data.Results We found that engagement in MRB was strongly associated with poorer educational attainment and each additional risk equated to a one grade reduction in capped GCSE score or a reduced odds of attaining five or more A*-C grades of 23%. The average cohort member engaged in 3 MRB and would therefore have an associated reduction in their GCSE examinations equivalent to three grades in one examination, or reduced odds of attaining five or more A*-C grades of 69%.Conclusion Our findings show for the first time that engagement in a greater number of MRB during adolescence is strongly associated with poorer educational attainment at age 16 years. Preventing MRB could improve educational attainment and thereby directly and indirectly improve longer term health outcomes. Keywords: Multiple risk behaviours, educational attainment, ALSPAC

OP46

CHANGES IN CONSUMPTION OF ADDED SUGARS FROM AGE 13 TO 30 YEARS: A SYSTEMATIC REVIEW AND META-ANALYSIS OF LONGITUDINAL STUDIES

E M Winpenny1, T L Penney1, K Corder1, M White1, E M F van Sluijs1

1MRC Epidemiology Unit, University of Cambridge, Cambridge, UK

Background Intake of added sugar among adolescents is generally above recommended levels. Added sugar intake has been associated with weight gain and cardiometabolic risk factors during adolescence. Dietary habits developed during this period may persist into adulthood, increasing chronic disease risk in later life. A better understanding of the underlying trajectories of sugar consumption across adolescence and early adulthood will help inform appropriate interventions for this age group. This systematic review investigated changes in intake of added sugars and sugary foods and drinks, and determinants of such changes, between the ages of 13 and 30 years.Methods In accordance with the registered protocol (PROSPERO: CRD42015030126), 7 electronic databases were searched in January 2016 for longitudinal studies of diet during adolescence or early adulthood. The papers retrieved were screened for studies including multiple measures of intake of sugars or sugary foods or drinks in cohort participants between the ages of 13 and 30 years. Data from included studies were extracted and analysed using random-effects meta-analysis, by the three main nutrient and food group categories identified.Results We identified 23 papers reporting longitudinal data on intake of added sugar or sucrose (n=5), sugar-sweetened beverages (SSBs) (n=20) and/or confectionery (n=9).  Eight papers reported data from the US, with Norway, Sweden and Australia also contributing several papers. On average, we found a per-year of age decrease in added sugar or sucrose intake (-0.28% total energy intake (95%CI -0.44; -0.12)), a decrease in confectionery consumption (-0.20 servings/week (95%CI -0.41; -0.001)) and a non-significant decrease in SSB consumption (-0.15 servings/week (95%CI -0.32; 0.02)). A small number of studies reported associations between behavioural determinants (e.g. screen time and fast food consumption) and change in intake of added sugar. Only three papers reported data beyond the age of 21 which weakens applicability of these findings to the early adulthood years.  Conclusion This review demonstrates a decrease in added sugar intake from adolescence to early adulthood, which may suggest an opportunity to capitalise on such changing preferences with interventions to further improve dietary choices within this age range. Improved longitudinal data is needed to further develop our understanding of changes in added sugar consumption into early adulthood and determinants of these changes.Keywords: sugar adolescence longitudinal

OP47

RELATIONSHIPS BETWEEN SUGAR INTAKE FROM SWEET FOOD AND BEVERAGES, COMMON MENTAL DISORDER AND DEPRESSION: PROSPECTIVE FINDINGS FROM THE WHITEHALL II COHORT STUDY

A Knüppel1, M J Shipley, C H Llewellyn, E J Brunner1Department of Epidemiology and Public Health, University College London, London, UK

Background In the face of high rates of mood disorders and an overconsumption of sugar, intake of sweet food, beverages and added sugars has been linked with depressive symptoms in several populations. It remains unclear if this association is causal or due to reverse causation (the influence of mood on sugar intake). Our study aimed to investigate systematically cross-sectional

and prospective associations between sugar intake from sweet food/beverages, common mental disorder (CMD) and depression, and to examine the role of reverse causation as potential explanation for the observed linkage.Methods We analysed repeated measures (23,245 person-observations) from the Whitehall II cohort study. The study recruited non-industrial civil servants from 1985 to 1988. The most recent data used in this analysis was collected in 2013. CMD was measured with the 30-item General Health Questionnaire and depression with the 20-item Center for Epidemiologic Studies Depression scale. Sugar intake from sweet food/beverages was assessed using food frequency questionnaires. Prospective analyses included 2-, 5- and 10-year follow-up periods. We modelled associations using random effects regression using Stata 14.Results Cross-sectional analyses showed positive associations. In prospective analyses we found a positive association of sugar intake from sweet food/beverages with incident CMD in men and with recurrent depression in women. Men in the highest tertile of intake had a 24% increased odds of incident CMD after 5 years (95% CI: 1.02, 1.48) independent of health behaviours, socio-demographic and diet-related factors, adiposity and other diseases (fully adjusted model). Women had a 36% increased odds for recurrent depression per 30g increment (95% CI: 1.03, 1.80) in fully adjusted models, associations using tertiles of sugar intake from sweet food/beverages were similar but not statistically significant when adjusted for diet-related factors. Notably, neither CMD nor depression predicted changes in sugar intake from sweet food/beverages.Conclusion Our findings using repeated measures follow-up data over 22 years suggest an adverse effect of sugar intake from sweet food/beverages on long-term psychological health. These results add further support for public health interventions which promote reduced sugar intake to improve overall health.Keywords: Epidemiology, mental health, diet,

OP48

AGE, SEX AND SOCIOECONOMIC INEQUALITIES IN FRUIT AND VEGETABLE INTAKE IN UK ADULTS, 1986-2012

A Yau, J Adams, P Monsivais1Centre for Diet and Activity Research, MRC Epidemiology Unit , University of Cambridge, Cambridge, United Kingdom

Background Low fruit and vegetable (FV) consumption is associated with cardiovascular disease and some cancers. The UK recommendation is that adults should consume at least 400g (5x80g portions) of FV per day. However, the most recent available data indicate that over 80% of UK adults do not adhere to the ‘5-a-day’ recommendation. This study investigates whether the level of adherence has changed over time. Further, we look at number of FV portions consumed by age, sex and socioeconomic position across the three time points between 1986 and 2012.Methods We used three surveys for this repeated cross-sectional analysis: the National Diet and Nutrition Survey (2000/01 and 2008/12) and a similar survey from 1986/87, the Dietary and Nutritional Survey for British Adults. These surveys employ multistage random sampling. FV intake was assessed using food diaries and demographic information collected by an interviewer. We included adults aged 19 to 64, who adequately completed a food diary, which gave sample sizes of 2047, 1724 and 1512, in 1986/87, 2000/01 and 2008/12 respectively. To test for socioeconomic inequalities, respondents were categorised into four groups (professional/managerial (I/II); skilled non-manual (IIINM), skilled manual (IIIM); partly skilled/unskilled (IV/V)) based on the Registrar General Social Class classification and we looked at absolute and relative differences between I/II and IV/V. To test differences in mean intake by demographic characteristics, we used t-tests and linear regression. All analyses were conducted in StataSE 13.

Results Overall FV intake increased by 20% from 1986/87 (2.67 portions, 95% confidence interval [2.59,2.74]) to 2008/12 (3.21 portions [3.10,3.32]), but still fell short of the recommended 5 portions. We found higher FV intake in socioeconomic group I/II compared to IV/V across all years (difference of 0.96-1.28 portions; 26.09%-36.36%). FV intake also differed by age across all surveys, with older respondents eating more than younger respondents (1.03-2.08 portions; 51.76%-129.19%). Men ate significantly more FV than women in 1986/87 (0.18 portions; 6.52%) but there was no significant difference in the latter time points.Conclusion While FV consumption has increased in the UK since 1986, and sex differences in consumption have disappeared, socioeconomic and age-related inequalities persist. Population-level strategies to improve diet and increase FV intake are still needed in the promotion of public health. In order to redress inequalities, it is crucial that these strategies are at least or more effective in lower versus higher socioeconomic groups, and in younger versus older adults. Keywords: socioeconomic inequality; diet; time trends

Ageing 1

OP49

ASSOCIATION BETWEEN TEMPERATURE AND DEATH AMONG ELDERLY PEOPLE IN ENGLAND 2012/13-2013/14: A CASE-CROSSOVER DESIGN

P Tammes1, C Sartini2, I Preston3, AD Hay1, D Lasserson4, RW Morris1

1School of Social and Community Medicine, University of Bristol, Bristol, UK2Department of Primary Care & Population Health, University College London, London, UK3Household Energy Services, Centre for Sustainable Energy, Bristol, UK4Nuffield Department of Medicine, University of Oxford, Oxford, UK

Background Around 24,000 extra deaths occur annually in winter in England and Wales. NICE guidance suggests GPs should identify patients most at risk. We investigated whether socio-demographic and clinical characteristics could predict cold-related mortality.Methods Data on over 500,000 patients aged 65+ from the Clinical Practice Research Datalink (CPRD) were linked with ONS death registration, yielding 34,777 patients who died between April 2012 and March 2014. We used daily temperature data from the Met Office to calculate (i) absolute mean temperature and (ii) difference from average monthly temperature (relative temperature) for the date of death and three days previously. In a case-crossover analysis, we also calculated both temperature measures for the 14th day before and the 14th day after the date of death. Patients assumed to live in an institution were identified using the CPRD family number. From linked Hospital Episode Statistics, we determined whether an emergency hospital admission occurred two years before death to indicate previous health status. Deprivation level and house energy efficiency were determined from patient’s and practice’s Lower Super Output Area respectively: the latter used information from the Centre for Sustainable Energy. Conditional logistic regression models were applied to estimate the odds ratio (OR) of death associated with temperature and interactions between temperature and socio-demographic, medical and house quality characteristics were expressed as relative odds ratios (RORs).Results Higher absolute temperature was associated with lower risk of death (OR 0.985 per 1oC; 95%CI 0.975-0.992; p=<0.001). There was weak evidence of a positive association between risk of death and higher relative temperature (OR 1.008 per 1oC; 95%CI 0.999-1.017; p=0.056). No interactions were found between temperature measures and age, gender, living in urban/rural areas, deprivation level, or house energy efficiency in either bivariable or multivariable analyses. There was some evidence for a stronger effect of higher relative temperature for those living in an institution (ROR 1.025; 95%CI 1.002-1.048; p=0.03), but not in multivariable analysis. Effects of temperature measures differed between those who had none vs at least one previous emergency admission: ORs for absolute temperature were 0.970 and 0.988 per 1oC, with ROR 1.018, 95%CI

0.998-1.039, p=0.079. For relative temperature ORs were 1.033 and 1.003, with ROR 0.974, 95%CI 0.951, 0.997, p=0.025, suggesting less impact of relative temperature for those with a previous emergency admission.Conclusion Recommendations for GPs to identify those at highest risk during cold weather cannot be supported by these results.Keywords: winter mortality, risk predictor, electronic patients records, meteorological measures

OP50

INEQUALITIES IN TIME BETWEEN STOPPING WORK AND DEATH: ONS LONGITUDINAL STUDY

ET Murray1, E Carr1, P Zaninotto1, J Head1, N Shelton1

1Epidemiology and Public Health, University College London, London, United Kingdom

Background Due to the financial challenges of increasing life expectancy, many industrialised countries are raising state pension eligibility ages (SPA).  However, use of average life expectancy to calculate SPA ignores inequalities in health and life expectancy across socio-economic groups.Methods Data was used from 120,552 members of The Office for National Statistics Longitudinal Study (LS) who were aged 50-75 at the 2001 census and had information on work status at the 2011 census, or died between 2001 and 2011.  First, multinomial logistic regression was used to examine the odds of being dead or not being in work at the 2011 census date, compared to being in paid work, by Registrar General occupational social classes.  Then, right-censored linear regression was used to examine mean social class differences in age of stopping work, age of death, and years of life between stopping work and death before aged 85; separately for each outcome.  All models were adjusted for gender and self-rated health status in 2001.Results By the 2011 census date, 12.8% of the sample had died, 66.8% were alive but not working and 20.4% remained in work. Women were less likely to both to die in the next 10 years or to remain in work than men.   Gender-adjusted analyses showed that lower social class was associated with earlier mean age of stopping work [unskilled manual vs professional: -2.76 years (95% CI -3.04,-2.48)] and mean age of dying (before aged 85) [-3.92 (-4.56,-3.27)].  Before aged 65, work exit was mostly due to exit from paid employment, rather than mortality; although unskilled workers in this age group were still 1.92 times (1.69,2.19) more likely to die over the 10-year period than professional workers.  Most of these differences were explained by health status in 2001.  For those who stopped work (n=89,330), gender-adjusted mean differences in time from stopping work to death (before aged 85) were greatest for managerial and skilled non-manual groups [0.76 (0.02,1.50) and 1.01 (0.24,1.77)], compared to professional occupations, with no differences seen for lower social groups.  This U-shaped relationship was explained by the combination of higher social class occupations staying in work longer but experiencing lower rates of mortality, and lower social class occupations leaving work at earlier ages but experiencing higher rates of mortality.  Conclusion Lower occupational social groups were both more likely to stop work before state pension age and also could expect less years in receipt of state pension.Keywords: mortality pensions policy

OP51

FUNCTIONAL AND MENTAL HEALTH TRAJECTORIES PREDICTING DEMENTIA INCIDENCE: LATENT CLASS ANALYSIS IN THE ENGLISH LONGITUDINAL STUDY OF AGEING

D. Cadar1, H. Davies1, D. Llewellyn2, G.D. Batty3, A. Steptoe1

1Department of Behavioural Science and Health, University College London, London, UK

2Mental Health Research Group, University of Exeter Medical School, University of Exeter, Exeter, UK3Department of Epidemiology and Public Health, University College London, London, UK

Background Functional disability and depression might be related to an increased risk of future dementia.  We aimed to examine the occurrence of functional disability and depression over an eight-year period and evaluate the predictive role of these changes in dementia incidence.Methods The data used for these analyses are from 1,670 men and women aged 65 and older from the English Longitudinal Study of Ageing (ELSA), an on-going, open, prospective cohort study.  Seven waves of data between 2002 and 2014 were analysed. Dementia was determined by doctor-diagnosis combined with a score above the threshold of 3.38 on the Informant Questionnaire on Cognitive Decline in the Elderly. Independent linear growth models with time-invariant covariates and a categorical distal outcome (dementia incidence in 2014) were used to examine whether different trajectories of functional abilities (activities of daily living, ADL; and instrumental activities of daily living, IADL) or depressive symptoms (Center for Epidemiologic Studies Depression Scale, CES-D) between 2002 and 2010 were predictive of new dementia cases four years later.Results We identified three independent patterns of functional and depressive symptoms trajectory.  Most people showed no functional impairments over time (‘class III’) in either ADL (86%) or IADL (90%), a small group showed a steep deterioration (‘class II’) in ADL (5%) and IADL (3%), while the others show a substantial disability in either ADL (9%)or IADL (7%) at baseline with only a slow improvement over time (‘class I’).  Similarly, most participants (79%) did not experience depressive symptoms during the study period (‘class III’), a small group (9%) showed increased depressive symptoms over time (‘class II’), while the others (2%) started with depressive symptoms and experienced a minor improvement over time (‘class I’).  After adjustment for age, sex, education and wealth, relative to participants in ‘class III’, study members who experienced a sharp deterioration in either measure of functional capabilities or depressive symptoms (‘class II’) were markedly more likely to be classified with dementia four years later (ADL: Odds Ratio (OR)=3.29 (95% Confidence Intervals (CI) 0.37-6.21); IADL: OR=14.59 (95% CI 2.09-26.09) & CES-D: OR=3.94 (95% CI 1.27-6.61)).Conclusion Our results showed heterogeneity in all measures of functional abilities or depressive symptoms over time, suggesting that a steeper decline in each of these functions is not necessarily a normative process, but could constitute an early indication of neurodegeneration and pre-clinical symptomatology.  Keywords: Dementia, functional impairment, depression

OP52

ENJOYMENT OF LIFE AS A PREDICTOR OF HEALTHY LIFE EXPECTANCY: EVIDENCE FROM THE ENGLISH LONGITUDINAL STUDY OF AGEING

P Zaninotto1, A Steptoe1

1Epidemiology and Public Health, UCL, London, UK

Background In older adults evidence showed that subjective wellbeing is associated with reduced risk of mortality and the incidence of chronic conditions. With the dramatic increase in life expectancy, it is important to gain a better understanding of the lives of older individuals. While life expectancy is a useful indicator of health, it is becoming increasingly recognized that health expectancy, which quantifies the quality of remaining years of life, is also crucial. The aim of this study is to estimate health expectancy in men and women according to enjoyment of life.Methods We used data from a nationally representative sample of 8,852 individuals aged 50 and over participants of the English Longitudinal Study of Ageing (ELSA), followed from 2002 to

2013.The main outcome measures are chronic disease-free and disability-free health expectancy, the main exposure is enjoyment of life, adjustment variables are social class, sex and age.Discrete-time multistate life table models were used to estimate health expectancies from the ages of 50 to 100, by age and sex and adjusted for social class.Results Men and women experiencing high enjoyment of life could expect to live an additional 32 and 34 years of their remaining lives free from disability. Disability-free life expectancy at the age of 50 and 65 was approximately four years shorter for men and women who experienced low enjoyment of life compared to those who experienced high enjoyment of life. At the age of 50 men and women who experienced low enjoyment of life could expect to live an additional 16 years of their lives free from a chronic condition, compared to 20 and 22 additional years for men and women who experienced high enjoyment of life. At the age of 65 chronic disease-free life expectancy is 6 years for men and women experiencing low enjoyment of life and 9 years for those who experienced high enjoyment of life. Conclusion Our study showed that those who experienced enjoyment of life have higher life expectancy than those who don’t. This study also showed that the quality of remaining life is also important, individuals  experiencing high enjoyment of life could expect to live a higher number of years in good health (without disability and chronic conditions) compared to those experiencing low enjoyment of life.Keywords: Enjoyment of life, Healthy life expectancy, Older age

Interventions

OP53

SHOULD THE CURRENT NATIONAL CHILD MEASUREMENT PROGRAMME BE EXTENDED TO INCLUDE YOUNGER CHILDREN?

N Foster, M Cortina-Borja, L Griffiths, C Dezateux1Population, Policy and Practice Programme, University College London, Great Ormond Street Institute of Child Health, London, United Kingdom Background In 2016 the National Child Measurement Programme (NCMP), which measures the height and weight of all 4-5 and 10-11 year-olds in England, reported that 19.8% of 11 year-olds were obese. The parliamentary Health Committee report - ‘Childhood obesity – brave and bold action’ - recommended evaluating extension of NCMP measurements to include younger children. We used longitudinal data, unavailable from NCMP, from the Millennium Cohort Study (MCS) to evaluate two additions to the current NCMP schedule: a preschool measurement at age three years, and a primary school measurement at age seven years. We hypothesised that these would achieve earlier detection of children at risk of obesity at later ages, and that parental concern about their child’s future risk of overweight would be greatest among parents of obese children.  Methods MCS children’s heights and weights were measured, and their parents interviewed, at ages three, five, seven and 11 years. We included data for 14,789 singletons (51.0% boys) seen at age five whose height, weight and BMI values met NCMP criteria. We imputed missing height and weight measurements, calculated BMI and estimated obesity prevalence using UK90 clinical cut-offs, weighted for survey design (Stata: Release 14). We calculated prevalence rate ratios (PRR) to examine associations between parental concern about their child’s future risk of overweight, reported at the age five interview, and child’s obesity at age five years.Results At three, five, seven and 11 years respectively, 1009 (6.6%), 1010 (6.4%), 1153 (7.5%) and 1772 (11.8%) children were obese. A preschool measurement would identify 79.6% (95% CI: 76.3, 82.4) of obese five year-olds as overweight (28.6%) or obese (51.0%) at age three, and an additional primary school measurement 68.1% (64.4, 71.4) of newly obese 11 year-olds (i.e. not obese at age five) as overweight (39.0%) or obese (29.1%) at age seven. At the age five interview,

66.6% (63.1, 69.9) of parents of obese five year-olds reported concern about their child’s future overweight risk compared to 22.3% (21.4, 23.2) of parents of healthy-weight five year-olds (PRR: 2.94; 95% CI: 2.69, 3.20). Conclusion Extension of the current NCMP schedule to include younger children would achieve earlier identification of those at risk of obesity at later ages in primary school. At age five, parental concern about their child’s future overweight risk is strongly related to their child’s current obesity status. Further work is needed to understand parental views about, and to evaluate the cost-effectiveness of, options to extend current programmes.Keywords: Obesity, childhood, intervention

OP54

INDIVIDUAL-, FAMILY- AND SCHOOL-LEVEL INTERVENTIONS TO PREVENT MULTIPLE RISK BEHAVIOURS IN YOUNG PEOPLE AGED 8-25 YEARS: A COCHRANE SYSTEMATIC REVIEW AND META-ANALYSIS

GJ MacArthur1, DM Caldwell1, J Redmore2, S Watkins1, V Er1, R Langford1, J White3, R Lingam4, C Chittleborough5, K Pasch6, DJ Gunnell1, M Hickman1, RM Campbell11School of Social and Community Medicine, University of Bristol, Bristol, UK2College of Medicine and Veterinary Medicine, University of Edinburgh, Edinburgh, UK3Centre for Trials Research, Cardiff University, Cardiff, UK4Institute of Health and Society, Newcastle University, Newcastle-upon-Tyne, UK5School of Public Health, University of Adelaide, Adelaide, Australia6Department of Kinesiology and Health Education, University of Texas, Austin, USA

Background Little is known about the impact of interventions that target multiple risk behaviours in young people and there remains no quantitative estimate of their effect. We reviewed and quantified the effect of individual, family or school-level interventions that targeted engagement in two or more risk behaviours (including; tobacco use, alcohol use, illicit drug use, gambling, self-harm, sexual risk behaviour, antisocial behaviour, vehicle-risk behaviour, physical inactivity and poor nutrition), among young people aged 8-25 years.Methods Eleven databases were searched (including Medline, Embase, PsycINFO, CINAHL and CENTRAL), alongside grey literature, until November 2016 without geographic or language restriction. Randomised controlled trials that targeted two or more eligible behaviours were included. Screening, data extraction and risk of bias assessment were conducted in duplicate. Quality of the studies was rated using GRADE (Grading of Recommendations Assessment, Development and Evaluation). Data from eligible studies were pooled in a random effects meta-analysis, conducted in Revman 5.3, sub-grouped by intervention type (individual, family or school-level and universal or targeted in focus).  Results A total of 27,691 unique titles were screened, and 62 eligible studies included. The majority were universal school-based interventions (n=26), followed by targeted family interventions (n=13). A substantial proportion of interventions targeted substance use and/or antisocial behaviour (n=40, 65%), while none targeted prevention of gambling or self-harm alongside the other behaviours. 52 of 62 studies (84%) were conducted in the USA.Universal school based interventions were associated with benefit in relation to alcohol use (OR 0.72, 0.56-0.92, p=0.009; n=8 studies), tobacco use (OR 0.77, 0.60-0.97, p=0.03; n=9 studies), and engagement in any antisocial behaviour (OR 0.79, 0.64-0.97, p=0.02; n=11 studies), though there was evidence of heterogeneity (I2 57-68%, χ² 16.55-36.95; all p‹0.05). There was also evidence that such interventions improved the odds of vigorous activity (OR 1.32, 1.16-1.51, p‹0.0001, I2 0%, n=4 studies). Targeted family-level interventions were associated with reduced odds of antisocial behaviour only (OR 0.67, 0.46-0.96, p=0.03; n=4 studies), though fewer of these studies were identified overall. Meta-analyses did not support the effectiveness of individual-level interventions

for any outcome (n≤4 studies for each outcome). Some caution is warranted since the evidence was rated as being of low or very low quality using the GRADE approach.Conclusion Available evidence suggests that universal school-based interventions that target multiple risk behaviours may be effective in preventing engagement in alcohol use, tobacco use and antisocial behaviour, as well as improving physical activity, among young people. (Protocol: Cochrane Database CD009927)Keywords: Risk behaviour, adolescent, systematic review

OP55

HEALTH, HAPPINESS AND WELLBEING IN THE TRANSITION FROM ADOLESCENCE TO ADULTHOOD: APPLYING AN “EQUITY LENS” TO AN OVERVIEW OF SYSTEMATIC REVIEWS OF POPULATION LEVEL INTERVENTIONS

AK Macintyre1, J McLean2, P Campbell3, J Williams4, C Torrens3, M Maxwell5, H Biggs2, A Pollock3, A Woodhouse6

1Centre for Health Policy, University of Strathclyde, Glasgow, Scotland 2Research & Development, Mental Health Foundation , Edinburgh, Scotland 3Nursing, Midwifery and Allied Health Professions Research Unit, Glasgow Caledonian University, Glasgow, Scotland4Clinical & Health Psychology , University of Edinburgh , Edinburgh, Scotland 5Nursing, Midwifery and Allied Health Professions Research Unit, University of Stirling , Stirling, Scotland 6Policy Team, Children in Scotland , Edinburgh , Scotland

Background Adolescence is a key period for health and development. There is evidence for interventions to promote health in adolescence, yet health inequalities persist into adulthood. Whilst there is robust data describing socioeconomic inequalities in adult health, there is comparatively less on adolescent health, and we have little insight on what works to tackle inequities in adolescence. We aimed to apply an “equity lens” to systematic review evidence on population interventions for adolescent health. Specifically, we aimed to synthesise systematic review evidence on the equity impact of population interventions designed to improve health, happiness and wellbeing in children and young adults aged 10-24 years.Methods A rapid systematic review of reviews (an overview) was conducted. We systematically searched the Cochrane Database of Systematic Reviews; Health technology assessments (HTA); Campbell Collaboration; EPPI; Joanna Briggs Library; and Database of Reviews of Effects (DARE), MEDLINE, EMBASE, CINAHL, PsycINFO and PubMed for systematic reviews published in English between 01 January 2005 and 07 March 2016. Two reviewers independently applied pre-specified selection criteria. Methodological quality was independently assessed using the ROBIS tool. Disagreements were resolved through discussion. Key data relating to the review focus, including information on the review aim, participants, interventions, comparisons and outcomes were extracted. Data on whether reviews reported the socioeconomic characteristics of primary studies, differential effects, or the delivery of population interventions with disadvantaged groups were also collated. Data were synthesised within evidence tables, categorised according to themes and subthemes.Results We identified 35310 reviews, screened 1953 abstracts and retrieved 566 full text articles. 150 reviews, assessed as either low risk of bias or unclear risk of bias were included in the final synthesis. Evidence was mapped across 9 topic areas; mental health and wellbeing; tobacco free living; preventing drug abuse and excessive drinking; sexual and reproductive health; violence and abuse free living; active living; healthy eating; obesity and general health. The majority of reviews did not describe the socioeconomic characteristics of primary studies, and examination of the differential impact of population interventions varied widely across topic areas.

Conclusion Systematic review evidence on the equity impact of population interventions for adolescents is limited. To the best of our knowledge, this is the first overview to synthesise review evidence with a focus on transition from adolescence to adulthood. This work provides important insights on how we can advance a focus on equity and considers the implications for addressing inequalities in adolescent health.   Keywords: Equity; Adolescence; Interventions

OP56

EFFECTS OF NATIONAL HOUSING QUALITY STANDARDS ON HOSPITAL EMERGENCY ADMISSIONS: A QUASI-EXPERIMENT USING DATA LINKAGE

S E Rodgers1, W Poortinga2, R Bailey1, R Johnson1, F Dunstan3, D Berridge1, R A Lyons1

1Swansea University Medical School, Swansea University, Swansea, UK2Welsh School of Architecture, Cardiff University, Cardiff, UK3Institute of Primary Care and Public Health, Cardiff University, Cardiff, UKBackground National housing quality standards are being applied throughout the UK. A housing improvement programme was delivered through a local authority to bring nearly 9000 homes up to the Welsh Housing Quality Standard (WHQS). Homes received multiple elements, including new kitchens, bathrooms, windows and doors, insulation, and heating and electrical systems, through an eight-year rolling work programme. The study aimed to determine the impacts of the different housing improvements on hospital emergency admissions for all residents.Methods Intervention homes, council homes that received at least one element of work, were data linked to individual health records of residents. Counts of admissions relating to respiratory and cardiovascular conditions, and falls and burns, were obtained retrospectively for each individual in a dynamic housing cohort (January 2005 - March 2015). The intervention cohort criterion was for someone to have lived in any one of the intervention homes for at least three months within the intervention period. Counts were captured for up to 123 consecutive months for 32,009 individuals in the intervention cohort and analysed using a multilevel approach to account for repeated observations for individuals, nested within geographic areas. Negative Binomial regression models were constructed to determine the effect for each element of work on emergency admissions for those people living in homes in receipt of the intervention element, compared to those living in homes that did not meet quality standards at that time. We adjusted for background trends in the regional general population, as well as for other confounding factors.Results People of all ages had 34% fewer admissions for cardiovascular and respiratory conditions, and fall and burn injuries while living in homes when the electrical systems were upgraded, compared to the reference group (IRR=0.66, 95% CI 0.58-0.76). Reduced admissions were also found for new windows and doors (IRR=0.78, 0.70-0.87), wall insulation (IRR=0.80, 0.73-0.87) and garden paths (IRR=0.81, 0.73-0.90). There were no associations of change in emergency admissions with upgrading heating (IRR=0.92, 0.85-1.01), loft insulation, (IRR=1.02, 0.93-1.13), kitchens (IRR=1.01, 0.87-1.18), or bathrooms (IRR=0.99, 0.87-1.13).Conclusion Improving housing to national standards reduces the number of emergency admissions to hospital for residents. Strengths of the data linkage approach included the retrospective collection of complete baseline and follow up using routine data for a long-term intervention, and large scale regional adjustment.Keywords: data linkage, emergency admissions, housing intervention

Environment 1

OP57

SOCIOECONOMIC PATTERNING OF FOOD AND DRINK ADVERTISING AT PUBLIC TRANSPORT STOPS IN THE CITY OF EDINBURGH, UK

T Robertson1, K Lambe2, L Thornton3, R Jepson2

1Faculty of Health Sciences and Sport, University of Stirling, Stirling, UK2Scottish Collaboration for Public Health Research and Policy, University of Edinburgh, Edinburgh, UK3Faculty of Health, Deakin University, Melbourne, Australia

Background Outdoor advertising has been shown to disproportionately promote unhealthy foods, with links to food preferences, purchasing and consumption habits. Given the socioeconomic inequalities in food quality, consumption and obesity, it is also possible that advertisers are disproportionally advertising unhealthy food products in lower socioeconomic areas. The aim of this study was to explore the socioeconomic patterning of food advertising at bus stops in Edinburgh, UK. Our hypothesis was that there would be a higher prevalence of unhealthy (fast food, soft drinks, confectionary etc.) versus healthy (water, low sugar beverages etc.) food advertisements situated in more deprived areas within the city.Methods This study took place in Edinburgh, Scotland in 2015. In the city, there were 2,227 bus stops, 447 of which had advertising shelters (20%). ‘ODK collect’, an open-source application, was used to record data at each bus stop including GPS coordinates, type of area (residential, industrial etc.), details about the food or drink product being advertised and whether a price or special offer was featured. GPS coordinates were converted to postcodes and then the area level measure of socioeconomic deprivation (Scottish Index of Multiple Deprivation; SIMD). SIMD scores were further converted into quintiles for analysis. Generalized Linear Models were used to compare the patterning of food adverts by area-level deprivation, with and without adjustment for ward size, population and area type. All analyses were conducted using SPSS version 21. Results In total, 562 food advertisements were recorded across 298 bus stops, with a mean of 1.89 food advertisements per shelter. Over 85% of all advertisements were for food products. Nine categories of food and related advertisements were identified including alcohol, confectionary, coffee, fast food outlets, food stores, fruit juices, frozen desserts, iced coffee and soft drinks. No adverts for fresh fruit or vegetables (besides juices), water or low sugar beverages were recorded. Across all nine food categories there were no associations between increased prevalence of these adverts and deprivation level. For example, fast food outlet advertisements (which made up 39% of all food adverts) were no more likely to be present in lower versus higher socioeconomic areas (B=0.248, 95% CI: -0.082, 0.578, p=0.140).Discussion While food advertisements were abundant across the city, there were no patterns related to the socioeconomic status of the areas where these were located. However, all could be classed as advertising unhealthy food and drink products. This study was limited by not considering other forms of outdoor advertising.Keywords: food, advertising, obesity, inequality, socioeconomic status

OP58

EXAMINING ASSOCIATIONS BETWEEN NEIGHBOURHOOD BUILT ENVIRONMENTS AND ADIPOSITY IN THE UK BIOBANK COHORT

KE Mason1, N Pearce1, S Cummins2

1Non-Communicable Disease Epidemiology, London School of Hygiene & Tropical Medicine, London, United Kingdom2Social and Environmental Health Research, London School of Hygiene & Tropical Medicine, London, United Kingdom

Background Persistent inequalities in obesity-related health outcomes may be partly due to unequal distribution of resources in local built environments. For example, differential neighbourhood access to physical activity (PA) facilities or exposure to unhealthy food environments may influence health behaviours and ultimately be reflected in uneven population distributions of overweight and obesity. Despite much research, evidence on health effects of neighbourhood environments remains inconclusive, making it difficult to generalise to suitable interventions, if indeed intervention is warranted.Methods Using cross-sectional data from ~270,000 adults aged 40-70 from the UK Biobank cohort residing across England and Wales, including linked person-centred environmental data, we examined whether features of the fast food and PA environments near an individual’s place of residence were independently associated with measures of adiposity. We also constructed a composite exposure measure to examine the PA and fast food environments operating together, classifying people’s neighbourhoods on a scale of obesogenicity, from high (limited PA facilities and close to a fast food outlet) to low (many PA facilities and far from a fast food outlet). Multilevel models were used to account for clustering due to an area-based sampling design and were adjusted for potential confounding effects of individual and area-level variables, including each exposure on the other.Results Considered separately we found that greater density of PA facilities and greater distance to the nearest fast food outlet were independently associated with smaller waist circumference (WC) e.g. having ≥6 formal PA facilities within a 1km street network distance of home was associated with 0.91cm lower WC (95% CI: 0.32-1.49) than having no nearby PA facilities, and living >2km from the nearest fast food outlet was associated with 0.42cm lower WC than living within 500m (95% CI: 0.11-0.72). Similar patterns were observed for other adiposity outcome measures. Preliminary results from ongoing analyses using the composite exposure measure indicate a possible dose response of WC to increasing combined neighbourhood obesogenicity. A series of sensitivity and other additional analyses currently underway, including the use of propensity scores, will also be presented to explore the possibility that residual confounding may explain the findings.Conclusion Combining a very large sample with wide geographical coverage and robust statistical methods we sought improved clarity on the potential health impact of two built environment exposures, and present evidence suggesting that improving neighbourhood access to PA facilities and minimising proximity to fast food outlets may help reduce adiposity in the UK mid-aged adult population.Keywords: neighbourhood effects; physical activity environment; fast food environment

OP59

PREVALENCE AND PATTERNING OF HEALTHY, LOW-CARBON LIFESTYLES IN THE UK: A CROSS-SECTIONAL ANALYSIS OF UK BIOBANK BASED ON COMBINATIONS OF TRAVEL AND DIETARY BEHAVIOUR

MA Smith1, JR Böhnke1,2, H Graham1, PCL White3, SL Prady1

1Department of Health Sciences, University of York, York, UK2Hull York Medical School, University of York, York, UK3Environment Department, University of York, York, UK

Background There is considerable policy interest in promoting behaviours with health and environmental co-benefits, but current research has typically focused on single behaviours in isolation rather than on understanding healthy, sustainable lifestyles more broadly. The aim of this study was to describe the prevalence and socio-demographic patterning of healthy, low-carbon (HLC) lifestyles in the UK population by identifying clusters of travel and dietary behaviours that have implications for both human health and carbon emissions.

Methods We analysed self-reported data from participants in UK Biobank (aged 39-72) who completed a 24-hour dietary recall questionnaire (n=211,049). Measures of travel behaviour included transport mode(s) for both commuting and non-work journeys (car, public transport, walking, cycling) as well as average daily driving time. Measures of dietary behaviour included consumption of red and processed meat (RPM), fruit and vegetables, and vegetarian status. We used latent class analysis (LCA) to identify unique clusters of travel and dietary behaviour and characterised each group as ‘higher-carbon’ or ‘lower-carbon’ based on its indicators. Best-fitting LCA models were selected using information criteria and interpretability. Multinomial logistic regression was used to examine socio-demographic differences between each cluster, compared to the highest-carbon class. All analyses were stratified by sex due to gender differences in diet and active travel behaviour.Results The best-fitting models identified 10 different classes among females and 9 among males. The largest classes were characterised by higher car use (2-4 hours/day) and higher RPM consumption (>1 serving/day) representing 72% of males, and 65% of females. The proportion leading entirely HLC lifestyles (composed of female cyclists, urban vegetarians) was very small (3%). Several groups comprised a much larger segment whose lifestyles were partially or predominantly HLC (20% of males, 27% of females). In fully adjusted multinomial models, the most consistent predictors of HLC lifestyles (across all classes) were having higher qualifications, residing in an urban postcode, and living in and around London.Discussion This is the first study to measure HLC lifestyles in the UK based on combinations of travel and dietary behaviour. We found that wholly HLC lifestyles are very rare and particularly scarce outside of the most highly urbanised areas, however a sizable minority of the population engages in behaviours that are partially or predominantly HLC. The existence of clustering between travel and dietary behaviours suggests that there is a policy role for establishing stronger links between these areas and for promoting HLC lifestyles more holistically.Keywords: Environment, Active travel/Physical activity, Food/Dietary consumption

OP60

AGE-FRIENDLY ENVIRONMENTS AND PHYSICAL ACTIVITY: A CROSS-SECTIONAL, OBSERVATIONAL STUDY OF THE OVER 55S IN IRELAND 

S Gibney1,3, M Ward1,2, E Sexton1, S Shannon1

1Healthy and Positive Ageing Initiative, Department of Health , Dublin , Ireland 2, Trinity College Dublin , Dublin , Ireland 3Geary Institute for Public Policy, University College Dublin , Dublin , Ireland

Background The benefits of physical activity extend beyond improving or maintaining physical health. Regular activity reduces the risk of dependency in old age, enables social participation, promotes mental health and well-being, and in combination with other lifestyle factors, may reduce the chances of developing dementia. Despite the benefits of regular physical activity, one-in-three older adults in Ireland have low activity levels. Therefore, the objective of this study is to examine the effect of health, social connectedness, and perceived accessibility and safety of the local environment on physical activity in the over 55s in Ireland.  Methods Data was from the Age-Friendly City and Counties Survey (2016), a population-representative cross-sectional survey of community-dwelling adults aged 55+ administered in 21 Local Authority areas in Ireland (n=10,540).  Data was collected face-to-face using Computer Assistant Personal Interviews. Mixed-effects negative binomial regressions were used to estimate the effect of 1) health status and behaviours, 2) social connectedness, 3) availability and accessibility of recreational green areas, and 4) perceptions of safety in the local area, on physical

activity. Moderate and vigorous activity was measured using a brief version of the International Physical Activity Questionnaire and reported as minutes-per-week. Models were adjusted for socio-demographic characteristics. Results are reported as Beta (β) Coefficients, with Standard Errors (S.E).Results In the fully adjusted model, area-level differences explained 8% of the observed variance in physical activity. Poor health (β -.74, S.E. 0.22, p<0.001), loneliness (β -.11, S.E. 0.02, p<0.001), community participation (β .34, S.E. 0.5, p<0.001), and difficulty accessing green spaces (β -.19, S.E. 0.09, p<0.05) partially explained physical activity differences. Several socio-demographic characteristic were also associated with physical activity. Women (β -.03, S.E. 0.09, p<0.001) older adults (aged 75+) (β -.02, S.E. 0.07, p<0.001), and those looking after a family or home (β -.02, S.E. 0.08, p<0.01), were less physically active than their peers. These findings are limited to self-reported perceptions of the local environments whereas geographical data could add further relevant information about area-level social deprivation and distance to services and green spaces. Conclusion In Ireland, like many other cities and countries that have subscribed to the World Health Organisation’s Age-friendly Cities and Counties Programme, locally-directed social and health strategies are increasingly being developed. These results shows that, combined with individual-level behaviour change interventions, improvements to the local environment and promoting social connectedness may be useful in promoting physical activity among the over 55s.Keywords: Ageing, Physical Activity, Environment

Mental health and wellbeing

OP61

UNDERSTANDING SOCIAL INEQUALITIES IN CHILD MENTAL HEALTH: FINDINGS FROM THE UK MILLENNIUM COHORT STUDY

VSS Straatmann, MC Campbell, CR Rutherford, SW Wickham, DTR Taylor-Robinson1Department of Public Health and Policy, University of Liverpool, Liverpool, United KingdomBackground Child mental health is poor in the UK, with the most disadvantaged children experiencing worse outcomes and consequences over the course of their lives. Using a contemporary U.K. birth cohort, we therefore explored the social gradient in poor child mental health and the extent to which it is explained by other known risk factors for adverse child mental health outcomes.Methods Analysis of the U.K. Millennium Cohort Study (MCS), based on 9,818 children who participated in five survey sweeps (9 months, 5, 7 and 11 years old). The main outcome was child socio-emotional behavioural problems using the Total Difficulties Score of the Strengths and Difficulties Questionnaire (SDQ), at age 11. Relative Risk (RRs) and 95% confidence intervals (CI) for socioemotional behavioural problems were estimated using Poisson regression, according to maternal education, which was used as a measure of socio-economic circumstances at birth. Sequential models adjusted for risk factors for child mental health problems included demographic factors (sex, ethnicity and maternal age), family poverty, maternal mental health, and being bullied. Analyses were conducted using Stata/SE with svy commands to account for the sampling design and attrition.Results By age 11, 10.4% (95%CI 9.6%-11.2%) of children had socioemotional behavioural problems. Children of mothers with no qualifications were more than five times as likely to have mental health problems compared to degree level (RR 5.4 [95%CI 4.0-7.4]). Male sex, younger maternal age, poor maternal mental health, family poverty and being bullied, were all independently associated with an increased risk of child mental health problems. Adjusting for maternal mental health, family poverty and being bullied attenuated the RR for mental health problems in the lowest maternal education group compared to the highest (4.2 [95%CI 3.0-5.9]); 4.0 (95%CI 2.8-5.7); and

4.9 (95%CI 3.5-6.8) respectively. Adjusting for all risk factors attenuated the RR to 3.4 (95%CI 2.3-5.0). Conclusion In a representative U.K. child cohort, we found one in ten children faced socioemotional behavioural problems at age 11. The risk was much greater in disadvantaged children. This was partially explained by the social patterning of maternal mental health, family poverty, and being bullied. The self-reported outcome is a limitation of this study. Future research should investigate critical/sensitive periods for these exposures over the life-course. Efforts to reduce inequalities in child mental health problems should focus on reducing socioeconomic inequalities and action on risk factors such as maternal mental health, child poverty, and bullying. Keywords: mental health; inequalities; socioeconomic; longitudinal; cohort; child health

OP62

WHAT ARE THE PHYSICAL AND PSYCHOLOGICAL HEALTH EFFECTS OF SUICIDE BEREAVEMENT ON FAMILY MEMBERS?: A QUALITATIVE STUDY

A Spillane1,2, C Larkin3, P Corcoran1,2, K Matvienko-Sikar1, E Arensman1,2

1Department of Epidemiology and Public Health, University College Cork, Cork, Ireland2National Suicide Research Foundation, University College Cork, Cork, Ireland3Department of Emergency Medicine, University of Massachusetts Medical School, Worcester, United States

Background Research indicates that experiencing the suicide of a relative can have a significant impact on family members’ psychological health. However, research incorporating the impact of suicide bereavement on family members’ physical health is sparse. Therefore, the aim of this study was to examine how family members have been physically and psychologically affected by a relative’s suicide.Methods This exploratory qualitative study is a follow-up to a larger case-control study, which examined the psychological, psychiatric and work-related factors associated with suicide in Ireland (SSIS-ACE, 2014-2017). Participants for the SSIS-ACE study were next-of-kin of persons who died by suicide or probable suicide, who were identified via coroner’s records. All participants who completed the SSIS-ACE interview and who consented to further follow-up were invited by letter to take part in the current study. Semi-structured interviews, with the use of a topic guide were conducted with 18 relatives experiencing suicide bereavement. Eleven participants were female and seven were male; participant’s ages ranged from 25-73 years. Duration of bereavement ranged from 15 to 37 months. Thematic analysis was used to analyse the data which was facilitated by the use of NVIVO 11 to organise the data.Results Preliminary findings indicate the emergence of five themes in two main domains, psychological and physical outcomes.Psychological outcomes: For most participants, the suicide was viewed as a predominantly negative turning point in their lives, where the death forced them to confront a new reality without their loved one. Secondly, immediate emotional reactions, including shock, disbelief, guilt, anger and surprise were reported. Suffering from persistent mental health difficulties was a recurrent theme among bereaved relatives: difficulties included depression, anxiety, stress, posttraumatic stress disorder, suicidal thoughts and suicide attempts. Coping mechanisms, both positive and negative, utilised by suicide survivors emerged as the fourth theme.Physical outcomes: Immediate physical reactions, including nausea, breathlessness, palpitations, chest pains and losing consciousness were physical reactions reported. Some of these physical conditions did not improve in the months after the death but rather persisted and sometimes worsened over time.Conclusion Family members bereaved by suicide are at risk of mental and physical health sequelae, while also being vulnerable to suicidal thoughts and suicide attempts. Participants were

drawn from a small geographic area and the findings of this study may not be generalisable to other settings. From a policy perspective, this study highlights the importance of providing support services for this group following suicide bereavement.Keywords: suicide, bereavement, health

OP63

THE EFFECTS OF THE 2007-9 FINANCIAL CRISIS ON MENTAL HEALTH IN THE UK: A LONGITUDINAL ANALYSIS OF NON-SUICIDE MENTAL HEALTH TRENDS

B Amies1, L Munford2, M Sutton2

1Centre for Primary Care, University of Manchester, Manchester, UK2Manchester Centre for Health Economics, University of Manchester, Manchester, UK

Background Existing research using area-level data has identified big increases in suicides in the UK and elsewhere through the financial crisis of 2007-09. Areas with larger increases in unemployment tended to have greater increases in suicide rates. The evidence base in this area has two particular weaknesses that this study aims to address. Firstly, this study uses individual-level data rather than aggregated area data, and secondly it examines non-suicide mental health trends rather than suicide rates.Methods A nationally-representative sample of 12,816 individuals from the British Household Panel Survey in England from 2000-2013. Data were not collected for 2009. Fixed effects models were used to examine the data for associations between 12-item General Health Questionnaire score (measuring mental health) and demographic features, employment status and equivalised household income on the unweighted sample. Dummy variables for 2008-2010 (the period from the start of recession up to the start of government budget cuts) and 2011-13 (from the start of budget cuts) separated out the influence of changing economic conditions. Lagged and advanced variables were used to explore associations between mental health and the loss of income or employment through time.Results Poor mental health peaked in 2008, before the peak in unemployment in 2010. Women and poorer people had significantly worse mental health throughout (p<0.000, beta 1.24 (95% confidence interval 1.08-1.40), p<0.000, beta 0.129 (-0.192- -0.065) respectively). Unemployment was significantly associated with worse mental health (p<0.000, beta 1.84 (1.52-2.16)). The association between mental health and income was weaker, with the most persistent effects for the period 2008-2010 (p=0.000, beta -0.64 (-0.88- -0.40)). Worse mental health appeared to preced job or income loss (p=0.028, beta 0.46 (0.049-0.88); p=0.011, beta 0.23 (0.051-0.40), respectively) but mental health did not deteriorate following job-loss, while increasing income appeared to compromise mental health (p=0.006, beta -0.24 (-0.42- -0.069)). Weighted sensitivity analysis confirmed a persistent association between unemployment and mental health but significance was lost between income and mental health.Conclusion Non-suicide mental health deteriorated during the 2007-2009 financial crisis. It appears that worse mental health preceded job loss or income loss. A possible explanation could be that poor mental health conferred weakness in the job market. There are a number of inferences to the research area, for example that a strong social safetynet may protect mental health (by maintaining income) and that in-work mental health support may benefit poorer workers.Keywords: depression unemployment recession

OP64

DO RELIGIOUS PRACTICES AND BELIEFS MODERATE THE ASSOCIATION BETWEEN STRESSFUL LIFE EVENTS AND SUBSEQUENT MENTAL HEALTH AND WELLBEING?

A Kaushal1, D Cadar2, M Stafford1, M Richards1

1MRC Unit for Lifelong Health and Ageing, University College London, UK2Research Department of Behavioural Science and Health, University College London, UK

Background Religion may provide coping strategies for stressful life events (SLEs). Aspects of religious practices and beliefs, e.g. provision of meaning in life, interpretation of difficult events and supportive social contact can all contribute to successful coping strategies. The aim of this project was to investigate if SLEs are associated with mental health and wellbeing in early old age and if religious practices and beliefs moderate these associations.Methods Participants were study members from the MRC National Survey of Health and Development (1946 British birth cohort) who participated in data collection at age 68-69 (n=2148). Mental health and wellbeing were measured using the 28-item General Health Questionnaire and the Warwick-Edinburgh Mental Wellbeing Scale. SLEs were measured using a checklist of events on five occasions from age 26 to 60-64. Religious practices and beliefs were ascertained from age 11 to 68-69 and included upbringing, beliefs and attendance. Initial analyses investigated associations between SLEs, and mental health and wellbeing. This was followed by analyses to determine if religious practices and beliefs were more common in people who had good mental health or high wellbeing despite experiencing a high number of SLEs compared to those with fewer SLEs and poor mental health or wellbeing. Analyses were conducted using linear and logistic regression models adjusted for gender and education.Results Experiencing SLEs was associated with lower wellbeing (β=-0.31, CI=-0.46,-0.16) and worse mental health at age 68-69 (β=0.02, CI=0.02,0.03). High wellbeing at age 68 despite a substantial number of SLEs was associated with frequent religious attendance (for men only) from age 36 to 60-64 (OR=2.60, CI=1.10,6.14), religious importance (OR=1.83, CI=1.13,2.97) and meaning in life provided by religion (OR=2.07, CI=1.31, 3.29). There were no differences in religious practices and beliefs between study members with good mental health and a high number of SLEs compared to those with fewer SLEs and poor mental health.Conclusion SLEs across the life course were associated with poor mental health and wellbeing in early old age. We also found evidence that religious practices and beliefs moderate the effect of SLEs on wellbeing but not mental health, and in particular that religious attendance is beneficial for men and not women. Future work will investigate if SLEs are associated with changes in religious attendance, mental health and wellbeing, and if this varies by different types of SLEs, e.g. personal and interpersonal, health and work-related events.Keywords: Religion, Stress, Coping, Mental Health, Wellbeing, Life Course

Behaviours 2

OP65

A SYSTEMATIC REVIEW OF THE PREVALENCE OF SMOKING IN HEALTHCARE STUDENTS

AE Granville1, T McKeever1, R Murray1, K Nilan1

1Epidemiology and Public Health, University of Nottingham, Nottingham, UK

Background Smoking continues to pose a huge cost to an individuals’ health and the healthcare economy.  Healthcare professionals are known to have an authoritative influence over patients and are well placed to promote abstinence from tobacco.  Indeed, Articles 12 and 14 of the Framework Convention on Tobacco Control (FCTC) make several recommendations concerning smoking behaviour and cessation training amongst healthcare professionals.  The current study estimates the prevalence of smoking in healthcare students, healthcare professionals of the future, across the six World Health Organisation regions.

Methods Five databases (Medline, Embase, CINAHL, CAB abstracts, LILACS and the WHO Global Healthcare Professional Survey database) were searched to identify studies including any profession of healthcare students.  Studies were published between January 2000 and March 2016, and no restrictions were placed on language of publication. Titles, abstracts and full texts were checked for eligibility independently by two authors and the quality of the included studies was assessed. Pooled prevalence with 95% confidence intervals (CI) were estimated using random effect models, with heterogeneity quantified using I2.Results 417 papers were included: 214 studies and 203 Global Health Professional student surveys. Healthcare professions included medicine, nursing, dentistry, pharmacy, and mixed groups. The prevalence of smoking across all healthcare students was 19% (95% CI 17-21%, I2 99.98). Subgroup analysis by year shows the prevalence of smoking in healthcare students appears to be increasing; from 16% (95% CI 16-17%, I2 99.23) up to and including 2010 to 19% (95% CI 6-31%, I2 99.99) between 2011 and 2016.  Pooled estimated smoking prevalence within the WHO regions for medical students ranged from 10 to 25%, and nursing students from 0 to 30%.  Estimates for both were highest in Europe and lowest in Africa. Conclusion Smoking prevalence among healthcare students varies widely across professions and WHO regions, however remains worryingly high in light of the key role healthcare professionals play in tackling the global smoking epidemic. In order to continue to make progress with implementation of the FCTC, urgent efforts need to be made to reduce smoking behaviour amongst healthcare students, which will ultimately contribute to the reduction of smoking prevalence amongst their patients.Keywords: Smoking, Healthcare, Students

OP66

MOTIVATION TO QUIT SMOKING AND CHANGES IN CIGARETTE CONSUMPTION, AMONG SMOKERS WHO USE E-CIGARETTES, FINDINGS FROM THE HEALTH SURVEY FOR ENGLAND

L Ng Fat1, S Scholes1, JS Mindell11Epidemiology and Public Health, UCL, London, United Kingdom

Background The majority of people who use e-cigarettes are dual users with tobacco cigarettes. E-cigarettes may aid smokers with their quit attempts and reduce cigarette consumption or reinforce nicotine addiction. This study explores the motivations for current and previous use of e-cigarettes, and whether use is associated with reporting lower or higher cigarette consumption than a year ago. It makes comparisons with other traditional nicotine delivery products (NDPs).Methods This study uses a sample of current smokers aged 16+ (N=3,039) from the nationally representative, cross-sectional Health Survey for England, HSE2013-2014, (HSE2015 data will be included when archived).  Firstly, multinomial logistic regression models were conducted on the odds of a) Never use of e-cigarettes versus b) Current use of e-cigarettes c) Previous (not current) use of e-cigarettes, and key exposure included the intentions to quit smoking scale (No intention/Pre-contemplation/Contemplation/Preparation (within next 3 months)). Secondly multinomial logistic regression was carried out on the odds of reporting smoking a) the same number of cigarettes versus b) more c) fewer than a year ago. Models were repeated using never, current and previous use of other NDPs. All models adjusted for sex, age-group, highest qualification and cigarette consumption. Analyses were conducted using Stata.Results 12% were current users, and 20% previous users of e-cigarettes. Compared with never use of e-cigarettes, no association was found with age and current use, while being younger was associated with previous use (45-54 v. 16-34years, OR=0.68[95%CI 0.46-0.79]). Conversely, current and previous users of NDPs were more likely to be older than never users of NDPs (45-

54years 2.07[1.29-3.30]). Quit intentions had a dose response relationship with the odds of current e-cigarette use (e.g. Preparation versus No intention (3.14[2.24-4.42]); for previous e-cigarette users the magnitude was smaller (1.39[1.04-1.87]). However, ‘Preparation’ had stronger associations with other NDPs, for both current (8.93[5.54-14.40]), and previous use (3.18[2.47-4.09]). Being a current user of e-cigarettes (1.77[1.36-3.20]) or other NDPs (1.72 [1.19-2.50]) increased the odds of reporting smoking fewer cigarettes than the previous year; previous use was not significant. E-cigarette use was not associated with reporting smoking more than the previous year, however current use of NDPs was (1.84[1.13-3.01]).Conclusion Current or previous e-cigarette use is unlikely to increase consumption of cigarettes compared with a year ago, but smokers who used them had weaker intentions to quit smoking than smokers who used other NDPs. Longitudinal research is needed to track changes in consumption involving duration of e-cigarette use to further verify findings.Keywords: smoking, quantitative, behaviour change

OP67

ACTIVITY LEVELS IN MOTHERS AND CHILDREN DURING THE TRANSITION TO PRIMARY SCHOOL: FINDINGS FROM THE SOUTHAMPTON WOMEN’S SURVEY

KR Hesketh1, S Brage1, U Ekelund1,2, C Cooper3,4,5, K Godfrey3,4, NC Harvey3,4, H Inskip3,4, S Robinson3,4, EM van Sluijs1

1CEDAR and MRC Epidemiology Unit, University of Cambridge, Cambridge, UK2Department of Sport Medicine, Norwegian School of Sport Sciences, Oslo, Norway3MRC Lifecourse Epidemiology Unit , University of Southampton, Southampton, UK4NIHR Southampton Biomedical Research Centre, University of Southampton and , University of Southampton & University Hospital Southampton NHS Foundation, Southampton, UK5NIHR Musculoskeletal Biomedical Research Centre, University of Oxford, Oxford, UK

Background Parental physical activity (PA) is frequently investigated as a potential correlate of preschool-aged children’s PA, yet there is little information about how the association between parent-child PA changes during the transition to formal schooling. We aimed to determine the association between objectively measured maternal and 6-year-old children’s PA, exploring how this association differs by demographic and temporal factors; and 2) identify how this association changes during the transition to school (from 4-6 years). Methods Data were from the UK Southampton Women’s Survey. PA levels of 530 6-year-olds and their mothers were measured concurrently using accelerometry for up to 7 days. Two-level mixed-effects linear regression was used to model the association between maternal-child PA at age 6 [average activity intensity (ACC); minutes spent sedentary (SED); and in moderate and vigorous PA (MVPA)]. Interactions with demographic and temporal factors, and how the association differed across the day (morning (6-9am); school (9am-3pm) and evening (3-11pm)), were tested. Change in the association between maternal-child PA (at age 4 and 6, n=170) was also assessed. Results At age 6, mother-child daily PA were positively associated at all activity intensities: ACC: β=0.24 [95% CI: 0.19, 0.30] counts per minute; SED: 0.23 [0.20, 0.26] minutes/hour; MVPA: 0.53 [0.43,0.64] minutes/hour. The association was stronger between mother-child PA at all intensities at the weekend (vs. weekdays: ACC: βinteraction=0.16 [95% CI: 0.06, 0.25] counts per minute; SED: 0.07 [0.02,0.12] minutes; MVPA: 0.44 [0.24,0.64] minutes). For SED, the mother-child association was stronger for children with older siblings (vs. none); for MVPA, the relationship was stronger for those who had both younger and older siblings (vs. none). Longitudinally, the mother-child association did not differ with age for SED and light PA (LPA); mother-child ACC and MVPA were significantly weaker at age 6 compared with age 4 (difference in ACC: -0.23 [-0.37, -0.10], MVPA: -0.16 [-0.31, -0.00]). This difference was driven by a weaker relationship in the mornings and during the school day (9-3pm).

Discussion Maternal-child PA levels are positively associated at age 6, with stronger associations at weekends, and in those with siblings in the home. From age 4 to 6 years, the mother-child ACC and MVPA association weakened. This may reflect decreasing co-participation with age, as children gain independence/ engage in more structured PA at school. Different intervention foci may be needed before and after the transition to school, but family-based PA remains an important element of children’s overall PA.Keywords: Physical Activity Family

OP68

PHYSICAL ACTIVITY TRAJECTORIES AND PREDICTORS DURING THE TRANSITION TO OLD AGE

DAA Aggio1,2, EP Papachristou1, OP Papacosta1, LL Lennon1, SA Ash1, PHW Whincup3, SGW Wannamethee1,2, BJJ Jefferis1,2

1UCL Department of Primary Care & Population Health, UCL, London, England2Physical Activity Research Group, UCL, London, England3Population Health Research Institute, St George’s University of London, London, England

Background Maintaining physical activity during later life is associated with optimal health; however, research on the long-term trajectories of physical activity into old age and their predictors has been limited. This study aimed to identify distinct 20-year trajectories of physical activity spanning from midlife to old age and predictors of physical activity trajectories.Methods Participants were men drawn from the British Regional Heart Study, a prospective cohort study, involving 7735 men recruited from Primary Care Practices in 1978–80. Men were followed up after 12, 16 and 20 years, reporting physical activity levels (walking, cycling, recreational activity and sport/exercise), health status and socio-demographic characteristics. Group-based trajectory modelling was applied to identify distinct trajectories of physical activity and to examine the effects of predictor variables on trajectories. Predictors of trajectory group membership were examined using multinomial logistic regression. The effects of developing cardiovascular disease and changing employment status on trajectories were estimated for each trajectory group.Results 7658 men (mean baseline age 50.2 ± 5.8 years) providing valid questionnaire and physical activity data (of which 78% provided activity data at ≥2 surveys) were included in analyses. Three distinct trajectories of physical activity emerged: low decreasing (34.1%), low stable (45.8%) and moderate increasing (20.1%). Membership of the moderate increasing trajectory group was predicted by being married, having children, drinking alcohol and eating breakfast. Men with manual occupations, men residing in regions other than the South, men who were overweight/obese, men with doctor-diagnosed health conditions and men who smoked were less likely to be in the moderate increasing trajectory group. Being employed was associated with an increase in physical activity in the low decreasing group (β 0.43, p<0.001) but a decrease in the low stable (β -0.22, p<0.001) and moderate increasing groups (β -0.84, p<0.001). Development of cardiovascular-related conditions was associated with a decline in physical activity in the low decreasing (β -0.42, p<0.001) and low stable groups (β -0.13, p<0.05) but was not associated with any change in the moderate increasing group (β -0.06, p=0.34).Conclusion This study highlights the heterogeneity in physical activity levels over time in older adults and that activity levels in midlife are likely to dictate trajectories into old age. Efforts to promote physical activity in later life may need to focus attention earlier in the lifecourse. The effects of retirement and cardiovascular disease on physical activity may depend on prior activity trajectories. Thus, different strategies may be needed in these groups. Keywords: physical activity; trajectories, ageing

Ageing 2

OP69

CIRCULATING N-3 POLYUNSATURATED-FATTY ACIDS AND THE MAINTENANCE OF HEALTHY AGING IN OLDER ADULTS, THE CARDIOVASCULAR HEALTH STUDY

HT Lai1, MC de Oliveira Otto2, RN Lemaitre3, B McKnight3, PHM Chaves4, MC Odden5, AB Newman6, DS Siscovick7, D Mozaffarian1

1Friedman School of Nutrition Science and Policy, Tufts University, Boston, Massachusetts, USA2Department of Epidemiology, University of Texas Health Science Center at Houston, Houston, Texas, USA3Cardiovascular Health Research Unit, University of Washington, Seattle, Washington, USA4Benjamin Leon Center for Geriatric Research and Education, Florida International University Hebert Wertheim College of Medicine, Miami, Florida, USA5School of Biological and Population Health Sciences, Oregon State University, Corvallis, Oregon, USA6Department of Epidemiology, University of Pittsburgh Graduate School of Public Health, Pittsburgh, Pennsylvania, USA7, New York Academy of Medicine, New York City, New York, USA

Background Maintaining healthy aging (HA) is a crucial priority in older adults worldwide, given global population aging, increased number of years living with disability, and the need for new treatments. Omega-3 polyunsaturated fatty acids (n3-PUFA) from seafood and plants exert favorable physiologic effects that could benefit HA. However, relationships between n3-PUFA and HA are not well-established, especially using serial biomarkers which provide highly objective measures.Methods We investigated the longitudinal association between serial circulating n3-PUFAs and maintenance of HA in the Cardiovascular Health Study, evaluating 2,342 older U.S. adults with mean age 75 years and successful HA to-date at baseline in 1992-93. Individual plasma phospholipid n3-PUFAs (expressed as % of total fatty acids) including alpha-linoleic acid (ALA), eicosapentaenoic acid (EPA), docosapentaenoic acid (DPA), and docosahexaenoic acid (DHA) were quantified using gas chromatography in 1992-93, 1998-99, and 2005-06. HA was defined as survival free of cardiovascular disease, cancer, lung disease, and severe chronic kidney disease, with no difficulties with activities of daily living and intact cognitive function (Mini-Mental State Examination ≥80th percentile); dying with a lifetime meeting this criteria was also considered as HA. Events were centrally adjudicated or determined from medical records and diagnostic tests. Multivariable-Cox proportional hazards models with time-varying covariates evaluated the association between time-varying, cumulative average n3-PUFAs and unsuccessful HA.Results During 22 years of follow-up, 267 (11%) participants experienced successful HA. After multivariable-adjustments, the interquintile range of total n3-PUFAs and seafood-derived n3-PUFAs was associated with lower risk of unsuccessful HA by 17% (0.74-0.93 95% CI, p=0.002) and 16% (0.75-0.94 95% CI, p=0.002), respectively. Individually, EPA, DPA and DHA each associated with lower risk of unsuccessful HA by 12% (0.80-0.97 95% CI, p=0.009), 14% (0.77-0.97 95% CI, p=0.010) and 15% (0.76-0.96 95% CI, p=0.009), respectively.  Plant-derived ALA levels were not significantly associated with HA. Sensitivity analyses including freedom from atrial fibrillation, milder chronic kidney disease, and diabetes within the HA definition did not appreciably alter results.Conclusion Among older adults with mean age 75 years and HA to-date, a higher cumulative level of circulating seafood-derived n3-PUFAs (combined and individually), but not plant-derived ALA, was associated with maintainence of HA. These novel findings support guidelines for increased fish intake among older adults; and need for further investigations into plausible biological mechanisms and interventions for effects of n3-PUFAs on maintenance of HA.Keywords: Omega-3 polyunsaturated fatty acids; Healthy aging; Epidemiology

OP70

WEALTH DIFFERENCES IN AGE-TRAJECTORIES OF BODY SIZE: FINDINGS FROM THE ENGLISH LONGITUDINAL STUDY OF AGEING

C Lassale1, A Steptoe1, P Zaninotto2

1Behavioural Science and Health, University College London, London, United Kingdom2Epidemiology and Public Health, University College London, London, United Kingdom

Background While the obesity epidemic has devastating health consequences at all ages, underweight is also associated with an increased mortality risk. Lower socioeconomic status is associated with higher obesity rates and greater weight gain, but evidence from prospective studies in older adults is scarce. Our aim was to describe age-trajectories of body mass index (BMI) and waist circumference (WC) in a population-based study of older adults in England and to assess the association with wealth.

Methods Data come from a nationally representative sample of 3,259 men and 3,966 women aged 52y and over from the English Longitudinal Study of Ageing (ELSA) who had a measurement of BMI and WC on three occasions (2004-2005; 2008-2009; 2012-2013). We used latent growth curve modelling to estimate baseline status (intercept) and rate of change (linear slope) interpreted as the change per year. Intercept and slope were regressed on wealth tertile, and covariates (age, ethnicity, marital status, physical inactivity, smoking status and limiting long-standing illness). Gender- and age-specific (<70, ≥70y) models were fitted.Results In the <70y group, a man aged 60y in the richest wealth tertile had a baseline BMI of 27.7kg/m2 and WC of 100.5cm and a woman a BMI of 26.9kg/m2 and WC of 88.2cm. BMI increased by 0.04 kg/m2 every year in men and 0.05kg/m2 in women; and WC increased by 0.15cm/year in men and 0.21cm/year in women. Being in the poorest wealth group was associated with highest baseline BMI (28.7kg/m2 for men and 29.4kg/m2 for women) and WC (103.1cm in men and 93.6cm in women). However, there was no difference in the rate of change between those in the richest and poorest wealth tertiles. In the ≥70y group, for a man aged 77y in the richest tertile, the baseline BMI was 26.7kg/m2 and WC 100.0cm and for a woman it was 26.3kg/m2 and 88.7cm. The rate of change was non-significant for both anthropometrics markers. An individual of the same age in the poorest tertile had higher baseline BMI and WC (man: 27.5kg/m2 and 102.1cm; woman: 28.1kg/m2 and 92.1cm).Conclusion In this population-based study, BMI and WC increased significantly over time in both men and women until the age of 70, thereafter it remained stable. Less wealth was strongly associated with a higher BMI and WC at any given age, however the rates of change were similar, indicating that the socioeconomic gap associated with excess adiposity did not close with ageing.Keywords: Health inequalities; Ageing; Obesity

OP71

THE EFFECT OF LONGITUDINAL CHANGES IN PHYSICAL AND MENTAL HEALTH ON CONTINUING SOCIAL PARTICIPATION IN OLDER IRISH ADULTS: ANALYSIS FROM THE IRISH LONGITUDINAL STUDY OF AGEING

S Leahy1, C McGarrigle1, D Carey1, RA Kenny1, 2

1The Irish Longitudinal Study on Ageing, Trinity College Dublin, Dublin, Ireland2Mercer's Institute for Successful Ageing, St. James's Hospital, Dublin, Ireland

Background Social engagement and participation in leisure activities are recognised as beneficial to the physical and mental health and wellbeing of older adults and have been shown to lower the risk of negative health outcomes and early mortality. Identifying factors that constrain or enable social participation in older age can help to facilitate continuing engagement and thus improve future health outcomes. This study aimed to investigate the longitudinal relationship between accrual of chronic health conditions and changes in depressive symptoms and continuing social participation in community dwelling older adults over a 4-year period.Methods Data from waves 1 (2009-2011, n=6051), 2 (2012, n=5487) and 3 (2014-2015, n=4623) of the Irish Longitudinal Study of Ageing (TILDA), a stratified probability sample prospective cohort, was analysed. Frequency of participation in seven social activities (‘Go to films, plays, concerts’, ‘Attend classes or lectures’, ‘Travel for pleasure’, ‘Play cards, bingo or games’, ‘Go to pub’, ‘Eat out of house’ and ‘Participate in sport or exercise’) was collected.  Depressive symptoms were assessed using the 8-item Centre for Epidemiological Studies Depression (CESD) scale,  and chronic disease count included self-reported doctors’ diagnosis of cardiovascular (heart attack, angina, stroke, transient ischaemic attack, heart failure) and non-cardiovascular chronic conditions (high blood pressure, diabetes, arthritis, lung disease, osteoporosis). Multilevel mixed effects logistic regression modelling was employed to assess the effect of changes in 1) chronic disease count and 2) depressive symptoms on each item of social participation (defined as monthly participation or less) over three waves, adjusted for socio-demographic and health covariates.Results Mean age at baseline was 63.2y and 46.9% of the sample were male. Rates of social participation remained stable across waves. Each additional chronic disease accrued was associated with decreased participation in ‘Attend classes or lectures’ (Odds Ratio (OR): 0.86, 95%CI: 0.74-0.99) and ‘Participate in sport or exercise’ (OR: 0.86, 95%CI: 0.77-0.97) and an increase in ‘Go to pub’ (OR: 1.28, 95%CI: 1.09-1.50). A one unit increase in depressive symptoms over time was associated with decreased participation in ‘Participate in sport or exercise’ (OR: 0.96, 95%CI: 0.93-0.99) only.Conclusion This longitudinal analysis suggests that deterioration of physical and mental health may influence specific domains of social participation in community dwelling older adults. Holistic approaches to disease management and mental health interventions in older age should include programmes to facilitate and maintain social and leisure time activities.Keywords: ageing, participation, longitudinal

OP72

RELIGIOUS ATTENDANCE, LONELINESS AND DEPRESSIVE SYMPTOMS IN MIDDLE AGED AND OLDER WOMEN IN IRELAND 

J Orr, K Tobin, RA Kenny, C McGarrigle1The Irish Longitudinal Study on Ageing, Trinity College Dublin, Dublin, Ireland

Background Evidence for an association between mental health and religiosity largely supports a positive effect of religiosity on mental health. However, there remains a lack of research into the underlying mechanisms of these associations involving other social and health factors, particularly in older women. We aimed to investigate causal pathways between religious attendance and depressive symptoms, and test whether this relationship is mediated by loneliness.Methods We analysed three waves of The Irish Longitudinal Study on Ageing (TILDA) (2009-2011, 2012, 2014-2015), a stratified probability cohort of men and women aged over 50 resident in Ireland. A total of 3,400 women were included in this analysis. A theoretical longitudinal model of religious attendance and depressive symptoms was tested using Structural Equation Modelling (SEM), adjusted for age, marital status, self-rated health, education and recent adverse life events. Log likelihood tests were used to compare model fit. Depressive symptoms were measured using the 8-item Centre for Epidemiologic Studies Depression Scale (CES-D) and loneliness with the

UCLA Loneliness Scale. Religious practice and beliefs were also collected. Changes in religious attendance between waves were calculated. All analyses were conducted using Stata 14. Results A majority of women attended religious services (86%) and 60% attended at least once a week at baseline, with a decrease in attendance at subsequent time points (85% and 57% at Wave 2; 84% and 55% at Wave 3). Mean (SD) depressive symptoms were 3.38 (4.07) at Wave 1; 3.19 (3.99) at Wave 2; and 3.62 (4.03) at Wave 3. Regular attendance at Wave 1 and Wave 2 predicted fewer depressive symptoms at Wave 2 (Incident Rate Ratio (IRR):0.81 95%CI:0.73-0.89) and Wave 3 (IRR:0.92 95%CI:0.86-1.00) respectively. When loneliness was included in the model, the effect of regular attendance remained unchanged between Wave 1 and Wave 2. The effect was attenuated, but not mediated, between Wave 2 and Wave 3 (IRR: 0.94 95%CI: 0.87-1.02). Depressive symptoms consistently predicted changes in attendance, with higher depressive symptoms predicting subsequent increased and decreased attendance.Conclusion Longitudinal analyses of religious attendance and depressive symptoms suggest this is a complex relationship which is at least in part bidirectional. Evidence did not support loneliness as a mediator of this relationship. Analyses using other measures of religiosity may help further elucidate these associations. Keywords: Mental health, Ageing, Religiosity

Policy analysis

OP73

SYSTEMS SCIENCE FOR CARIBBEAN HEALTH: THE DEVELOPMENT OF A SYSTEM DYNAMICS MODEL FOR GUIDING POLICY ON DIABETES IN A RESOURCE LIMITED SETTING

L Guariguata1, C Guell2, 3, TA Samuels1, EAJA Rouwette4, J Woodcock2, IR Hambleton1, N Unwin1, 2

1Faculty of Medical Sciences, The University of the West Indies, Bridgetown, Barbados2MRC Epidemiology Unit, University of Cambridge, Cambridge, United Kingdom3European Centre for Environment & Human Health, University of Exeter, Truro, United Kingdom4Nijmegen School of Management, Radboud University, Nijmegen, The Netherlands5MRC Epidemiology Unit, University of Cambridge, Cambridge, United Kingdom

Background Type 2 diabetes (T2DM) is a major cause of morbidity and mortality in the Caribbean and a threat to development. Halting its rise by 2025, a WHO target, requires interventions addressing the determinants of unhealthy diet and physical inactivity. The objective of this project is to engage with stakeholders in the development of a system dynamics (SD) simulation model on the effect of different policy interventions on diabetes prevalence and mortality in the Caribbean. Methods Following SD methodology, we used a mixed methods approach to combine interviews (n=13) with stakeholders from multiple sectors across four countries (Barbados, Belize, Jamaica, and Saint Vincent and the Grenadines) with existing data on regional NCD policy and quantitative evidence. An additional twenty-two interviews from stakeholders in 7 Caribbean countries that were conducted as part of an ongoing policy evaluation study were also reviewed. Participants were sampled from existing contacts and their referrals. Analysis was guided by iterative thematic analysis using a grounded approach.A subgroup of stakeholders participated in a group model-building workshop to develop causal maps on the determinants of physical inactivity, unhealthy diet and effective healthcare delivery in the Caribbean. Data from the qualitative and quantitative reviews were applied to further develop the causal structures.Results Stakeholders reinforced the need for upstream interventions including the provision of a healthy food environment and sustainable urban development facilitating active living, supported by

fiscal incentives (subsidies) and disincentives (taxes). Stakeholders emphasized car-based development as a major driver of physical inactivity including a lack of infrastructure for active transport, a high social regard for car usage and a lack of social support for physical activity. Unhealthy diet, according to stakeholders, was driven by changing social structures in the home, the marketing and distribution of highly processed foods, and limited access to locally produced fresh fruits and vegetables. The causal maps are serving as the basis for a stakeholder-driven policy simulation model that will be validated against quantitative evidence.Conclusion The interconnectedness of diabetes determinants highlights the importance of a multi-sectoral response and a systemic approach to decreasing the diabetes burden. The simulation model aims to aid Caribbean policymakers in evaluating the potential impact of different policy options on future diabetes prevalence and mortality.  The overall utility of this approach to stakeholder engagement and option appraisal will be assessed and developed further for use in other resource-constrained settings. Keywords: noncommunicable disease, policy, developing countries, systems science

OP74

UNDERSTANDING THE PROCESS OF DEVELOPING AND IMPLEMENTING CHRONIC DISEASE POLICIES IN THE CARIBBEAN REGION: A QUALITATIVE POLICY ANALYSIS

C Guell1, 2, MM Murphy3, TA Samuels3, 4, L Bishop3, N Unwin2, 4

1European Centre for Environment and Human Health, University of Exeter, Truro, United Kingdom2MRC Epidemiology Unit and Centre for Diet and Activity Research (CEDAR), University of Cambridge, Cambridge, United Kingdom3Faculty of Medical Sciences, University of the West Indies, Cave Hill Campus, Barbados 4Chronic Disease Research Centre, University of the West Indies, Bridgetown, Barbados

Background The Caribbean, like other middle income regions, experiences a significant burden of non-communicable diseases (NCDs), and responded in 2007 with a heads of government declaration committing to a range of policy measures. In 2015, this multi-method qualitative evaluation study investigated the progress made in seven Caribbean countries; this paper reports the findings of an analysis of the process that helped or hindered the development and implementation of policies.Methods We conducted analysis of relevant policy documents, and 76 semi-structured interviews with 80 stakeholders from government (ministries of health, education); civil society (academia, agricultural, health and church organisations) and private businesses. Interviews were conducted by two regional teams. The initial pragmatic coding and categorisation of data with the software Dedoose was deductively framed by the WHO NDC Action Plan, and guided by a Multiple Streams policy approach and realist principles. This was complemented by further inductive thematic analysis and synthesis by the authors.Results Two overarching themes emerged related to the process that guided the development, adaptation and implementation of regional policy goals into local settings.First, stakeholders reported a commonly shared recognition of the need for multi-sectoral partnerships but were frustrated by the difficulties of achieving this in practice. Country-specific NCD commissions served as case studies for both positive experiences of achieving ‘whole of society’ collaboration and difficulties in equitable representation and translation into government-led action. In settings with more effective multi-sectoral working it was possible to identify ‘policy entrepreneurs’ who had the profile and traction to promote cross sectoral action. Second, stakeholders highlighted the importance and opportunity provided by policy transfer. When comparing policy creation across small islands, most success has been made in areas such as tobacco control that could use an international framework and roadmap as a template. As policy efforts were most advanced in larger, more resource-rich countries, some stakeholders called for more synergetic working on policy

solutions across countries. They felt that policy transfer could be improved in a region of relatively small countries but benefitting from a strong regional agenda, cooperation and professional exchange.Conclusion Stakeholders’ local and regional insight, expertise and experience helped to explain the processes that supported or hindered effective policy-making in different settings or different policy areas. Triangulating their reflections with a document review helped to corroborate and strengthen their assessment, provide contextual information to reports and also identified gaps in documentation and narratives.Keywords: policy analysis, chronic diseases, middle-income regions

OP75

CATALYSING GLOBESITY: AN ANALYSIS OF THE IMPACT OF US FREE TRADE AGREEMENTS ON CALORIC INTAKE FROM A NATURAL EXPERIMENT IN CANADA USING THE SYNTHETIC CONTROL METHOD

P Barlow1, M McKee2, D Stuckler1

1Department of Sociology, University of Oxford, Oxford, UK2Department of Health Services Research and Policy, London School of Hygiene and Tropical Medicine, London, UK

Background Free trade and investment agreements (FTAs) are often implicated in the global pandemics of overweight, obesity, and related diseases. Concerns centre on how FTAs increase population exposure to unhealthy, high calorie diets rich in salt, sugar, and fat. Yet the empirical evidence to support these concerns is limited, both methodologically and substantively. Thus, few existing studies of FTAs and dietary change have accounted for unobserved confounding, and many were unable to differentiate the impact of FTAs from contemporaneous macroeconomic and policy changes. Moreover, few have examined any potential effect of FTAs on total caloric intake, despite its critical role in the aetiology of obesity. Here we address these limitations by analysing a unique natural experiment from the unanticipated implementation of the Canada-US Free Trade Agreement (CUSFTA) in 1989.Methods We use data from the UN Food and Agricultural Organisation and the synthetic control method to test the hypothesis that CUSFTA increased caloric intake in Canada. We also anlayse whether  these changes were accompanied by increased US investment and imports in the Canadian food and beverage sector. Finally, we  simulate the population impact on weight gain of these changes in caloric intake using the models developed by Hall and colleagues (2011) and data from the 1990 Canadian Health Promotion Survey.Results Our analysis identifies a 95.8 kilocalorie per capita per day increase in caloric intake after CUSFTA in Canada compared with the synthetic control. These changes coincided with a $1,820 million (95% CI: 1,179.2 to 2,464.1) increase in US investment in the Canadian food and beverage sector, and a $5,258.3 million (95% CI: 4,894.8 to 5,621.8) rise in food and beverage imports from the US. This estimated rise in caloric intake corresponds to an average individual weight gain of 5.7 kg for women and 5.1 kg for men aged 40 and with low physical activity levels, or 2.5kg for women and 2.2kg for men aged 40 and who are very active.Conclusion Our findings suggest that US FTA can substantially alter  dietary behaviour by increasing caloric intake. FTA negotiations may  be a critical window for shaping dietary behaviours to prevent overnutrition, obesity and related diseases.Keywords: Globalization; diet; obesity; Brexit; trade; nutrition;

OP76

EVALUATING THE HEALTH IMPACTS OF RESTRICTIONS TO INCOME SUPPORT FOR LONE PARENTS: A NATURAL EXPERIMENT STUDY USING UNDERSTANDING SOCIETY

S V Katikireddi1, O Molaodi1, M Gibson1, R Dundas1, P Craig1

1MRC/CSO Social & Public Health Sciences Unit, University of Glasgow, Glasgow, UK

Background Lone parents experience poorer health and socioeconomic disadvantage compared to coupled parents. The UK Government has restricted access to Income Support, the primary income replacement benefit for lone parents. Under Lone Parent Obligations (LPO), lone parents are transferred from Income Support to Jobseeker’s Allowance once their youngest child reaches an age threshold (which has been reduced incrementally), and they must prove they are actively seeking work or face sanctions. We investigated the effects of introducing LPO on the health of lone mothers in the UK.Methods We analysed 2009-2013 data from Understanding Society, a panel study representative of the UK general population.  Our primary outcome was the mental health component of SF-12, with the physical health component and self-rated health assessed as secondary outcomes. We identified two intervention groups: lone parents newly exposed following the cut-off change from seven to five years and from ten to seven years. Each of these intervention groups were compared to two control groups: remaining unexposed (since the youngest child was below the age cut-off), and already exposed (since the youngest child was older than the existing age cut-off). We estimated a pooled LPO effect to increase precision. We conducted a difference-in-difference analysis using linear regression to estimate the ‘intention to treat’ causal effect, adjusting for maternal age, number of children and maternal education. Multiple imputation was used to address item missingness.Results Our primary analysis included a total of 2257 participants. Mental health of lone parents consistently declined in intervention groups compared to control groups, whereas physical health and self-rated health showed little change. For lone parents with children aged 5-7 years who were newly exposed, the mental health score of SF-12 changed by -1.39 (95%CI -4.08,1.29) compared to those unexposed and -2.29 (95%CI -4.57,0.00) compared to those remaining exposed. Equivalent figures for the cut-off change from 10-7 years were -2.45 (95%CI -5.48,0.57) and -1.28 (95%CI -4.00,1.45), while for the pooled effect were -2.13 (95%CI -4.17,-0.10) and -2.21 (95%CI -4.13,-0.30). A complete case analysis and inclusion of males within the analytical sample led to similar results.Discussion Increasing conditionality attached to the receipt of welfare benefits adversely impacted mental health of lone mothers but had no short-term impacts on physical health. Our study had limited statistical power and was only able to investigate short-term effects, but the pattern of findings was consistent across comparison groups. Planned extensions to LPO should be reconsidered.Keywords: welfare policy, natural experiments, evaluation

Neighbourhoods & communities

OP77

EVALUATING THE ASSET MODEL: FINDINGS FROM A RAPID REVIEW OF EVALUATION STRATEGIES  

J South1, S Rippon2, K Kinsella1, A-M Bagnall11School of Health and Community Studies, Leeds Beckett University, Leeds, UK2, Aligned Consulting Ltd, , UK

Background Ten years on from Morgan and Ziglio’s call to revitalise the evidence base for public health through an asset model, there are still gaps in understanding about how best to evaluate asset-based approaches. The demand for better evidence poses some challenges as asset-based working emerges from radically different traditions from the mainstream deficit model in public health. This paper will present findings from a rapid evidence review on the measurement of asset-based approaches.Aims The rapid review aimed to improve understanding of categories of measurement in the evaluation of asset-based approaches for health and wellbeing and to identify indicators and frameworks that can used in practice. The review was part of a project to develop an asset-based theory of change in health, care and wellbeing, funded by the Health Foundation.Methods Rapid Evidence Assessment techniques were used to produce a map of evaluation approaches and measures. The search strategy used strings of common terms (eg Salutogensis, Asset Based Community Development) combined with synonyms of measurement/evaluation. Academic databases were searched from 2005 along with relevant websites. Additionally, the results of a previous search were screened and reference lists mined. Publications were included if they explicitly discussed the asset model and measurement.  Data were then extracted across various fields including population, conceptual framework, measurement type and application. Findings were summarised in tables, with some additional thematic analysis.Results 8689 publications were identified and 91 full text articles assessed for eligibility. In total, 33 publications were included in the review. Many of these flagged up the challenges of evaluation and the need for better outcome measurement. Reported research strategies varied considerably from the practical to the theoretical. Also the purpose, level and specificity of measurement ranged from validated indicators through to broad domain frameworks.  Seven clusters were identified: Asset Based Community Development; Asset Mapping; Community-based evaluation; Conceptual frameworks for measurement; Resilience; Salutogensis; Other. One key theme was the importance of communities being involved in the assessment of assets.Conclusion The review provides an overview of research strategies and measures with application and purpose described. Producing a map of how asset-based approaches are evaluated and articulating key methodological differences helps researchers and practitioners select appropriate evaluation methods.  There are a number of limitations, including the use of rapid review methods which may have missed other relevant evaluation approaches. Further methodological development is needed in this field and we welcome debate about ways to evaluate asset-based approaches. Keywords: health assets; evaluation; methodology

OP78

NEIGHBOURHOOD SOCIAL COHESION, ETHNICITY AND PHYSICAL ACTIVITY IN ADOLESCENTS: LONGITUDINAL EVIDENCE FROM THE ORIEL STUDY

N Berger1, DJ Lewis1, EN Njagi2, S Cummins1

1Social and Environmental Health Research, London School of Hygiene and Tropical Medicine , London, UK2Non-communicable Disease Epidemiology, London School of Hygiene and Tropical Medicine , London, UK

Background Most adolescents do not achieve the recommended level of physical activity (PA) in the UK. Cultural norms – captured by ethnic identity – and social cohesion are aspects of the social environment that have the potential to influence health (behaviours). This study examines the relationships between social cohesion, ethnicity and three common types of PA in adolescents. The objectives are to test whether different types of PA have similar patterns of associations with social cohesion and ethnicity and to investigate confounding and interaction effects. 

Methods We used longitudinal data from the Olympic Regeneration in East London (ORiEL) study. In 2012, 3,088 adolescents aged 11-12 were recruited from 25 schools in four deprived and ethnically diverse boroughs of East London. Adolescents were followed-up in 2013 and 2014. Social cohesion was operationalised as trust in people living in the neighbourhood, measured on a four-point scale at wave 2. We grouped ethnic identities into eight categories. The outcomes were self-reported binary variables:  walking to school, walking for leisure, and outdoor PA in the neighbourhood. We identified potential confounders. We fitted Generalised Estimated Equation models for each outcome, with a time trend and investigated potential interactions between the exposures and time. Final available case sample size was 4,811 observations from 2,157 adolescents. Results We found evidence of associations between ethnicity and the PA outcomes. Adjustment for potential confounders (socioeconomic status, health and gender) did not attenuate the relationships. In the fully adjusted models, compared to White British, Black Africans had 1.49 higher odds of outdoor PA (95% CI 1.06-2.07), Bangladeshis had 1.42 higher odds of walking to school (95% CI 1.02-1.97) and all ethnic groups had lower odds of walking for leisure (estimated ORs varied between 0.37 and 0.71). While social cohesion was not associated with walking, there was evidence of a dose-response relationship with outdoor PA. A one-category increase in cohesion increasedthe odds of outdoor PA by 1.18 (95% CI 1.07-1.30). We found no evidence of time*exposure interaction; exposures did not predict change in PA.Conclusion This study contributes new findings to the evidence base on the social environment and PA. The ethnic diversity of the ORiEL study and its large sample size provided sufficient power to reveal how PA behaviours are patterned by ethnic groups. Further analyses will jointly model the three PA outcomes to better capture the dependency and associations between the exposures and the outcomes. Keywords: physical activity, ethnicity, social cohesion

OP79

INDIVIDUAL AND NEIGHBOURHOOD-LEVEL SOCIOECONOMIC FACTORS AND INCIDENCE OF TYPE 2 DIABETES IN OLDER AGE: RESULTS FROM A 14 YEAR FOLLOW-UP OF A COHORT OF OLDER BRITISH MEN

SE Ramsay1, D Roberts2, AO Papacosta3, LT Lennon3, PH Whincup5, SG Wannamethee3

1Institute of Health & Society, Newcastle University, Newcastle upon Tyne, UK2Whittington Health NHS Trust , Whittington Health NHS Trust , London, UK3Department of Primary Care & Population Health, UCL, London, UK4Population Health Research Institute, St George's, University of London, London, UK

Background Current evidence linking socioeconomic factors to incident Type 2 Diabetes Mellitus (T2DM) in older populations is conflicting. We investigated the prospective association of individual socioeconomic position and neighbourhood-level socioeconomic deprivation with incident T2DM in older British men, and examined possible underlying factors.Methods A socially-representative cohort of 3487 men, aged 60-79 years in 1998-2000, from 24 British towns was followed-up for 14 years for incident cases of T2DM. Individual socioeconomic position was based on social class derived from the longest-held occupation in middle-age, and was categorised into non-manual and manual groups. Neighbourhood-level socioeconomic deprivation was based on national Index of Multiple Deprivation (IMD) quintiles; a composite score of neighbourhood-level factors (income, employment, education, disability, crime, housing and living environment), with a higher score indicating greater deprivation. Follow-up on type 2 diabetes was obtained from reviews of general practitioner records and self-reported from questionnaires. Cox proportional hazards models were used to obtain hazard ratios (HR) and 95%CI for incident

diabetes according to social class and IMD quintiles. Prevalent cases of diabetes at baseline were excluded from the analyses.Results During the follow-up of 14 years, there were 289 incident cases of T2DM (7.1 per 1000 person-years). Diabetes risk increased from higher to lower social class groups and from IMD quintile 1 (least deprived) to quintile 5 (most deprived) (P for trend=0.001). Compared with non-manual social class groups, age-adjusted HR for manual groups was 1.58 (95%CI 1.24-2.01) – this was largely attenuated (1.38; 95%CI 1.08-1.76) on adjustment for body mass index (BMI); adjustment for blood pressure, smoking, alcohol, physical activity, diet, medication and family history resulted in little attenuation while further adjustment for triglyceride levels attenuated the association. Compared with IMD quintile 1, the risk of incident T2DM was highest in IMD quintile 4 (HR=1.79; 95%CI 1.24-2.54). This largely attenuated on adjustment for BMI (HR=1.46; 95%CI 1.02-2.10), and became non-significant after adjustment for lifestyle factors (smoking, alcohol, physical activity and diet).Conclusion Manual social class and neighbourhood-level socioeconomic deprivation was associated with an increased risk of T2DM in older British men. For social class this was mostly explained by BMI and triglycerides. For neighbourhood-level socioeconomic deprivation it was largely explained by BMI and lifestyle factors. Our results support the need for public health initiatives specifically targeting obesity as a means towards reducing socioeconomic inequalities in type 2 diabetes in later life.Keywords: Health inequalities, diabetes, older people.

OP80

ASSOCIATION OF OBJECTIVE AND PERCEIVED NEIGHBOURHOOD CHARACTERISTICS WITH POOR ORAL HEALTH IN OLDER AGE: RESULTS FROM A CROSS-SECTIONAL STUDY OF OLDER BRITISH MEN

SE Ramsay1, E Papachristou2, AO Papacosta2, LT Lennon2, PH Whincup3, SG Wannamethee2

1Institute of Health & Society, Newcastle University, Newcastle upon Tyne, UK2Department of Primary Care & Population Health, UCL, London, UK3Population Health Research Institute, St George's, University of London, London, UK

Background Socioeconomic inequalities in oral health are established. However, the influence of neighbourhood-level socioeconomic factors on the oral health of older people is not well-established. We investigated both objective and perceived neighbourhood characteristics and their associations with a range of oral health measures in older age.Methods The British Regional Heart Study comprises a representative sample of men drawn from 24 general practices across Britain at 40-59 years in 1978-80. In 2010-12, the participants when aged 71-92 years had a follow-up a physical examination including a dental assessment (n=1622), and completed a questionnaire (n=2137). Oral health assessment included objective measures (tooth count and periodontal (gum) disease), and self-reported fair/poor oral health and dry mouth symptoms. Neighbourhood deprivation was based on the Index of Multiple Deprivation (IMD); a composite score of neighbourhood-level factors (income, employment, education, disability, crime, housing, living environment). Perceived neighbourhood characteristics included local area services, safety, environment, and a cumulative index of these characteristics. Multilevel and multivariate logistic regression models were used to obtain odds ratios according to quintiles of IMD and perceived neighbourhood characteristics. Results The risk of periodontal disease and tooth loss increased from IMD quintile 1 (least deprived) to 5 (most deprived); age-adjusted odds ratios (OR) for quintile 5 were 3.25 (95% CI 2.05–5.17) and 3.58 (95% CI 2.38–5.39) respectively, compared to quintile 1. These associations were attenuated only slightly on adjustment for individual social class, smoking, depression, social interactions and history of cardiovascular disease or diabetes, and remained statistically significant.

Age-adjusted odds of dry mouth was increased only in quintile 2 (OR = 1.41, 95% CI 1.04–1.91) and quintile 5 (1.50, 95%CI 1.09-2.07) compared to quintile 1 and was not significant after adjustments for the remaining covariates. The odds of self-reported fair/poor oral health was greater only in quintile 5 (OR=1.73, 95%CI 1.28–2.35), and remained statistically significant after adjustment for covariates. For perceived neighbourhood characteristics, significant trends were observed across quintiles of local area services, safety and a cumulative index of neighbourhood characteristics, with greater levels of tooth loss, periodontal disease, fair/poor self-rated oral health and dry mouth from quintile 1 (best rated) to quintile 5 (worse rated).Conclusion Markers of poor oral health in older age were associated with both objective and perceived neighbourhood-level socioeconomic factors. Wider socioeconomic determinants are potentially important influences on the oral health of older people. Prospective studies are needed to establish these associations.Keywords: Oral health, deprivation, older age

Friday 8 September 2017

Physical activity interventions

OP81

LONG-TERM OBJECTIVE PHYSICAL ACTIVITY DATA FROM TWO PRIMARY CARE PEDOMETER-BASED RANDOMISED CONTROLLED TRIALS IN MIDDLE-AGED AND OLDER ADULTS– ARE THERE STILL POSITIVE TRIAL EFFECTS AT 3 AND 4 YEARS?

E Limb1, T Harris1, S Kerry2, C Victor3, S Iliffe4, M Ussher1, P Whincup1, C Furness1, C Wahlich1, D Cook1, U Ekelund5, J Fox-Rushby6, J Ibison1, S DeWilde1

1Population Health Research Institute, St George's University of London, London, UK2Pragmatic Clinical Trials Unit, Queen Mary's University of London, London, UK3Gerontology & Health Services Research Unit, Brunel, University of London, London, UK4Research Department of Primary Care and Population Health, University College, London, London, UK5Department of Sports Science, Oslo University, Norway6Health Economics Research group, Brunel, University of London, UK

Background The PACE-UP 3-armed primary care trial recruited 45-75 year olds into a 12-week pedometer-based intervention, with one postal intervention arm and one nurse support. The PACE-Lift 2-armed primary care trial recruited 60-75 year olds into a 12-week nurse-supported pedometer-based intervention. Both trials increased step-counts by around a tenth and time in moderate-to-vigorous physical activity (MVPA) in bouts by around a third at 12 months, with no difference between nurse and postal arms in PACE-UP. Long-term physical activity (PA) maintenance, particularly MVPA in bouts, is important for a wide range of health benefits, but few trials provide objective PA measures beyond 12 months. We followed up PACE-UP and PACE-Lift cohorts at 3 and 4 years respectively, to investigate whether intervention effects persisted.Methods 3-year (PACE-UP) and 4-year (PACE-Lift) accelerometer outcomes were regressed on baseline accelerometry to estimate change in average daily step-counts and average weekly time in MVPA in ≥10 minute bouts in the treatment groups compared to control groups. Imputation analyses were conducted to account for missing data effects.Results PACE-UP 3-year follow-up rate was 67% (681/1023). Both intervention groups were still doing more steps/day than the control group:  postal 627 (95% CI 198, 1056); nurse 670 (95% CI 237, 1102); nurse plus postal 648 (95% CI 272, 1024). The pattern was similar for total weekly MVPA in bouts: postal 28 (95% CI 7, 49); nurse 24 (95% CI 24 (3, 45); nurse plus postal 26 (95% CI

8, 44). PACE-Lift 4-year follow-up rate was 76% (225/298). In PACE-Lift the intervention versus control comparisons were 407 (95% CI -177, 992) for steps and 32 (95% CI 5, 60) for MVPA in bouts; though the steps comparison is not statistically significant, both the steps and MVPA estimates are consistent with the PACE-UP findings.  Both trials showed no wear-time differences between groups and imputation analyses did not change results interpretation.Conclusion Over two thirds of both trial cohorts provided long-term data. Both trials showed persistent effects on time in MVPA in bouts at 3 or 4 years. PACE-UP showed a significant persistent effect on step-counts and no difference between nurse and postal group outcomes.  In PACE-Lift the step-count effect, while not statistically significant was consistent with both the PACE-UP findings and with the persistent significant effect for MVPA in bouts in PACE-Lift.  These findings suggest that primary care pedometer interventions, delivered by post or with minimal contact, can make an important long-term contribution to addressing the public health physical inactivity challenge.Keywords: physical activity RCT follow-up

OP82

SYSTEMATIC REVIEW OF DIGITAL INTERVENTIONS TO IMPROVE THE DIET AND PHYSICAL ACTIVITY BEHAVIOURS OF ADOLESCENTS

T Rose1, M Barker1,2, CM Jacob3, W Lawrence1,2, S Strommer1, C Vogel1,2, K Woods-Townsend2,4, H Inskip1,2, J Baird1,2

1MRC Lifecourse Epidemiology Unit, University of Southampton, Southampton UK2NIHR Southampton Biomedical Research Centre, University Hospital Southampton NHS Foundation Trust, University of Southampton, Southampton UK 3Academic unit of Human Development and Health, Faculty of Medicine, University of Southampton, Southampton UK 4Southampton Education School, University of Southampton, Southampton, UK

Background Adolescents have poor diets and insufficient physical activity (PA) levels, which can lead to non-communicable diseases in later life. Adolescence is a key stage for establishment of lifelong health behaviours. Digital platforms offer inexpensive means of delivering health interventions to large numbers of adolescents, but little is known about their effectiveness in supporting sustained changes. This systematic review was conducted to synthesise evidence on the effectiveness of digital interventions to improve diet quality and increase PA in adolescents, to identify effective intervention components and to assess the cost-effectiveness of these interventions. Methods A systematic review of digital interventions to improve the diet and/or physical activity behaviours of adolescents was carried out according to established guidance. A search of MEDLINE, PsycINFO, CINAHL, PubMed Central, Embase, ERIC, the NHS EED and CENTRAL was conducted using a combination of MeSH and free text terms. Abstracts were assessed by two independent researchers against review inclusion criteria (intervention studies with and without a control group, participants aged 10-19, interventions that included a digital component, studies that measured a diet or PA outcome at baseline and follow-up, any setting and any population of adolescents). For included studies, data extraction and quality assessment were performed using a form designed to capture key information to answer the research questions. We then analysed data to identify key intervention features that were associated with significant improvements in behaviour. We also aimed to collect data on cost-effectiveness of included interventions. Results The search returned 6792 results, of which 86 full texts were selected for screening. After assessing these against eligibility criteria, 27 intervention studies were included in the review. Most (n=15) were website interventions. Other delivery methods were text messages (n=4), gamified

interventions (n=3), multi-component interventions that included a digital element (n=3), email (n=1), and social media (n=1). Significant behaviour change was often seen when interventions included diet and physical activity education, goal-setting in conjunction with self-monitoring, and parental involvement. None of the included publications reported cost-effectiveness. Due to heterogeneity of studies, meta-analysis was not feasible. Conclusion It is possible to effect significant health behaviour change in adolescents through targeted digital interventions that incorporate education, goal-setting, self-monitoring and parental involvement. Most of the evidence relates to websites and further research is needed to assess effectiveness of interventions delivered via alternate media such as smartphone apps. Longer term outcomes should be evaluated. Future trials of digital interventions need to evaluate cost-effectiveness. Keywords: Adolescent Diet Exercise

OP83

DIET AND PHYSICAL ACTIVITY-BASED WEIGHT MANAGEMENT PROGRAMMES FOR ADULTS: RE-ANALYSIS OF A SYSTEMATIC REVIEW THROUGH THE LENS OF USER VIEWS TO IDENTIFY PATHWAYS TO EFFECTIVENESS

GJ Melendez-Torres1, K Sutcliffe2, HED Burchett3, R Rees2, M Richardson2, J Thomas2

1Warwick Medical School, University of Warwick, Coventry, UK2EPPI-Centre, UCL Institute of Education, London, UK3Policy Innovation Research Unit, London School of Hygiene and Tropical Medicine, London, UK

Background Previous systematic reviews of weight management programmes have not been able to account for heterogeneity of effectiveness within programmes using top-down behavioural change taxonomies and standard meta-regression methods. This could be due to the complex nature of these interventions, the intersecting aspects of different intervention components, and the presence of overlapping causal pathways to effectiveness (or lack of effectiveness). Qualitative comparative analysis is a method that, when used in the synthesis of complex interventions, can help identify the ‘how’ and ‘why’ of intervention effectiveness while accounting for these overlapping and intersecting causal pathways.Methods We updated a previous systematic review of studies relating to the views of programme users and trials of weight management programmes in adults that included dietary and physical activity components. We thematically analysed views studies to identify intervention features perceived to be important by service users and providers. We selected the most and least effective trials as measured by amount of weight loss at 12 months compared to minimal treatment. Using intervention components suggested by the synthesis of user views, we labelled interventions as to the absence or presence of these components and, using qualitative comparative analysis, we identified combinations of components that created the conditions sufficient for interventions to be most effective and least effective.Results Our analysis included 21 studies of user views reported in 26 papers, and 20 interventions—10 most effective and 10 least effective—reported in 15 trials. The synthesis of user views suggested three facets of interventions for analysis: provider-user alliance, or the quality of the relationship between users and providers; provider directiveness, or the perceived high need for guidance from providers; and group relationships, or the quality of the relationship between peers in weight management programmes. Across our analyses, we found that pathways to intervention effectiveness were characterised by the presence of opportunities to develop supportive relationships with providers or peers, clear direction from providers together with components to develop self-regulation of exercise and dietary behaviour.Discussion Revisiting complex interventions through a lens that accounts for complexity can facilitate action on a pressing public health problem like obesity and overweight. Though QCA is an

inductive method, this innovative approach has enabled the identification of neglected critical aspects of WMPs, such as the nature of relationships within them, which were previously not considered to be as important as more concrete content such as dietary focus.Keywords: diet and physical activity, user views, systematic review

OP84

ARE WE THERE YET? A CUMULATIVE META-ANALYSIS OF THE ACCUMULATION OF EVIDENCE FOR PHYSICAL ACTIVITY PROMOTION

R Love1,2, E van Sluijs1, J Adams1, D Humphreys2

1Centre for Diet and Activity Research (CEDAR), MRC Epidemiology Unit, University of Cambridge School of Clinical Medicine, Cambridge, United Kingdom2Department of Social Policy and Intervention, University of Oxford, Oxford, United Kingdom3Department of Social Policy and Intervention, University of Oxford, Oxford, United Kingdom

Background Despite a large evidence base and increasing number of physical activity interventions, rates of physical inactivity and associated chronic diseases are rapidly increasing globally. It is an imperative time to examine if new intervention trials have contributed novel evidence to the field. The purpose of this overview of systematic reviews is to investigate the evolution of evidence for physical activity promotion interventions over time through the analysis of changes in effect sizes for physical activity (PA) and cardiorespiratory fitness (CR) outcome measures. Methods The search process was conducted in two stages. In the first phase, the Cochrane Central Register of Controlled Trials, MEDLINE, EMBASE, CINAHL, PsycINFO and Web of Science were searched for systematic reviews that examined the effects of interventions targeted at increasing physical activity within individuals, and included randomized controlled trials. In the second phase, primary studies included within the eligible systematic reviews were pooled. Interventions aiming to increase physical activity in healthy adults at the individual level, assessed through randomized controlled trials, were included. A cumulative meta-analysis was performed separately for interventions with PA and CR outcome measurements. Results When assessed through cumulative meta-analysis, the totality of the evidence demonstrates that intervention effectiveness has not changed very much over the past fifteen years and that further trials are unlikely to overturn the positive and stabilized findings. The analyses indicate that the effect size for interventions became precise and stable in 2001 after the conduct of 12 physical activity (PA) and 11 cardiorespiratory (CR) outcome trials. In the time following the attainment of these thresholds of stability and precision, 70 further full scale RCTs (59 for PA and 11 for CR outcomes) have been conducted.Conclusion Substantial evidence exists demonstrating that physical activity interventions can modify individual behavior. However, there is limited evidence of advancements in intervention effectiveness. Given the stability and sufficiency of the evidence base, researchers are urged to shift focus towards the optimization and scaling-up of interventions across populations.Keywords: Scientific progress, Cumulative meta-analysis, Randomised controlled trials

Methods

OP85

A QUALITATIVE EXPLORATION OF TRIAL-RELATED TERMINOLOGY IN A STUDY INVOLVING DEAF BRITISH SIGN LANGUAGE USERS

A Young1, C Dodds1, R Oram1, C Nassimi-Green1, R Belk1, K Rogers1, L Davies2, K Lovell31Social Research with Deaf People, School of Health Sciences, University of Manchester and MAHSC, Manchester, UK2Health Services Research and Primary Care, Institute of Population Health, University of Manchester and MAHSC, Manchester, UK3Division of Nursing, Midwifery and Social Work, University of Manchester and MAHSC, Manchester, UK

Background British Sign Language (BSL) is a fully grammatical, living language, independent of spoken English. Deaf BSL users are routinely excluded from clinical trials either on grounds of language and/or assumptions of confounding factors associated with disability.  Few clinical trials specifically targeted at Deaf people exist internationally.  Appropriate terminology in BSL for key concepts in clinical trials is needed to support informed consent.. Barriers to conceptual (mis)understanding of trial participation and sources relevant to the Deaf community are unexplored.  This study set out to:(i) Explore, in BSL, the meaning and understanding of key concepts and common vocabulary used in recruitment and informed consent to a clinical trial(ii) Enable signs/signed expressions to emerge that are semantically accurate and support Deaf people’s informed consent in any future trialMethods A community participatory approach underpinned data collection. This comprised Deaf-led, qualitative, exploratory group discussions (19 self-selected participants in five groups), facilitated in BSL.  Discussion was structured to explore the meaning of key terminology (such as trial, randomisation, consent) and through that exploration generate signed explanations of concepts to use in future trials. Data were video-recorded and analysed in BSL, using a phenomenological approach, built from the application of 4 themes and 8 sub-themes using NVIVO8 for visual data. Results All participants were over 30 years old, with a self-declared ‘strong Deaf identity’; BSL was their first or preferred language. Six necessary conditions for developing optimal trial information were identified.  (i) developing appropriate expressions and terminology from a community basis, rather than testing out previously derived translations from a different language; (ii) paying attention to language-specific features which support best means of expression (in the case of BSL expectations of specificity, verb directionality, handshape); (iii) accounting for bilingual influences on comprehension; (iv) deliberate orientation of information to avoid misunderstanding not just to promote accessibility; (v) sensitivity to barriers to discussion about intelligibility of information that are cultural and social in origin, rather than linguistic; (vi) the importance of using contemporary language-in-use, rather than jargon-free or plain language, to support understanding.Conclusion These conditions need to be met to develop signed participant information for Deaf people that is: acceptable, accessible, transmitted accurately and understood as intended. They are required to address the cultural preferences and lower background knowledge of Deaf people (using signed languages) as well as in spoken/written languages when participants’ language use is different from the dominant language of the country.Keywords: informed consent - trials - sign language

OP86

UNDERSTANDING INTERVENTION ACCEPTABILITY: CHANGING PARTICIPANT VIEWS OVER TIME AND THE IMPORTANCE OF CONTEXT

S Morgan-Trimmer1, J Gregory2, J Townson3, S Channon3, I Chestnutt4, B Chadwick4, S Murphy5, S Hutchings6

1Psychology Applied to Health Group, University of Exeter Medical School, Exeter, UK2School of Medicine, Cardiff University, Cardiff, UK

3Centre for Trials Research, Cardiff University, Cardiff, UK4School of Dentistry, Cardiff University, Cardiff, UK5Centre for the Development and Evaluation of Complex Interventions for Public Health Improvement (DECIPHer), School of Social Sciences, Cardiff University6Simbec-Orion Group Ltd

Background Acceptability is an important aspect of the quality of health interventions and also has implications for the feasibility of future implementation. Process evaluations of complex interventions often assess the acceptability of interventions through qualitative interviews with participants, addressing participants’ experience of, satisfaction with and preferences for treatment/services received. Acceptability can be influenced by multiple aspects of an intervention and its context. This paper reflects on two process evaluations which produced complex findings on acceptability: the Delivering Early Care In Diabetes Evaluation (DECIDE) study, a trial of home vs. hospital care following a diagnosis of Type I diabetes (T1D) in children, and the Seal or Varnish (SoV) study which compared fluoride varnish and fissure sealant dental treatments to prevent caries in children.Methods In the DECIDE study, semi-structured interviews with 11 (pairs of) parents and seven children were conducted 15-20 months post-diagnosis about their experience of hospital or home care immediately following diagnosis of T1D. In the SoV study, all children completed a ‘smiley face’ acceptability questionnaire immediately after treatment. In addition, paired semi-structured interviews were conducted with children at the beginning (50 children) and end (32 children) of the intervention, within a few days of receiving treatment. Interviews were divided evenly between trial arms; questions included experience of and acceptability of the interventions, diabetes management (DECIDE) and management of dental health (SoV).Results In the DECIDE study, most interviewees wanted to be randomized to the ‘home’ arm initially but expressed a retrospective preference for whichever trial arm they had been in. This shift in preference may have been influenced by ‘positive attitude’ coping strategies adopted by families. In the SoV study, acceptability immediately post-treatment was related to which treatment was received, but acceptability in the week following treatment was strongly influenced by wider aspects of treatment such as receiving a sticker and there was little difference in overall acceptability by trial arm. Perceptions of overall treatment may have been influenced by the intervention being delivered through a well-established, child-friendly dental service in a school setting.Conclusion Both studies found that acceptability of an intervention can change over time, and indicated that participant restructuring of acceptability can be influenced by wider contextual factors of the intervention. Implications for future research are that the timing of data collection on acceptability may influence findings, and that acceptability (and implications for future implementation) should be interpreted in relation to intervention context.Keywords: Acceptability; process evaluation

OP87

DAGITTY AND DIRECTED ACYCLIC GRAPHS IN OBSERVATIONAL RESEARCH: A CRITICAL REVIEW

PWG Tennant1,2, J Textor3, MS Gilthorpe1,4, GTH Ellison1,4

1Leeds Institute for Data Analytics, University of Leeds, Leeds, UK2School of Healthcare, University of Leeds, Leeds, UK3Department of Tumour Immunology, Radboud University Medical Center, Nijmegen, The Netherlands4Leeds Institute for Cardiovascular and Metabolic Medicine, University of Leeds, Leeds, UK

Background Empirical researchers working with observational data have been slow to adopt modern statistical methods for causal inference, which remain poorly recognised among applied quantitative researchers. First introduced in 2010, DAGitty is a free web application (and R package) that enables empirical researchers to draw directed acyclic graphs (DAGs) and identify minimally-sufficient adjustment sets without explicit knowledge of graphical model theory. This review examines empirical research articles that have used DAGitty as an aid for analysing observational data.Methods Articles citing 'DAGitty' published before 1 July 2016 were identified through searching Web of Science, Medline, Scopus, PubMed, and Google Scholar. Original articles describing the analysis of observational data were identified by inspecting the published manuscripts. Information on the use and presentation of DAGs and adjustment sets were extracted into a standardised table. Bibliographic details (including journal discipline) were obtained from Thompson-Reuter's Journal Citations Reports.Results 124 original articles describing the analysis of observational data were identified from 151 unique articles citing DAGitty. Two (2%) were published in 2012, seven (6%) in 2013, 23 (19%) in 2014, 46 (37%) in 2015, and 46 (37%) in the first half of 2016. The first authors came from 18 countries, most commonly the USA (n=36, 29%), Germany (n=19, 15%), Australia (n=14, 11%), Sweden (n=12, 10%), the UK (n=10, 8%), and Denmark (n=6, 5%). The host journals represented 43 academic disciplines, most commonly 'Public, environmental, and occupational health' (n=29, 23%), 'environmental studies' (n=13,10%), 'multidisciplinary sciences' (n=11, 9%), 'oncology' (n=10, 8%), 'nutrition and dietetics' (n=9, 7%), and 'immunology' (n=8, 6%). 29 (23%) articles included a DAG in the manuscript, 41 (33%) in supplementary material, while 53 (44%) contained no DAG. DAGs varied greatly in scope from three-variable overviews to graphs with 30+ variables. Very few DAGs were saturated, whether completely or in order of transit. At the extreme, some researchers omitted all arcs except those that were explicitly evidenced. Adjustment sets were often modified beyond minimally-sufficient set(s) by adding: competing exposures (for 'improve precision'), mediators (to 'improve face validity'), and interaction terms; or by removing variables using stepwise (p-value) methods or criteria for 'minimum change'.Conclusion Use of DAGitty in empirical research is increasing exponentially. There is however huge variation in practice, with many choosing to blend DAG-based methods with more traditional/accepted approaches to model specification. Guidelines for 'best practice' should be developed and included in teaching material and/or journal guidelines.Keywords: Methods, Causal Inference, Graphical Model Theory, Directed Acyclic Graphcs

OP88

ASSESSING THE POTENTIAL UTILITY OF ‘BIG DATA’ FROM THE PRIVATE SECTOR FOR HEALTH RESEARCH: LINKING EXPERIAN™ MOSAIC GROUPS TO DEPRIVATION INDICES

WM Wami, OR Molaodi, R Dundas, AH Leyland, SV Katikireddi1MRC/CSO Social and Public Health Sciences Unit, University of Glasgow, Glasgow, UK

Background Socioeconomic circumstances are routinely measured using government-held data e.g. the Index of Multiple Deprivation (IMD) for health research and service planning. However, alternative approaches may be necessary as key datasets (e.g. the decennial census and specific welfare claims) may not be available in the future. ‘Big data’ from the private sector could help meet this need. Experian collates diverse information to produce socio-demographic and lifestyle variables, primarily for marketing purposes. Unlike traditional deprivation measures, these variables are intended to reflect affluence and consumption patterns. We assessed the overlap between deprivation measures and Experian’s British population segmentation (Mosaic Groups) to establish the utility and feasibility of bringing together commercial and health data for public health purposes. 

Methods Experian Mosaic classifies all UK consumers into 15 distinct groups at the postcode level which comprehensively describes their socio-economic and sub-cultural behaviour, based on 1,754,408 unique postcodes (mean population 39). Thus, each postcode can be allocated to only one Mosaic Group. These data were linked to widely used deprivation measures, i.e. Index of Multiple Deprivation (IMD) and its devolved equivalents for Scottish (SIMD) and Welsh (WIMD) populations and Carstairs quintiles ranked from 1 to 5 (most to least deprived) based on full postcode. Cochran-Armitage tests were used to determine if there were significant trends in the proportions of deprivation quintiles within each Mosaic Group. Results The Experian segmentation showed good population coverage; 92% of the UK population was included in the dataset. IMD quintiles and Mosaic Groups were associated (all P-values for trend <0.05); for example, the proportions increased with deprivation in Municipal Challenge (0.1% and 73.6% in least and most deprived quintiles) and Family Basics (1.1% and 50.2%, respectively). For some Mosaic Groups the proportions increased with decreasing deprivation, e.g. Prestige Positions (0.3% and 67.1% in most and least deprived, respectively). However, there were Mosaic Groups (e.g. City Prosperity, Country Living) that showed no consistent trend with deprivation, indicating Mosaic is capturing a different aspect of socio-economic circumstances. Similar patterns were found for Carstairs and measures of IMD for Scottish and Welsh populations.Conclusion Experian Mosaic Groups appear to capture different aspects of socioeconomic circumstances to deprivation measures. A key limitation of our study to date is the lack of investigation of health outcomes. These marketing data may provide new insights into the social determinants of health at a very small area level and could help plan service delivery. Keywords: public health, deprivation, epidemiology

Gender

OP89

PRE-PREGNANCY AND POSTNATAL DEPRESSIVE SYMPTOMS ARE ASSOCIATED WITH QUALITY OF MOTHER-CHILD RELATIONSHIPS; LONGITUDINAL DATA FROM THE SOUTHAMPTON WOMEN’S SURVEY

J Baird1, M Barker1,2, W Lawrence1,2, T Kendrick3, S Crozier1, SM Robinson1,2, C Cooper1,2, KM Godfrey1,2, HM Inskip1,2

1MRC Lifecourse Epidemiology Unit, University of Southampton, Southampton, UK2NIHR Southampton Biomedical Research Centre, University of Southampton and University Hospital Southampton NHS Trust, Southampton, UK3Primary Care and Population Sciences Academic Unit, University of Southampton, Southampton, UK

Background Maternal mental health difficulties have been associated with poor outcomes for children, possibly because they affect the quality of the relationship between mother and child. We examined associations between maternal depressive symptoms before pregnancy, and again 6 months postnatally, and quality of the mother-child relationship in the Southampton Women’s Survey (SWS).Methods The SWS is a population based cohort study of 12,583 women aged 20-34 years who were assessed when not pregnant; those enrolled during the latter half of the recruitment period completed the General Health Questionnaire (GHQ-12), a short screening instrument with good sensitivity for depression and anxiety disorders. Women who became pregnant and their children (n=3158) were followed up. Six months post-partum, mothers completed the Edinburgh Postnatal Depression Scale (EPDS) questionnaire to determine depressive symptoms. Established cut-offs for each scale were used to determine the presence or absence of depressive symptoms at the two

time points. A summary pre-pregnancy and postnatal depressive symptoms 4-level categorical variable was derived indicating symptoms at neither, one or other, or both periods. When the child was 3 years old, mothers completed the Pianta Child-Parent Relationship Scale (CPRS), from which closeness and conflict scores were derived. To normalise these scores a Fisher-Yates transformation was used so that the scores are presented in standard deviation (SD) units. Linear regression was used to relate the two mother-child relationship measures to depressive symptoms, adjusting for confounding factors identified using a Directed Acyclic Graph: receipt of benefits, employment, parity and educational attainment.Results Among women with GHQ-12 and EPDS data, 1,441 completed the Pianta CPRS. Presence of depressive symptoms was identified before pregnancy in 28% of mothers and postnatally in 42%. After adjustment for confounders, and compared with those with no symptoms during either time period, those who only had depressive symptoms postnatally had conflict scores that were elevated by 0.28SD (95%CI: 0.16-0.41), whereas scores for those with depressive symptoms at both time periods were elevated by 0.53SD (95%CI: 0.39-0.68). Closeness scores were reduced in those with symptoms at both time periods, being 0.26SD (95%CI: 0.08-0.43) lower than for those with no symptoms, but were not reduced in those with depressive symptoms at one or other period only.Conclusion Women who suffer from depressive symptoms both before and after pregnancy appear vulnerable to poorer relationships with their children. This emphasizes the importance of maternal mental health for the mother and for her relationship with her child.Keywords: Maternal depression, preconception, child behaviour

OP90

THE EMBODIMENT OF GENDER DIVISIONS OF LABOUR: GENDER DIVISIONS OF LABOUR AND INFLAMMATORY MARKERS IN THE UK HOUSEHOLD LONGITUDINAL STUDY

A McMunn, L Bird, E Webb, A Sacker1Epidemiology & Public Health, UCL, London, UK

Background While the gender division in paid work has progressively diminished, this is less true of unpaid labour. A gender perspective requires an extension of investigations of stress and health beyond paid work to include unpaid labour. This study investigates associations between gender divisions of ‘total labour’ and inflammatory markers thought to potentially link chronic stress exposure with disease outcomes amongst contemporary UK couples.Methods The study sample includes co-resident, opposite sex couples aged 16-65 in Wave 2 of the UK Household Longitudinal Study or Understanding Society. Gender divisions of labour were measured using the number of weekly hours that each member of the couple spent in paid work, housework, and caring, as well as who has responsibility for children. Inflammatory marker outcomes included C-Reactive Protein (CRP) and fibrinogen. High CRP was defined as greater than 3 mg/L. Cross-sectional multivariate regression models were used to investigate associations between gender divisions in paid work, caring, housework and childcare with inflammatory markers, adjusting for household income, educational qualifications, longstanding illness and age, stratifying by gender to investigate whether associations are differential within couples.Results Gender divisions of labour were associated with raised inflammation for men but not women. Men living in households in which women were providing care to an adult or doing all of the housework had significantly higher levels of Fibrinogen (caring: coef = 0.11, 95% CI = 0.04-0.19; housework: 0.08, 0.02-0.14) and CRP (caring: OR= 1.66, 95% CI = 1.15-2.39; housework: OR = 1.50, 1.06-2.14) compared with men in households in which neither partner provided care or both did few hours of housework. In addition, men in traditional ‘male breadwinner’ households, or childless households, were significantly less likely to have raised CRP levels (paid work: OR = 0.49,

0.36-0.68; parental status: 0.64, 0.44-0.92) than men in dual-earner households or those in which childcare was equally shared between parents.Conclusion Contrary to expectations, inflammatory markers were significantly associated with gender divisions of labour within couples for men and not for women. While potential stress-reducing benefits of participation in paid work and childlessness are aligned with prior research, further investigation is required to better understand the effects of women’s caring and housework on men’s inflammation. Next steps include investigation of the identity of care recipients, non-linear associations with hours spent in housework, and interaction effects between labour types.Keywords: Inflammatory markers, paid & unpaid work, gender, UK Household Longitudinal Study

OP91

Note: Presentation moved to Health Inequalities 2 session on WednesdayINFORMAL CAREGIVING AND MARKERS OF ADIPOSITY IN THE UK HOUSEHOLD LONGITUDINAL STUDY

R Lacey, A McMunn, E Webb1Department of Epidemiology and Public Health, University College London, London, UK

Background Providing informal care is known to be associated with poorer psychological health. However less is known about other, more objective health outcomes such as adiposity. The aim of this study was to investigate associations between being an informal caregiver with measures of adiposity using a large, representative UK longitudinal study.Methods Data on 20,669 participants of the UK Household Longitudinal Study were used to explore the relationship between caregiving and adiposity (body mass index (BMI) and waist circumference). We additionally investigated caregiving intensity (hours spent caregiving per week, number of people cared for and combining caregiving with paid work). Multiple imputation was used to account for missing data. Associations between caregiving and adiposity measures were tested using multivariable linear regressions. Analyses were stratified by gender and interactions with age were tested. Models adjusted for household income, educational attainment, social class, presence of a long-standing limiting illness, number of dependent children in the household and partnership status. All analyses included survey weights to account for design, unequal probabilities of selection, differential non-response and potential sampling errors. The analyses were conducted in Stata v14.Results Being an informal caregiver was associated with increased waist circumference (1.48cm, 95% CI: 0.42, 2.53) but not BMI for men in age-adjusted models. Women who were informal carers had higher waist circumferences (3.62cm, 95% CI: 2.77, 4.47) and BMIs (1.26, 95% CI: 0.89, 1.64) relative to non-carers. A caring-age interaction was present for women suggesting that younger women carers (aged 16-44) had particularly higher waist circumferences (5.44cm, 95% CI: 3.77, 7.10) and BMIs (1.90cm, 95% CI: 1.17, 2.62). Caregiving intensity was found to be important with increasing hours of caregiving associated with increasing adiposity.Discussion Being an informal carer was associated with increased adiposity amongst UK men and women. Caring appears to be particularly negatively associated with adiposity when occurring during non-normative life stages, such as early adulthood, and when high intensity. These findings are based on nationally representative longitudinal data. The main limitation of this study was the inability to investigate the reasons the care recipient requires care (i.e. dementia vs cancer), and to investigate the mechanisms involved. Given funding cuts for social care, advancements in medical treatment and increasing life expectancy, a greater proportion of the population will be expected to provide informal care for relatives and friends. The poorer health of carers should therefore be a priority for UK public health.Keywords: Caring; adiposity; family

OP92

A SYSTEMATIC REVIEW OF PROSPECTIVE RISK AND PROTECTIVE FACTORS FOR INTIMATE PARTNER VIOLENCE VICTIMISATION AMONG WOMEN

AR Yakubovich1, H Stöckl2, J Murray3, GJ Melendez-Torres4, JI Steinert1, CEY Glavin1, DK Humphreys1

1Social Policy & Intervention, University of Oxford, Oxford, United Kingdom2Global Health and Development, London School of Hygiene and Tropical Medicine, London, United Kingdom3Center for Epidemiological Research, Universidade Federal de Pelotas, Pelotas, Brazil4Warwick Medical School, University of Warwick, Warwick, United Kingdom

Background Rates of intimate partner violence (IPV) against women are unacceptably high worldwide. There has been no systematic review in over 10 years of all risk and protective factors without location or peer-review restrictions. Resultantly, there is no recent, systematically-developed model of the causes of IPV at all levels (individual, relationship, community, and structural) that accounts for differences, similarities, and evidence-gaps across low- to high-income contexts. This remains a barrier to the effective prevention of IPV, with significant uncertainty over what works and within which contexts. We aimed to systematically review all prospective, longitudinal risk and protective factors of IPV victimisation among women.Methods Systematic searches were conducted in 16 databases and references of relevant studies were hand-searched. Published or unpublished studies in English that prospectively analysed the association between any risk or protective factor(s) and self-reported IPV victimisation among women, controlling for at least one other variable, were included. Study quality was assessed using the Cambridge Quality Checklists. Study screening, extraction, and quality appraisal were completed and checked by three independent reviewers. Results were graphically synthesised using harvest plots, which allow for the synthesis of heterogeneous evidence and identification of trends towards negative, null, or positive associations.Results Searches retrieved 10,444 unique results. After title and abstract review, 387 studies were screened by full-text. Sixty studies from 35 cohorts met inclusion criteria. Most studies were from the USA (80.0%). A total of 71 risk/protective factors were identified, mostly at the individual- (N=21) or relationship-level (N=25) rather than the community- (N=7) or structural-level (N=18). Variables that showed positive or a mix of null-positive associations with women’s IPV victimisation were: at the individual-level, women’s identification as non-white, younger age, alcohol use, depressive symptoms, antisocial behaviour, aggressive personality, and experience of child abuse; at the relational-level, partners’ identification as non-white, alcohol use, antisocial behaviour, low relationship satisfaction, poor parental relationship quality, and experience of low parental monitoring; and at the structural-level, partners’ unemployment, women’s lower education, and financial difficulties. Other variables were under-studied (<2 studies) or showed mixed or mainly null effects.Conclusion Significant work is needed to develop an ecological model of IPV against women using prospective data. Many commonly accepted risk factors for IPV victimisation among women such as exposure to inter-parental violence and community factors have little (if any) prospective evidence from outside the USA. Further prospective research on the etiology of IPV against women is needed to inform rigorous prevention models.Keywords: violence prevention; systematic review; risk and protective factors

Food policy

OP93

STAKEHOLDERS' FRAMING OF EVIDENCE ABOUT THE UK SUGAR-SWEETENED BEVERAGE TAX: A NEWS MEDIA ANALYSIS

S Hilton1, CH Buckton1, SV Katikireddi1, F Lloyd-Williams2, C Patterson1, L Hyseni2, A Elliott-Green2, S Capewell21MRC/CSO Social and Public Health Sciences Unit, University of Glasgow, Glasgow, UK2Department of Public Health Policy, University of Liverpool, Liverpool, UK

Background In politically-contested health debates, such as sugar-sweetened beverage (SSB) taxation, stakeholders seek to present evidence and arguments for or against the specific policy initiatives, based on their interests. The news media play a crucial role in shaping public opinion by selectively choosing which messages to focus on. While the literature suggests that media debates should be a key concern for those interested in understanding public health policy processes, as yet there has been only limited research in this area. This study examined how stakeholders’ positions and evidence on SSB taxation were represented in the media to inform SSB advocacy strategies.Methods Quantitative and qualitative content analysis of 1,632 articles about sugar consumption and SSB taxation published in eleven national UK newspapers, chosen for diversity in political views and genre. We conducted a systematic search of the Nexis database to identify all articles relating to SSBs published between 1 April 2015 and 30 November 2016.A coding frame was developed. Two reviewers then coded a 10% random sample of articles to ensure consistency in the definition and application of codes. All remaining articles were coded by one reviewer. Data were analysed thematically, following the principle of constant comparison and attention to contradictory data. We used Beauchamp’s theory of market justice and social justice frames to analyse stakeholders’ messages on SSB taxation.Results A wide range of stakeholders sought to present evidence and arguments for or against SSB taxation. Stakeholder positions were largely shaped by their vested political interests. For example, corporate stakeholders were more likely to draw upon market justice frames promoting individual-level drivers for high rates of sugar consumption and individual-level solutions such as education. Whereas, public health advocates were more likely to draw upon social justice frames promoting population-level drivers for high sugar consumption and SSB taxation as a policy-level measure.Conclusion There is a complex, poorly-understood, inter-dependency between the framing of evidence in public policy debates, media representations of this evidence and the influencing strategies used by stakeholders. These early insights into stakeholders’ framing of evidence, both scientific and non-scientific, in the case of SSB taxation could potentially inform wider debates about the media strategies of global producers and marketers of unhealthy commodities to ‘directly lobby’ the public.Keywords: Public Health Policy, Media content analysis, SSB taxation

OP94

HOW DOES THE ALCOHOL INDUSTRY DEFINE “RESPONSIBLE DRINKING”? A QUALITATIVE ANALYSIS

NJ Maani Hessari1, M Petticrew1

1Department of Social and Environmental Health Research, London School of Hygiene and Tropical Medicine, London, United Kingdom

Background Alcohol is the third largest risk factor for disease burden worldwide, after hypertension and tobacco use. Although there is an extensive evidence base on the most effective interventions

to reduce alcohol harms at a population level (through targeting marketing, availability and price), the main focus of alcohol industry initiatives has been on providing information and education. “Responsible drinking” messaging (e.g. “Drink [product] Responsibly”) which frequently appears on product labels and adverts is a central element of such corporate social responsibility (CSR) activities. It has been argued that such messaging is vague, and potentially part of broader CSR activities to protect industry interests at the expense of public health. This study aimed to identify how industry defines responsible drinking, and in what contexts it is used. Methods Qualitative documentary analysis. A document search was carried out to identify publicly available documents (annual reports, shareholder communications, press releases and website content), published or available between January 2014 and July 2016, from two representative multinational alcohol producers (Diageo and AB InBev), Diageo’s DrinkIQ website, the Portman Group, the International Alliance for Responsible Drinking, the International Centre for Alcohol Policy or ICAP, and the DrinkAware Trust (all organisations funded by alcohol producers). These were compared to alcohol-related documents from Public Health England, WHO, Alcohol Concern and Addaction during this period. Coding was performed iteratively using NVivo 11 (version 11.2.2), and analysed using the hermeneutic approach, which involves reading and understanding meanings of individual texts, identifying sub-themes or ‘codes’, identifying thematic clusters of codes, triangulation between sources, checking reliability/validity, and illustrative use of representative case material.Results In total, 321 documents were evaluated, of which 101 referred to responsible drinking and were therefore included in the analysis.The term “responsible drinking” was used almost exclusively by industry or industry-funded organisations. Responsible drinking was not clearly defined with relation to any particular level of alcohol consumption, and government alcohol guidelines were rarely referenced. Long-term health harms (such as non-communicable diseases) were not mentioned in association with responsible drinking. Instead, responsible drinking was typically linked to behaviours (such as underage drinking). Conclusion Responsible drinking is a strategically ambiguous, industry-affiliated term allowing multiple interpretations. Industry sources rarely reference government drinking guidelines in the context of responsible drinking, instead stressing individual responsibility and risk management. Public health practitioners should be aware of these distinctions, and the industry framing of ‘responsible’ drinking, and use clear language regarding lower-risk drinking. Keywords: alcohol, industry, corporate social responsibility

OP95

QUANTIFYING THE POTENTIAL US HEALTH AND ECONOMIC EFFECTS OF THE FDA VOLUNTARY SALT REFORMULATION PROPOSAL 

J Pearson-Stuttard1,2, C Kypridemos1, B Collins1, Y Huang3, P Bandosz1, L Whitsel4, S Capewell1, D Mozaffarian3, P Wilde3, M Guzman-Castillo1, M O'Flaherty1, R Micha3

1Public Health and Pollicy, University of Liverpool, Liverpool, UK2School of Pubilc Health, Imperial College London, London, UK3Tufts Friedman School of Nutrition Science & Policy, Tufts University, Boston, USA4Policy Research, American Heart Association, Washington, USA

Background Salt consumption is a major modifiable risk factor for cardiovascular disease (CVD), the leading cause of mortality and morbidity in the US. Voluntary reformulation policies targeting salt have been deployed in several countries with varying effectiveness - high in Finland and the UK, low in Australia. The US Food & Drug Administration (FDA) has proposed voluntary salt reduction goals targeting processed and commercially prepared foods. We aim to quantify the potential CVD and economic impact of the FDA reformulation policy. 

Methods We extended the previously validated US IMPACT Food Policy Model. We then estimated the CVD cases averted, Quality Adjusted Life Years (QALYs) generated and cost-effectiveness from 2017-2036 of the proposed FDA reformulation policy. We used datasets including the National Health and Nutrition Examination Survey, cost information from the National Sodium Reduction Initiative and meta-analysis for salt consumption effects upon blood pressure and CVD. Costs included government costs to administer and monitor the policy and industry reformulation costs, under the assumption that estimated 75% of food products would be applicable for the salt reduction targets. Savings included healthcare and productivity costs. All costs were inflated to 2017 dollars and outputs were discounted at 3%.We modelled the 10-year reformulation targets under 2 scenarios:a)     Full industry compliance in all applicable food groupsb)     50% compliance in applicable food groupsWe then conducted a rigorous probabilistic sensitivity analysis.Results Achieving the salt reduction targets under a full compliance scenario could prevent approximately 516,000 CVD cases (95% uncertainty intervals 300,000 - 752,000) and gain some 2.7 (2.4 – 3.1) million discounted QALYs between 2017 and 2036.  The policy could produce discounted cost savings of approximately $62bn ($35.3bn - $86.2bn), with total net costs of approximately  +$15.7bn (policy), -$37.6bn (healthcare), and -$41.3bn (indirect costs) over the same period.Under the 50% compliance scenario, health gains would be approximately half as large, approximately 1.4 (1.3 – 1.7) million QALYs with discounted savings of $33bn ($19.4bn - 45.9bn).From a societal cost perspective, both scenarios would have an 80% chance of being cost effective after 4 years (Willingness to pay of $50,000/QALY) and cost saving after 10 years.Discussion Achieving the FDA salt reduction targets in processed foods could generate substantial health gains and cost savings in the US, assuming industry compliance. Policy makers should therefore focus on encouraging high compliance by industry to ensure that the powerful effects of salt reformulation are realised.Keywords: Salt, cost-effectiveness, cardiovascular disease

OP96

YOUNG PEOPLE’S EXPERIENCES OF NON-BROADCAST ADVERTISING OF UNHEALTHY FOOD

LE White, R Phipps, S Chambers, S Hilton1MRC/CSO Social and Public Health Sciences Unit, University of Glasgow, Glasgow, ScotlandBackground Critics argue that regulation of non-broadcast advertising for foods high in fat, sugar and salt (HFSS) offers less protection to children than regulation of broadcast advertising.  This is concerning as viewing habits now exist across a range of media platforms. There is a lack of research engaging with young people about the shifting nature of advertising for foods HFSS, particularly those aged 12-15 as they are often not included in industry self-regulatory initiatives. The study aims were to identify: 1) where young people experience advertising for foods HFSS; 2) their perceptions of this form of advertising; 3) the ways in which they believe they are influenced by this advertising.Methods We interviewed 65 UK 12-15 year olds in 15 focus groups.  Participants were recruited using snowball sampling techniques from initial local adult contacts. Potential participants were provided with a study summary sheet, and those who agreed to participate were asked to recruit a group of friends to take part in a discussion. Participants were drawn from a range of social backgrounds. Groups were held within participant’s homes or within the University.  All focus groups were audio-recorded. Topics included leisure time, viewing habits, and the perceived impact of advertising.  Young people were shown a range of broadcast and non-broadcast advertising to stimulate discussion. Interview transcripts were analysed thematically.

Results Young people reported that they rarely watched live television, and instead engaged in leisure activities that included watching programming via subscription services, and watching and socialising on digital platforms (such as video websites and social media). They recalled seeing extensive advertising for foods HFSS in non-broadcast media, both on- and offline. Participants reported scepticism and mistrust towards the healthfulness of many advertised foods. Nonetheless they believed they were influenced to purchase foods HFSS based on emotive techniques, such as togetherness, and were attracted to high quality advertising campaigns that made use of various techniques such as music, colour and humour.Conclusion Young people encounter advertising of foods HFSS across a wide range of non-broadcast media. It both attracts and frustrates them. Many young people believed advertising influenced their purchasing of food and drink. Regulation of non-broadcast advertising for foods HFSS must be updated to reflect these new and diverse viewing practices.Keywords: Food, Young People, Marketing

Cardiovascular disease

OP97

THE BURDEN OF HYPERTENSION AND ITS ASSOCIATED FACTORS IN THE GAMBIA: DATA FROM A NATIONAL HEALTH EXAMINATION SURVEY USING THE WORLD HEALTH ORGANISATION (WHO) STEPWISE APPROACH

B Cham, S Scholes, L Ng Fat, N Groce, JS Mindell1Epidemiology and Public Health, University College London(UCL), London, United KingdomBackground Non-communicable diseases are increasing in sub-Saharan Africa (SSA). They are estimated to account for 32% of deaths in The Gambia according to the WHO. Worldwide, prevalence of hypertension is highest in the African region (46%) and a very high proportion is undiagnosed.  There is very limited up-to-date information on the burden of diagnosed and undiagnosed hypertension and associated risk factors in The Gambia. This study aims to examine cardiovascular risk factors in in The Gambia adult population, with a particular focus on diagnosed and undiagnosed hypertension.Methods Data was collected from a random, nationally-representative sample of 4111 participants aged 25-64 years (78% response rate) in 2010 using the WHO STEPwise survey methods. Analysis was restricted to non-pregnant participants with three valid blood pressure (BP) measurements (n=3573). All analysis were weighted and adjusted for complex survey design using STATA14. The mean of the second and third BP measurements was used in the analysis. Hypertension was categorised into measured (SBP≥140mmHg and/or DBP ≥90 mmHg) and total (SBP≥140mmHg and/or DBP ≥90 mmHg and/or self-reported hypertension). Among people with total hypertension, we also looked at undiagnosed hypertension (proportion of participants with hypertension not aware of their status). Univariate and multivariate regression models were run to identify the most important factors associated with hypertension including sex, age, rural/urban residence, socioeconomic and anthropometric factors. Smoking status and fruit and vegetable intake were additional covariates.Results One third of adults were hypertensive; this was higher in rural regions (40%, p<0.001). Multivariate analysis revealed increased odds of total hypertension among the overweight/obese and rural residents. Abdominal obesity (OR1.8 [95% CI, 1.2-2.7]), rural residence (3.0[1.6-5.5]), and age were the most important predictors among men while in women it was generalised obesity (2.4[1.6-3.7]), rural residence (2.5[1.4-4.5]), and age. More than three-quarters of hypertensive participants were undiagnosed: this was higher among males (86% vs 71%, p<0.001).  Men (3.1[1.7-5.6]) and participants aged 25-34 years (4.8[1.4-3.5]) had higher odds of undiagnosed

hypertension after adjusting for other factors. However, obesity was protective for undiagnosed hypertension (0.4[0.2-0.6]).Conclusion Contrary to what is found in similar studies in SSA, where hypertension is highest in urban areas, we found that rural residence, abdominal obesity among men and generalised obesity among women were the most important predictors of hypertension. Intervention to reduce hypertension could be further targeted towards rural areas. Sensitisation campaigns should promote awareness of the risk factors, especially on the importance of maintaining a healthy weight.Keywords: Hypertension , STEPwise survey, Africa

OP98

REVISITING EARLY LIFE DEPRIVATION AND CARDIOVASCULAR DISEASE: AN ECOLOGICAL STUDY OF HISTORICAL TRENDS IN ECONOMIC DEVELOPMENT AND CURRENT CARDIOVASCULAR MORTALITY IN 1577 BRAZILIAN MUNICIPALITIES

PAC Mallinson1, S Kinra1

1Department of Non-communicable Disease Epidemiology, London School of Hygiene and Tropical Medicine, London, UK

Background Ecological findings from Europe in the 1970s demonstrated an interaction between deprivation during early life and adult cardiovascular disease (CVD) mortality. These findings hold renewed significance today in the context of emerging epidemics of CVD in rapidly developing countries. If generalizable to such settings, understanding of the interaction between early life deprivation and CVD mortality might improve disease projections and targeting of resources to high risk areas. To investigate this, we studied economic development rates since 1940 and current CVD mortality in Brazil.Hypothesis Currently, higher GDP/capita is associated with higher CVD mortality at municipality level in Brazil. We hypothesised that if deprivation was a risk factor for CVD mortality during the early life period, municipalities which have undergone the greatest shifts from low to high GDP/capita in the past 50-70 years would have higher CVD mortality rates today than those with consistently high or low GDP/capita.Methods We used municipality-level data on deaths, demographics and gross domestic product (GDP) from the Brazilian Institute of Geography and Statistics. Our primary outcome was CVD mortality rate in 2005-14, defined and adjusted according to the WHO Global Burden of Disease protocol. We compared the trajectory of municipality GDP/capita between time of birth and time of death, defined by tertiles of GDP/capita at each time point. Analyses were age-standardised and stratified by sex. Municipalities were grouped to reflect 1940s borders and excluded where this was not possible. We conducted analyses on R.Results In 1557 included municipalities, 367 had a low-low GDP/capita trajectory, 44 had low-high and 329 had high-high. Age-adjusted CVD mortality rates for >50 year-olds, per 100,000 person years, in low-low, low-high and high-high trajectory municipalities, respectively, were 656.7 (95%CI: 636.2, 677.2), 758.2 (95%CI: 713.8, 802.5) and 821.9 (95%CI: 810.7, 833.2) among men (p-value trend test <0.001); and 447.0 (95%CI: 430.5, 463.4), 414.4 (95%CI: 383.5, 445.3) and 449.5 (95%CI: 442.0, 457.0) among women (p-value trend test >0.1). These findings were not substantially altered in sensitivity analyses checking for the potential effects of internal migration.Conclusion Contrary to what we hypothesised, Brazilian municipalities which have shifted from low to high GDP/capita did not exhibit higher rates of CVD mortality than consistently high or low municipalities. This reminds us to be cautious extrapolating evidence generated in high-income settings to rapidly developing settings where social and economic contexts surrounding CVD differ markedly. Further individual-level studies with robust designs are needed, as inference from ecological studies has limitations.Keywords: Cardiovascular disease, Early life deprivation, Rapidly developing countries

OP99

Psychological distress and incident stroke risk in the 45 and Up Study

CA Jackson1,2, CLM Sudlow3, GD Mishra2

1Usher Institute of Population Health Sciences & Informatics, University of Edinburgh, Edinburgh, Scotland2School of Public Health, University of Queensland, Brisbane, Australia3Centre for Clinical Brain Sciences, University of Edinburgh, Edinburgh, Scotland

Background Few studies have reported on the association between psychological distress and stroke risk, with most investigating the effect of psychological distress on stroke mortality rather than incidence. We investigated whether psychological distress is associated with stroke incidence in a large population-based cohort study and examined whether associations differed by gender, age-group and pathological stroke type.Methods We included 228,955 participants without prior stroke from the New South Wales 45 and Up Study. Baseline psychological distress was measured using the 10-item Kessler psychological distress Scale and categorised as low, medium, and high/very high. We identified incident stroke through linkage to hospital admission and mortality records. We analysed men and women separately, using cox survival analysis to obtain unadjusted and adjusted hazard ratios (HRs) with 95% confidence intervals (CIs) for the association between psychological distress and all stroke and pathological stroke types. We serially adjusted for groups of confounders, including: sociodemographic factors; lifestyle factors; clinical disease history; family history of cardio- and cerebrovascular disease; physical disease co-morbidity (Charlson comorbidity index); and (among women) menopausal status and current hormone replacement therapy and oral contraceptive use.Results During 1,075,057 person-years of follow-up (mean follow-up time 4.7 ± 0.98 years), we identified 2682 incident strokes among men and women. In men, medium and high/very high psychological distress was associated with a 10% and 34% increased risk of stroke compared to low psychological distress (fully adjusted HRs 1.10, 95% CI 0.95, 1.26 and 1.34, 95% CI 1.11, 1.62, respectively). Similar effect estimates were observed in women (fully adjusted HRs for medium and high/very high versus low psychological distress: 1.17, 95% CI 1.00, 1.37 and 1.44, 95% CI 1.18, 1.77). Effect estimates were similar across age-groups and pathological stroke type, in both men and women.Conclusion Psychological distress is strongly associated with increased stroke risk, even after adjusting for a wide range of confounding factors. Further investigation is needed to establish whether this relationship is causal and to determine the underlying mechanism(s). Meanwhile, study findings support the need for renewed efforts: to encourage people with mental ill-health to seek medical help; for better screening and treatment for mental health conditions (which might itself reduce cerebrovascular and cardiovascular disease risk); and to encourage screening for established cerebrovascular risk factors in this high-risk, vulnerable group.Keywords: epidemiology; Psychological distress; stroke

OP100

THE COST-EFFECTIVENESS AND EQUITY OF THE NHS HEALTH CHECKS CARDIOVASCULAR DISEASE PREVENTION PROGRAMME: A MICROSIMULATION USING REAL-WORLD DATA FROM A DEPRIVED NORTHERN CITY

B Collins1, C Kypridemos1, P Parvulescu2, R Gosling2, S Capewell1, M O'Flaherty1

1Department of Public Health and Policy, University of Liverpool, Liverpool, UK2Public Health Department, Liverpool City Council, Liverpool, UK

Background The NHS Health Check Programmes’ stated objective is the early identification of otherwise healthy people at high risk of cardiovascular disease (CVD) and diabetes. However, the programme’s effectiveness, cost-effectiveness, and equity are still contested. This study therefore aimed to determine whether Health Checks (HCs) are cost-effective and equitable in a city with high levels of deprivation and CVD.Methods IMPACTNCD is a previously validated,R-based dynamic stochastic microsimulation policy model. We calibrated this model to Liverpool demographics, risk factor exposures, and CVD epidemiology. We modelled the current implementation of HCs using local and national data on effectiveness, costs, and participant risk profiles. Disease costs and health state utilities were drawn from standard sources and discounted at 3.5% annually using a healthcare perspective. We quantified the uncertainty of model outputs using second-order Monte Carlo simulation, and report 95% uncertainty intervals.We modelled three fifteen year scenarios from 2017 to 2031:A) continuing the current implementation of HCs;B) an optimal implementation of HCs assuming optimal coverage, uptake, treatment and lifestyle change;C) combining scenario A with structural policies targeting dietary consumption of salt, sugar, fruit and vegetables.We compared all three scenarios with a counterfactual of no HCs, and conducted a rigorous sensitivity analysis. Results The model suggested that over 15 years the CVD cases prevented or postponed would be approximately 310 (40 – 734) for scenario A, 870 (327 – 1,397) for scenario B, and 1,740 (815 – 2,939) for scenario C.Cumulative discounted net costs and quality-adjusted life years (QALYs) gained for the three scenarios respectively would be +£2.1m (-£1.5m – +£4.8m) and +90 QALYs (-124 – +376) for A; +£1.4m (-£6.1m – +£6.6m) and +434 QALYs (-76 – +1,133) for B; or -£16.9m (-£33.2m – -£5.9m) and +2,871 QALYs (+1,355 – +4,830) for C.We estimated the probability of scenarios A and B being cost-effective by 2031 at 25% and 74% respectively, valuing each QALY at £20,000.Scenario C would become cost saving by 2030.Scenario A may increase existing health inequalities; B is likely to be neutral, while C would substantially decrease inequalities.In extensive sensitivity analyses, the direction of the results did not change when the discount rate was varied from 0%-6%, nor when the time horizon was increased to 20 years.Conclusion Current NHS Health Checks implementation appears neither equitable nor cost-effective.  The addition of structural policies proved equitable and cost saving. Future research might now seek to identify the optimal combination of structural policies at local level. Keywords: CVD Prevention, Economic evaluation, Modelling

SSM Annual Scientific Meeting 2017 Plenary (PL) Presentations

PL01

THE IMPACT OF CO-LOCATED WELFARE ADVICE IN PRIMARY HEALTHCARE SETTINGS ON MENTAL HEALTH AND HEALTH SERVICE USE: A MIXED METHODS EVALUATION

C Woodhead1, M Khondoker2, R Raine1

1Applied Health Research, University College London, London, England2Medicine and Health Sciences, University of East Anglia, Norwich, England

Background Co-locating welfare advice services in primary healthcare settings has been one approach to tackling health inequalities by increasing income among socially deprived individuals. It is also hoped to relieve pressure on general practices in supporting patients with ‘non-clinical’ needs. Previous evaluations have been methodologically limited and lack theoretical underpinning. We aimed to examine the impact of co-located welfare benefits and debt advice on mental health and primary care service use, and to develop theory linked to pathways of effect.Methods A prospective, controlled quasi-experimental study with an embedded qualitative component was carried out (December 2015-December 2016) in eight intervention and nine comparator sites across North Thames, London. Before-and-after quantitative data were collected via self-report questionnaires. Comparison group members were propensity score weighted for analyses. Outcomes included change in symptoms of common mental disorder (CMD) (12-item General Health Questionnaire), well-being (Shortened Warwick and Edinburgh Mental Well-being Scale), three-month GP consultation rate and financial strain. Data from qualitative interviews with 24 primary care staff, funders and advice providers were analysed using a modified realist evaluation approach to understand how co-located welfare advice could influence practice outcomes.Results For the quantitative study, n=285 and n=633 individuals were recruited into advice and comparison groups respectively at baseline. 72% and 84% were retained at 3 month follow-up. Relative to controls, CMD caseness reduced significantly among female and Black/Black British advice recipients. Individuals whose advice resulted in positive outcomes demonstrated significantly improved well-being scores. Significant reduction in financial strain overall but no change in three-month consultation rate was found. Per capita, advice recipients received £15 per £1 of funder investment. Qualitative findings were used to inform underlying theory linking service activity to general practice outcomes. These were reduced GP consultations for ‘non-clinical’ issues and reduced practice staff time supporting patients with such issues. The findings revealed key implementation, context and agency factors that facilitated or hindered the potential for co-located advice to influence these outcomes.Conclusion Recipients of co-located welfare benefits and debt advice experience reduced financial strain and for sub-groups short term mental health is improved. Co-located advice services have the potential to support general practice work but not if co-location is limited to a physical sharing of space. Suggestions are made to facilitate joint working.Keywords: health inequalities, social welfare, co-location, welfare advice, primary care

PL02

TOBACCO CONTROL IN ENGLAND: USING MICROSIMULATION MODELLING TO QUANTIFY THE POTENTIAL IMPACT OF A TOBACCO-FREE GENERATION OR A TOTAL BAN

C Kypridemos1, J Reeve2, I Buchan3, A Gilmore4,5, S Capewell1, M O'Flaherty1

1Department of Public Health and Policy, University of Liverpool, Liverpool, United Kingdom2WMS - Social Science and Systems in Health, University of Warwick, Coventry, United Kingdom3Farr Institute @ HeRC, University of Manchester, Manchester, United Kingdom4Department for Health, University of Bath, and the UK Centre for Tobacco and Alcohol Studies, Bath, United Kingdom5UK Centre for Tobacco and Alcohol Studies (UKCTAS), Bath, United Kingdom

Background In 2015, almost one-fifth of English adults continued to smoke. Tobacco control policies in the UK are amongst the strongest in Europe, yet smoking prevalence remains stubbornly high, especially in deprived groups. Novel and radical approaches may be needed to control tobacco effectively. The British Medical Association backs a tobacco ban for those born after 2000, and a population-wide tobacco ban has already been implemented in Bhutan. We use microsimulation modelling to quantify the effectiveness and equity of these two radical tobacco policies.Methods IMPACTNCD is a previously validated dynamic stochastic microsimulation developed in R. It simulates the life course and smoking histories of synthetic individuals under alternative scenarios. We used IMPACTNCD to estimate the potential impact of two proposed changes to tobacco control policy in England – a sales ban restricted to those born in or after 2000 and a total sales ban – on a simulated English population over a 30-year time frame. Extrapolating from Bhutan’s implementation, we assumed 50% reductions in smoking initiation rate, active to ex-smoking ratio, and cigarette consumption. We compared both scenarios with a counterfactual that assumed current declining trends in smoking will continue in the future. We conducted a rigorous sensitivity analysis using second-order Monte Carlo simulation, and we report the median and interquartile range of the output distributions.Results The model suggested that under the tobacco-free generation caps scenario, smoking prevalence would fall to 12.4% (12.1% to 12.6%) for men and 7.9% (7.7% to 8.1%) for women, by 2045. This could prevent or postpone approximately 3,500 (-4,200 to 11,000) cardiovascular disease cases and some 230 (-3,100 to 3,600) lung cancer cases; resulting in approximately 190 (-2900 to 3400) fewer cardiovascular disease deaths, and 220 (-2,700 to 3,200) fewer lung cancer deaths.Under the total caps scenario, the English smoking prevalence would fall to 7.2% (7.0% to 7.4%) for men and 4.3 % (4.1 % to 4.5 %) for women, by 2045. This could prevent or postpone approximately 90,000 (70,000 to 120,000) cases of cardiovascular disease, some 79,000 (55,000 to 120,000) cases of lung cancer, approximately 14,000 (3,000 to 25,000) fewer cardiovascular disease deaths and some 54,000 (38,000 to 73,000) fewer lung cancer deaths.Both scenarios could reduce socioeconomic health inequalities in cardiovascular disease and lung cancer morbidity and mortality.Conclusion Strengthening existing English tobacco control policies through limiting access could substantially improve effectiveness and equity. Further research is now needed to explore the political and legal feasibility issues.Keywords: tobacco control, modelling, policy

PL03

HOSPITAL TREATED DELIBERATE SELF-HARM AND RISK OF SUICIDE AND DEATH FROM OTHER EXTERNAL CAUSES IN THE REPUBLIC OF IRELAND – A NATIONAL REGISTRY COHORT STUDY

IB O'Farrell1, IJ Perry1, P Corcoran1

1Epidemiology and Public Health, University College Cork, Cork, Ireland

Background Suicide is a major public health problem. The prediction of suicide is difficult, however research has identified that deliberate self-harm (DSH) is one of the strongest predictors of future suicide. To date, the risk of suicide in individuals who DSH is not well established internationally as relatively few countries have accurate data on DSH. This is the first registry based study to examine the risk of mortality on a national cohort of all individuals presenting to hospital due to DSH in Ireland. Methods A national prospective cohort of 26,168 DSH patients attending the 40 hospital emergency departments in Ireland from 2009 to 2011, were followed up until to the end of 2011 using national death recording systems. Gender specific age adjusted European standardised rates for external cause mortality were calculated. Additionally, Poisson regression was used to generate incidence rate ratios (IRRs). Potential risk factors were investigated using Cox Models.Results During the study follow-up 437 patients died from external causes. The average 1-year cumulative incidence for suicide, non-suicide external cause mortality and all external causes combined were 0.8%, 0.5% and 1.3% respectively. The risk of suicide within the first year after DSH was 46 times greater in DSH population compared to the general population. Risk of other non-suicide external cause mortality was also greater in the DSH population compared to the general population (females; IRR=30, males; IRR=20). While the relative risk of death was higher in the female DSH population compared to the female general population, the absolute risk of death was found to be higher in males than females. Older age and male gender were associated with an elevated risk of death. Risk of death from suicide varied depending on method of DSH. Compared with overdose alone attempted hanging had the greatest risk of suicide, particularly in females (females; HR=6.8, males; HR=2.6), major self-cutting was also associated with a 2-fold increased risk. DSH repetition was also found to be a strong predictor of subsequent death.Conclusion The findings from the world’s first national DSH Registry highlight the extremely high risk of death from suicide and other external causes following hospital treated DSH. Older age, male gender, DSH repetition (especially for females) and persons presenting with attempted hanging or major self-cutting are at a particular risk. The findings from this study highlight the need for well-structured, specialist and organised care for this vulnerable group attending emergency departments. Keywords: mortality follow up study, deliberate self-harm, suicide

SSM Annual Scientific Meeting 2017 Poster (P) Presentations

P01

SYSTEMATIC REVIEW OF REVIEWS ON THE NEGATIVE HEALTH EFFECTS AND PSYCHOSEXUAL BENEFITS OF HIGH-HEELED SHOES

MS Barnish1, HM Morgan2, J Barnish3

1Child Health, University of Aberdeen, Aberdeen, Scotland2Health Services Research Unit, University of Aberdeen, Aberdeen, Scotland3Retired health professional, London, England

Background Although frequently worn by many women in line with traditional female gender identity, high-heeled shoes (high heels) have been shown to be detrimental to musculoskeletal health and increase the risk of injury. However, no evidence synthesis has considered both the psychosexual benefits and the epidemiological and biomechanical impacts of their wear. We present a systematic review of reviews concerned with the public health challenge of high heels considering all three of these aspects.Methods Seven standard academic health bibliographic databases, including MEDLINE and EMBASE, were searched up to November 2016 using high heel-related keywords such as “high heel” and “stiletto”. Supplementary searches were also conducted in Google Scholar, Directory of Open Access Journals and bibliographies of relevant articles. Due to several review articles on aspects of our research question, we initially included all review articles that provided evidence linking high heels to psychosexual benefits or musculoskeletal health problems (osteoarthritis, hallux valgus, pain or injury) from an epidemiological or biomechanical perspective in participants without prior history of significant musculoskeletal conditions. We then considered additional primary studies addressing areas on which there was no review or an identified lack of evidence. Narrative synthesis was conducted using standard forms. Proportionate second review was conducted.Results A total of 506 unique records were identified, 27 full-text publications were screened and 20 publications (7 reviews and 13 additional primary studies) were included in our narrative synthesis. All available psychosexual studies show that high heels increase women’s attractiveness to men, although they are uncritical regarding heteronormativity. The most up-to-date epidemiological review provides clear evidence of a link to hallux valgus, pain and first-party injury. The body of biomechanical reviews also provides clear evidence of changes indicative of increased risk of these outcomes plus osteoarthritis. With regard to second-party injury, evidence is limited to one descriptive study and eight case reports.Conclusion Our evidence synthesis clearly shows that high heels bring psychosexual benefits to women, but are detrimental to their musculoskeletal health. Considering this dilemma, it is important that women’s freedom of choice is respected in social life, and that policy makers seek to address any remaining issues of explicit or implicit compulsion to wear high heels (e.g. at work). Further research is needed to assess second-party injury and any public safety implications. A limitation of our study is that there was no suitable quality assessment tool for the breadth of designs we included.Keywords: public health, health policy, musculoskeletal

P02

POTENTIAL BENEFIT OF SINGING FOR PEOPLE WITH PARKINSON’S DISEASE: A SYSTEMATIC REVIEW UPDATED TO 2017

MS Barnish1, RA Atkinson2, SM Barran3, J Barnish4

1Respiratory Research Group, University of Aberdeen, Aberdeen, UK2Adult Speech and Language Therapy, St Helier Hospital, London, UK3Children's and Young People's Speech and Language Therapy, Guy's and St Thomas', London, UK4Retired health professional, London, UK

Background There is evidence that participation in performing arts brings psychosocial benefits in the general population. In recent years, there has been substantial interest in the potential therapeutic benefits of performing arts for people with chronic neurological conditions. A lack of effective evidence synthesis, however, made it difficult to evaluate the evidence base and future research directions. We conducted the first systematic review of the potential benefit of singing for people with Parkinson’s disease (PD) on speech, communication, cognition, motor function and quality of life outcomes. Here, we present an updated version of this systematic review up to January 2017.Methods Seven standard academic health bibliographic databases, including MEDLINE and EMBASE, were searched up to January 2017 using MeSH terms and keywords corresponding to (Parkinson’s disease AND (Singing OR Music OR Music therapy). Supplementary searches were also conducted in Google Scholar and bibliographies of relevant articles. We considered full-text original articles assessing the potential benefit of singing for human participants with clinically diagnosed PD on speech impairment, functional communication, cognitive status, motor function or quality of life using any appropriate quantitative design. Narrative synthesis was conducted using standard forms. Proportionate second review was conducted. Study quality was assessed using the Threats to Validity tool.Results A total of 490 unique records were identified, 30 full-text publications were screened and eight studies included in the review. All eight studies assessed the impact of singing on speech, of which six found evidence of benefit. Two studies assessed quality of life, one finding evidence of benefit. One study assessed functional communication and found no evidence of benefit. No studies assessed cognitive or motor function. One study was assessed at low risk of bias, one at medium risk of bias and six at high risk of bias.Conclusion Included studies provide evidence that singing benefits the speech of people with PD. However, evidence is limited with regard to wider benefits, especially those of a psychosocial nature such as functional communication and quality of life, which are areas of high important to people with PD. A key limitation of our review is that most studies were at high risk of bias. Groups such as Parkinson’s UK run choirs for people with PD – this seems a good idea and has some evidence base, but further more rigorous studies are required to provide a stronger evidence base to support greater healthcare provider-community organisation partnerships.Keywords: health services research, community health, performing arts

P03

WHAT DO PEOPLE IN THE GENERAL POPULATION THINK ABOUT BACK PAIN – AND DOES IT MATTER? A SYSTEMATIC REVIEW

L Morton1234, M de Bruin3,4, GJ Macfarlane1,2,4

1Epidemiology Group, University of Aberdeen, Aberdeen, UK2Aberdeen Centre for Arthritis and Musculoskeletal Health, University of Aberdeen, Aberdeen, UK3Health Psychology Group, University of Aberdeen, Aberdeen, UK4Arthritis Research UK / MRC Centre for Musculoskeletal Health and Work, University of Southampton, Southampton, UK

Background Mass media campaigns have aimed to change the public’s beliefs about back pain to ultimately change behaviours like healthcare utilisation. Most campaigns demonstrated shifts in beliefs without concomitant changes in behaviours. This raises the question of whether the beliefs targeted by campaigns were prevalent and related to the behaviours they tried to change. The objectives of this review were to: i) describe the prevalence of particular beliefs about back pain, ii) identify any cross-sectional relationships between these beliefs and other (e.g. sociodemographic) factors, and iii) assess the evidence for beliefs about back pain to predict future back pain-related outcomes within the general population.Methods A systematic review of the literature was carried out using searches for ‘back pain’, ‘beliefs’ and ‘general population’ in five literature databases (MEDLINE, PsycINFO, Embase, ISI Web of Science, CINAHL). Articles were screened, data was extracted and quality was assessed by two reviewers. Results across the three aims were tabulated and narratively synthesised.Results 4480 unique records were identified and 30 articles were included. Beliefs about back pain were measured using eight questionnaire instruments and 57 unique items. These represented beliefs about: consequences, fearing and avoiding pain/activities, catastrophising, and other specific beliefs. Where a given construct was assessed by more than one study, most samples agreed, on average, with beliefs about negative consequences of back pain (8/12), while most disagreed with fear avoidance beliefs about activity (4/5). Negative beliefs about consequences were consistently associated with being older, unemployed, having completed less education, lower self-rated health, and more pain and disability; they were predictive of future back pain-related outcomes.Discussion Most samples agreed with beliefs about back pain’s negative consequences and these beliefs predicted some back pain-related outcomes. However, based on current evidence, we cannot say whether these beliefs impact future management behaviours. The relationships between other belief constructs and future management behaviours has not been investigated in the general population. Previous campaigns’ mixed success could, in part, be due to not targeting beliefs which were related to behaviours they hoped to change. Existing evidence suggests that this relationship has not been assessed in the general population; it must be elucidated prior to developing future campaigns.Keywords: illness beliefs, back pain, systematic review

P04

CHILD-PARENT SEPARATION DURING UPBRINGING AND LATER RISK OF VIOLENT CRIMINALITY AND SELF-HARM: A POPULATION-BASED COHORT STUDY

PLH Mok1, A Astrup2,3, CB Pedersen2,3, MJ Carr1, RT Webb1

1Centre for Mental Health and Safety, University of Manchester, UK2Centre for Integrated Register-based Research, CIRRAU, Aarhus University, Aarhus, Denmark3National Centre for Register-Based Research, Business and Social Sciences, Aarhus University, Aarhus, Denmark

Background Childhood experience of parental separation has been associated with subsequent increased risk for interpersonal and self-directed violence in offspring. However, conflicting and insufficient evidence exists regarding how links vary across age, duration, and types of separation. We conducted a large, detailed epidemiological study of the association between child-parent separation and later risk of violent criminality and self-harm in offspring. We investigated how risks varied by type of separation (from mother or father only, or both), offspring gender, age at separation, number and duration of separations, and by a range of separation scenarios.Methods Our study cohort was established using data from the Danish Civil Registration System and interlinked psychiatric and crime national registers. Persons born in Denmark 1971-1997 (N=1,346,772) were included. Residential addresses were used to determine if cohort members were living with either parent each birthday from birth to 15th. Members were followed up from reaching age 15 birthday until date of first violent offence conviction, date of first registered self-harm episode, or end of study (31st December 2012). Incidence rate ratios (IRRs) were estimated using survival analyses techniques in SAS 9.4.Results All types of child-parent separation scenarios are linked with increased risks for later violent offending and self-harm. Even those who were separated from one parent, but subsequently lived with both, were at increased risk. Compared to persons who always lived with both parents, those who lived with both parents at birth but were subsequently separated from the father - the most common type of separation scenario - had a 2-fold increased risk (IRR for violent offending=1.99, 95% CI 1.94-2.05; IRR for self-harm=2.07, 2.01-2.14). Risks increased with rising number of parental changes, and a steep gradient in risk with increasing age at separation was observed if children were first separated from both parents between age 10 and 15 years. Strengths of associations were much greater for female than male offspring for violent offending, but no gender differentials in risks were found for self-harm.Conclusion Our study adds to existing published evidence that experience of child-parent separation may increase the risk for harmful psychosocial development in offspring. However, we had no information on the reasons of separation and aggregated risk estimates would have masked any heterogeneity in the associations observed. Interventions should be tailored toward those young people with the most complex trajectories who have lived through many family changes as they grew up.Keywords: Child-parent separation; Self-harm; Violent crime

P05

MULTIPLE RISK BEHAVIOUR IN ADOLESCENCE AND ADVERSE HEALTH AND SOCIAL OUTCOMES IN EARLY ADULTHOOD: FINDINGS FROM A PROSPECTIVE BIRTH COHORT STUDY

R Campbell1, M Hickman1, R Kipping1, M Smith2, T Pouliou3, J Heron1

1School for Social and Community Medicine, University of Bristol , Bristol, UK2School for Clinical Medicine, University of Bristol , Bristol, UK3Administrative Data Research Wales, Swansea University, Swansea, UK

Background Adolescents engage in new and changed behaviours such as substance use or reduced physical activity. Evidence is accruing that these health risk behaviours co-occur at a population level and cluster within individuals. There has, however, been less focus on what the cumulative impact of engagement in multiple risk behaviours might be even in the short term. We sought to assess if engagement in multiple risk behaviours in mid-adolescence is associated with deleterious health and social outcomes in early adulthood.Methods 5591 young people enrolled in the Avon Longitudinal Study of Parents and Children provided data on 13 risk behaviours from the domains of physical inactivity, substance use, self-harm, sexual health, vehicle-related risk ehavior, and antisocial ehavior when aged 15 to 16. Logistic regression was used to examine any association between the total number of behaviours engaged in and rates of obesity, harmful alcohol use, problem gambling, anxiety, depression, trouble with the police and unemployment and lack of training at age 18.Results There was a strong association between multiple risk behaviour and all seven adverse outcomes. For each additional risk behaviour engaged in, the odds of having anxiety or being a problem gambler were 1.18 [1.12, 1.24] and 1.20 [1.13, 1.27] respectively, the odds for depression were 1.24 [1.17, 1.31], and the odds of getting into trouble with the police or of harmful drinking were 1.49 [1.42, 1.57] and 1.58 [1.48, 1.69] respectively. Adjustment for gender, parental socio-economic position, and maternal risk behaviours did little to alter the odds of these adverse outcomes. Confining analyses to adolescents not having experienced the adverse outcomes before age 15, to exclude reverse causality, also produced little change in odds.Conclusion Improvements in adolscent health are lagging behind those seen in child health. Investment in interventions and environments that effectively prevent multiple risk behaviour is likely to produce better adolescent health and wellbeing. Keywords: Adolescence, health risk behaviour, cohort study

P06

CORRELATES OF LEVEL AND LOSS OF GRIP STRENGTH IN LATER LIFE: FINDINGS FROM THE ENGLISH LONGITUDINAL STUDY OF AGEING AND THE HERTFORDSHIRE COHORT STUDY

LD Westbury1, HE Syddall1, EM Dennison1,2, C Cooper1,3,4, CR Gale1,5

1MRC Lifecourse Epidemiology Unit, University of Southampton, Southampton, United Kingdom2Victoria University of Wellington, Wellington, New Zealand3NIHR Southampton Biomedical Research Centre, University of Southampton and University Hospital Southampton NHS Foundation Trust, Southampton, United Kingdom4NIHR Musculoskeletal Biomedical Research Unit, University of Oxford, Oxford, United Kingdom5Centre for Cognitive Ageing and Cognitive Epidemiology, Department of Psychology, University of Edinburgh, Edinburgh, United Kingdom

Background Low grip strength in later life is a risk factor for subsequent disability and mortality and a key component of sarcopenia. Established determinants of low grip strength level include: older age; shorter stature; poor nutrition; low physical activity; socio-economic disadvantage and multimorbidity. However, little is known about risk factors for accelerated loss of grip strength in later life. Methods We examined socio-demographic, lifestyle and clinical predictors of baseline level and 8-year loss of grip strength in 3,703 men and women (aged 52-82 years at baseline) in the English Longitudinal Study of Ageing (ELSA). Data on 441 men and women (aged 59-71 years at baseline) who participated in a 10-year follow-up of the Hertfordshire Cohort Study (HCS) were used for replication. Change in grip strength from baseline to the end of follow-up was characterised using linear mixed-effects models in ELSA as grip strength was measured at three time-points. A residual change method was used in HCS to obtain a measure of change in grip strength between two time-points which was independent of baseline level. Linear regression models adjusted for age and gender, as well as mutually-adjusted models, were used to examine the relationship between baseline predictors and level and change in grip strength in both cohorts.Results Men in ELSA and HCS had higher average levels of grip strength at baseline, and accelerated rates of loss, compared with women. In ELSA, older age, shorter stature, and increased morbidity were associated with lower level, and accelerated rate of loss, of grip strength in both sexes. For example, accelerated loss of 0.04 (95%CI: 0.00, 0.08) standard deviation (SD) scores per additional morbidity were estimated from mutually-adjusted models; corresponding SD scores per SD decrease in height were 0.06 (95%CI: 0.03, 0.09). In mutually-adjusted analyses, socioeconomic disadvantage, low level of physical activity and poorer self-reported health were also associated with low grip strength level, but not with rate of loss. Analysis in HCS yielded similar results. Conclusion Our results identify multimorbidity as a modifiable determinant of both level and loss of muscle strength in later life. The association between shorter stature and accelerated loss of grip strength suggests that developmental influences may also impact on rate of loss, as well as on level, of muscle strength in older age. This research will inform the development of lifecourse interventions to promote maintenance, and reduce loss, of muscle strength in later life.Keywords: grip strength, risk factors, later life

P07

SENSORY IMPAIRMENTS AND COGNITIVE AGEING: FINDINGS FROM 11 EUROPEAN COUNTRIES

A Maharani1, N Pendleton1, G Tampubolon1, J Nazroo1, P Dawes1

1SENSE-Cog project, University of Manchester, Manchester, United Kingdom

Background Sensory impairment (hearing and/or visual) and cognitive decline commonly occur in the elderly. Whether they are in fact associated, and whether sensory impairment might contribute to intellectual decline, has been a subject of a number of investigations during the past three decades. The purpose of this study was to assess in an older European population: (1) any independent association between single and dual sensory impairment (hearing and/or vision) and cognitive decline; (2) cognitive trajectories according to the pattern of sensory impairment. Methods This study used data from the Survey of Health, Ageing and Retirement in Europe (SHARE) and English Longitudinal Study of Ageing (ELSA). We measured cognitive function by one key cognitive domain available in both surveys, namely episodic memory score (range 0-20). Sensory impairment was measured using self-reported hearing and vision quality in both SHARE and ELSA. Vision and hearing function in both surveys were each coded on a scale from 1 (excellent) to 5 (poor). We recoded the scores of hearing and vision function into two categories by combining the responses excellent, very good and good into `good' sensory function and collapsing fair and poor vision into `poor' sensory function. We defined sensory impairment as having poor sensory function and categorised it into three: no impairment, single impairment (vision or hearing), and dual impairment (vision and hearing). Results Using growth curve analysis, we found that older adults with single and dual sensory impairment (hearing and/or vision) were able to recall fewer words than those with no sensory impairment in final model in all 11 European countries included in this study, where the potential confounders were included. The cognitive trajectories of older adults with no sensory impairment followed curvilinear shapes, while those of older adults with single and dual sensory impairments showed more precipitous pattern trajectory of cognitive decline after the age of 50 in both surveys. Being female, having attained a higher level of education, having paid job and being relatively wealthy, were associated with higher cognitive function in older age both in SHARE and ELSA. Conclusion These findings demonstrate that hearing and/or vision impairment is a marker for risk of cognitive decline that could inform preventative interventions to maximise cognitive health and longevity. Further studies are needed to investigate how sensory markers could inform strategies to improve cognitive ageing, including hearing and vision rehabilitative intervention in combination with healthy aging interventions to promote social engagement, physical activity and positive health behaviors.Keywords: sensory impairment, cognition, ageing, Europe

P08

INCIDENCE OF DIAGNOSED DEMENTIA IN THE ENGLISH LONGITUDINAL STUDY OF AGEING IN ENGLAND: A 12-YEAR FOLLOW-UP

D. Cadar1, H. Davies1, D. Llewellyn2, G.D. Batty3, A. Steptoe1

1Department of Behavioural Science and Health, University College London, London, UK2Mental Health Research Group, University of Exeter Medical School, University of Exeter, Exeter, UK3Department of Epidemiology and Public Health, University College London, London, UK

Background There has been a suggestion that the incidence of dementia in England has recently declined. Crucially, however, these estimates have not been based on nationally representative samples. Accordingly, we calculated dementia incidence, the temporal trends and, additionally, examined various socioeconomic and geographical predictors of its occurrence.Methods Data from the English Longitudinal Study of Ageing (ELSA), a prospective cohort study at baseline known to be representative of the English population, were used to investigate the rates of dementia incidence over seven waves between 2002 and 2014.Dementia was determined by doctor-diagnosis and the Informant Questionnaire on Cognitive Decline in the Elderly (score >3.38 indicative of dementia).  Age- and sex-specific incidence rates of dementia were calculated for 5,131 individuals aged 65 and older at study entry.  Dementia incidence over time was examined by comparing age-specific incidence rates between two age-period-cohorts derived in ELSA using a median split in birth-year range (ELSA I: 1900-1925 and ELSA II: 1926-1950). The predictive value of education, wealth, geographical area, the index of multiple deprivations and the degree of urbanization for dementia incidence were examined using Poisson regression.Dementia was determined by doctor-diagnosis and the Informant Questionnaire on Cognitive Decline in the Elderly (score >3.38 indicative of dementia).  Age- and sex-specific incidence rates of dementia were calculated for 5,131 individuals aged 65 and older at study entry.  Dementia incidence over time was examined by comparing age-specific incidence rates between two age-period-cohorts derived in ELSA using a median split in birth-year range (ELSA I: 1900-1925 and ELSA II: 1926-1950).The predictive value of education, wealth, geographical area, the index of multiple deprivations and the degree of urbanization for dementia incidence were examined using Poisson regression.Results The average duration of follow-up was 7.7 years.  As anticipated, there was a significant increase in the rates of dementia incidence with age, from 5.90 per 1,000 person years at ages 65-69 to 31.59 at ages 85+. We observed an age-period-cohort effect on dementia incidence, with nearly 30% reduction in the incidence rates for participants of the same initial age group (75-79) at the study entry from ELSA I and II. Higher rates of dementia incidence were observed for London (Rate=15.52, Incidence rate ratio (IRR)=1.43 95% confidence intervals (CI) 1.01-2.03) compared with Northern England, and for those in the lowest wealth quintiles (Rate=14.61, IRR=2.18, 95% CI 1.55-3.06) compared with the highest.Conclusion In a nationally representative sample, we observed a reduction in the age-specific dementia incidence over time.  This incidence appeared to be patterned by group- and individual-level characteristics.Keywords: dementia incidence, education, wealth, geographical area

P09

AFFECTIVE PROBLEMS AND MEMORY DECLINE: A SYSTEMATIC REVIEW AND META-ANALYSIS

A John1, U Patel1, M Richards2, J Rusted1, D Gaysina1

1Psychology, University of Sussex, Brighton, UK2MRC Unit for Lifelong Health & Ageing, UCL, London, UK

Background Previous evidence suggests that the presence of affective problems, such as depression and anxiety, may confer an increased risk for late-life dementia. However, the extent to which affective symptoms may influence memory decline in particular, even many years prior to the clinical threshold for a diagnosis of dementia, is not clear. The present study systematically reviews and synthesises the current evidence surrounding the association between depression and memory decline across the life course.Methods An electronic search of PubMed, PsycInfo and ScienceDirect was conducted to identify studies on the association between depression and subsequent memory decline. Key inclusion criteria were prospective, longitudinal studies with a minimum follow-up period of one year. Cross-sectional, experimental, and clinical (case-control) studies were excluded. Reference lists of relevant papers were scanned for any additional articles of interest. Data extraction and methodological quality assessment using the STROBE checklist were conducted independently by two raters. Multi-level meta-analyses were conducted, with consideration of a number of potential moderators, including mean age of sample at baseline, length of follow-up, and quality of study.Results After removal of duplicate references, all papers were screened for eligibility using a three-step process: 1) title screening (N=20,954); 2) abstract screening (N=981); and 3) full text screening (N=172). Inter-rater reliability at each stage of screening was >90%. A total of 17 studies (with 8 assessing depression as a binary variable and 9 assessing depression as a continuous variable) met eligibility criteria and had sufficient statistical information for extraction. The results of the meta-analyses will be presented and discussed with a focus on the effects of the key moderators that may influence the link between depression and memory decline, such as mean age at baseline, length of follow-up, and quality of study. Preliminary analyses suggest that affective problems significantly increase risk for subsequent decline in memory (Binary meta-analysis: OR=1.47; 95% CIs=1.15, 1.87, p=.002; Continuous meta-analysis: B=-0.007, 95% CIs=-0.011, -0.002, p=.003). Conclusions Results of the present study improve current understanding of the temporal nature of the association between depression and memory across the life-course. This has important implications for the identification of individuals who are at a particularly high risk for accelerated decline in memory function and dementia. Keywords: Affective problems, memory decline, meta-analysis

P10

Diet quality, sarcopenia and frailty in older men: cross sectional analysis from the British Regional Heart Study

TJ Parsons1, E Papachristou2, SE Ramsay3, JL Atkins4, O Papacosta1, LT Lennon1, S Ash1, PH Whincup5, SG Wannamethee1

1Primary Care & Population Health, UCL, London, UK2Institute of Education, UCL, London, UK3Institute of Health & Society, Newcastle University, Newcastle, UK4Epidemiology & Public Health, University of Exeter, Exeter, UK5Population Health Research Institute, St George's University of London, London, UK

Background Frailty, a vulnerability to adverse health outcomes, and sarcopenia, a decline in muscle mass and strength or performance are associated with ageing. Frailty and sarcopenia predict increased mortality and hospitalisation, and sarcopenia often occurs with an increase in body fat known as sarcopenic obesity which elevates these risks further. Diet quality is well established as a predictor of mortality, but few studies have investigated diet quality in relation to frailty or sarcopenia, and findings are inconclusive. We have therefore examined the associations between diet quality, frailty and sarcopenic obesity categories.Methods We used cross sectional data from community-dwelling men aged 71-91 years (British Regional Heart Study) in 2010-2012 recruited from 24 primary care practices. Men completed a food frequency questionnaire, from which the Healthy Dietary Index (HDI) and Elderly Dietary Index (EDI) were derived, and attended a physical examination. Frailty was based on the 5 components of the Fried frailty phenotype and we used a sarcopenic obesity classification which defines 4 groups; optimal, sarcopenic, obese or sarcopenic obese based on waist circumference and mid-arm muscle circumference. We used logistic regression models to investigate whether diet quality was associated with frailty and sarcopenia and/or obesity.Results 1331/3137 men (42%) had data for sarcopenia/obesity, all covariates and diet quality and 1119 men (36%) for frailty, covariates and diet quality. After adjusting for age, social class, region of residence, smoking, alcohol consumption, cardiovascular disease and energy intake, men in the top quartile of the HDI score had a lower odds of being frail (0.58 95% CI 0.34, 0.96) compared with men in the bottom quartile, and men in the top quartile of either HDI or EDI had a lower odds of being obese compared with men in the bottom quartile (0.52 95% CI 0.33, 0.84 and 0.57 95% CI 0.38, 0.86 respectively). Neither the HDI or EDI was associated with sarcopenia or sarcopenic obesity, and the EDI was not associated with frailty.Conclusion Higher diet quality based on both the HDI and EDI is associated with obesity but we found no evidence that diet quality is associated with sarcopenia in these elderly British men. However, our findings suggest that a higher diet quality as indicated by the HDI, a measure of adherence to WHO nutrient intake guidelines, might be relevant for the prevention or reversal of frailty.Keywords: Diet, ageing, frailty

P11

THE ENDURING INFLUENCE OF CONTROLLING PARENTING ON PERSONAL MASTERY IN OLDER AGE

F H Harkness, M Stafford, T Cosco, M Richards1MRC Unit of Lifelong Health and Ageing, UCL, London, UK2MRC Unit of Lifelong Health and Ageing, UCL, London, UK3MRC Unit of Lifelong Health and Ageing, UCL, London, UK4MRC Unit of Lifelong Health and Ageing, UCL, London, UK Background Personal mastery is the subjective feeling of control over the events in one’s own life. It is associated with healthy ageing, including better cardio-metabolic health, immune function and physical functioning. As an adult mastery is strongly associated with achievements of education, income and social class. However, within-group differences indicate that there could be other ways to feel in control. Mastery is theorised to be a self-concept first learnt in adolescence, and as such family may play a role in shaping it. Those whose parents support them psychologically and allow them appropriate freedom as an adolescent may grow up perceiving themselves to be in control, over and above tangible socio-economic resources.     Data The Medical Research Council National Survey of Health and Development (NSHD) is a representative sample of births in mainland Britain that occurred during a week in March 1946. Participants were (n=1,037) study members who had provided data at ages 4, 26, 43 and 68. Controlling parenting was measured using the Parental Bonding Instrument (PBI). This measures percieved parental levels of psychological control (e.g. invasiveness, overprotection). The outcome was personal mastery assessed at age 68 using Pearlin’s 7 item scale. An example item is, “ I have little control over the things that happen to me.” Multivariable regression analysis was used to test the association between psychologically controlling parenting and personal mastery age 68, controlling for childhood and adult socio-economic markers. Results Higher perceived parental psychological control was associated with lower mean mastery -0.12 (95% 0.20, -0.04) aged 68.  This association was not attenuated when adjusting for childhood and adult socio-economic position. Conclusion These findings show the lifelong importance of parenting on psychological self-concepts. Restricting young-life psychological freedom may build an enduring feeling that one is not agent of one’s own control. Despite having access to a comfortable socio-economic position, those whose parents were more controlling in adolescence felt less in control of their own life age 68. Keywords: parenting, control, ageing

P12

EVALUATING POLICY RESPONSES TO UPSTREAM DETERMINANTS OF CHRONIC, NON-COMMUNICABLE DISEASES: SUPPORTING HEALTHY DIETS AND ACTIVE LIVING IN SEVEN CARIBBEAN COUNTRIES

MM Murphy1, C Guell2, 3, TA Samuels1, 4, L Bishop1, N Unwin3, 4

1Faculty of Medical Sciences, University of the West Indies, Bridgetown, Barbados2European Centre for Environment & Human Health, University of Exeter Medical School, Truro, United Kingdom3MRC Epidemiology Unit and Centre for Diet and Activity Research, University of Cambridge, Cambridge, United Kingdom4Chronic Disease Research Centre, University of the West Indies, Bridgetown, Barbados

Background In 2007, heads of government in the Caribbean Community (CARICOM) committed to concerted policy action to address non-communicable diseases, whose burden was recognized as a threat to regional development. In 2015, a large mixed-method evaluation study investigated the progress made in developing and implementing relevant policies. As part of this, a qualitative study in seven Caribbean countries aimed to identify, assess and compare existing policies, gaps in policy responses, and the factors influencing successful policy development and implementation.Methods Policy document analysis was complemented by 76 semi-structured interviews with 80 relevant stakeholders in government, civil society and the private sector. Data collection and analysis protocols were developed iteratively. Interviews were audio-recorded and analyzed pragmatically framed by the WHO NCD Action Plan, a Multiple Streams policy approach and realist evaluation principles. An analysis team coded using Dedoose software, after which two lead researchers synthesized the analyses.Results Policy gaps existed regarding alcohol, diet and physical activity. Most widely reported successes across countries were policies and health promotion initiatives to support healthy eating in communities and in schools, including the development of dietary guidelines. Physical activity was targeted primarily in schools, with public participation in public sports events. Successful initiatives were often marked by collaboration between government ministries such as health, education and agriculture. There were very few existing policies around alcohol harm. The impact of these initiatives was reported as limited by adverse upstream determinants, including reliance on food imports entwined with the consequences of upholding existing trade agreements, which constrained more impactful fiscal and legislative action to support availability, quality and affordability of healthy foods. There was little evidence of policy responses to create physical and social environments conducive to active living, such as to support greater walking. The lack of focus on policy creation, lack of monitoring and evaluation of existing interventions, and limited staffing and financial capacities limited success in policy implementation. For alcohol specifically, national and political cultural attitudes and its economic importance deemed it untouchable from a policy standpoint.Conclusion The least well developed policy responses concern upstream determinants of unhealthy diets, physical inactivity and alcohol harm. Local and regional political support is essential to accelerate action to support environments conducive to healthy eating and active living. Addressing reliance on food imports and existing trade agreements as well as developing and implementing policies to reduce alcohol related harm necessitates regional cooperation for a unified response.Keywords: policy evaluation, non-communicable diseases, upstream determinants

P13

SOCIAL, BIOLOGICAL, BEHAVIOURAL AND PSYCHOLOGICAL FACTORS AND PHYSICAL ACTIVITY DURING PREGNANCY: A CROSS-SECTIONAL STUDY

C FlanneryNUIG, D DahlyUCC, M ByrneNUIG, Ali KhashanUCC, S McHughUCC, L KennyUCC, FM McAuliffeUCD, PM KearneyUCC

1Health behaviour change research group, National University of Ireland, Galway, Ireland2HRB Clinical Research Facility Cork, University College Cork, Cork, Ireland 3Department Epidemiology and Public Health, University College Cork, Ireland4Department Obstetrics and Gynaecology, University College Cork, Ireland 5Obstetrics & Gynaecology, School of Medicine, University College Dublin, Ireland

Background Regular Physical Activity (PA) during pregnancy is associated with a number of health benefits such as preventing pregnancy complications, limiting pregnancy weight gain, and decreasing the risk of gestational diabetes. Despite this, women’s PA levels often reduce or cease during pregnancy. Various factors like body mass index, unemployment, and educational levels have been shown to be associated with PA levels during pregnancy. Using the Screening for Pregnancy Endpoint (SCOPE) data, this study aimed to examine the association of health behaviours and psychological well-being on PA levels in pregnancy using the biopsychosocial model. Methods Nulliparous women with a singleton pregnancy were recruited from a large academic maternity hospital in Cork, Ireland (CUMH) as part of the international SCOPE study. Data was collected at 15±1 week’s gestation including personal information, lifestyle and psychological measures. The outcome of interest was PA categorised as low, moderate or high levels and covariates were selected using the biopsychosocial model including social (age; years of schooling; socioeconomic status), biological (body mass index), behavioural (diet; smoking) and psychological factors (anxiety; response to pregnancy). Univariate and multivariable multinomial logistic regression examined the association between covariates and PA using Stata IC13. Results 1774 pregnant women were recruited to SCOPE and women were classified as having low (22.3%), moderate (54.4%) or high (23.4%) PA levels (n=1766). In the fully adjusted model, preliminary findings indicate that age categories 30-34 years (OR 2.34 [95% CI: 1.26-4.37]), ≥35 years (OR 2.06 [95% CI: 1.01-4.19]) compared to those <25 years were positively associated with high PA levels. Having more than 12 years of schooling and a higher socioeconomic status remained significant (P<0.05) for moderate PA levels. Five portions of fruit and veg a day (OR 1.89 [95% CI: 1.21-2.94]), fish consumption (OR 1.47 [95% CI: 1.07-2.03]) and psychological response to pregnancy (OR 1.04 [95% CI: 1.01-1.08]) were also positively associated with high levels of PA. Women who reported smoking in their first trimester were 31% less likely to be in the high physical activity group compared to those who reported no smoking (OR 0.69 [95% CI: 0.49-0.99]). None of the biological factors were associated with high levels of PA. Conclusion Social and behavioural factors appear to be the most influential factors driving PA levels. The results should be used to guide the development of educational campaigns and behaviour change intervention to encourage younger women to be active and to target those with lower education attainment and of lower socioeconomic status. Keywords: Physical activity levels, pregnancy, determinants, predictors, biopsychosocial model

P14

ASSOCIATION OF SLEEP DURATION AND SLEEP QUALITY WITH THE LOSS OF FUNCTIONING AMONG AUSTRALIAN PEOPLE

TL Lallukka1, 2, BS Sivertsen3,4,9 ,EK Kronholm1, YSB Bin5, SØ Øverland6, 7, NG Glozier8, JIH Halonen1

1Research and Service Centre for Work Organizations, Finnish Institute of Occupational Health, Helsinki and Turku, Finland2Department of Public Health, University of Helsinki, Helsinki, Helsinki3Department of Health Promotion, Norwegian Institute of Public Health, Bergen, Norway4The Regional Centre for Child and Youth Mental Health and Child Welfare, Uni Research Health, Bergen, Norway5Central Clinical School, Sydney Medical School, University of Sydney, Sydney, Australia6, Norwegian Institute of Public Health, Bergen, Norway7Department of Psychosocial Science, University of Bergen, Bergen, Bergen8Brain and Mind Centre, University of Sydney, Sydney, Australia9Department of Research and Innovation, Helse Fonna HF, Haugesund, Norway

Background While short and long sleep duration and poor sleep quality have individually been linked to work disability and ill-health, their joint impact on functioning has received less attention. The aim was to examine the joint associations of sleep quantity and quality with physical, emotional, and social functioning, while considering the contribution of key potential confounders.Methods Data from the nationally representative Household, Income and Labour Dynamics in Australia (HILDA) Study, collected in 2013, were used (n=14557). Poor sleep quality (yes/no) was examined in combination with normal (6-8 hours), short (<6) or long (>8) sleep. Short form SF-12 questionnaire was used to examine functioning. Covariates included sociodemographic factors, social connections (meeting with friends), health behaviours, obesity, pain, and mental and physical illnesses. Logistic regression analysis was used (odd ratios, OR and their 95% confidence intervals, 95% CI).Results After adjusting for sex and age, short, normal and long sleep duration in combination with poor quality were all associated with poorer physical, emotional and social functioning. The associations were relatively similar for physical and emotional functioning, and strongest for social functioning (OR 3.36, 95% CI 2.94-3.84 for normal sleep with poor sleep quality, OR 4.56, 95% CI 4.02-5.17 for short sleep with poor sleep quality and OR 7.83, CI 95% 6.09-10.09 for long sleep with poor sleep quality). Also people with good sleep quality but short or long sleep had poorer functioning, but the associations were weaker for all forms of functioning (ORs ranging 1.28-1.57 after adjusting for sex and age). Pain and comorbid illness attenuated all of the associations the most, while the contributions of the other covariates were relatively minor. However, the associations remained after full adjustments for people with poor sleep quality and short, normal or long sleep, and for people with good sleep quality but long sleep duration.Conclusion The results highlight that in particular poor sleep quality is likely to have the most severe consequences for daytime functioning in the general population, and that there is little difference between short and normal sleepers in the presence of poor quality sleep.Keywords: sleep; functioning; general population

P15

INEQUALITIES IN ALCOHOL-RELATED MORTALITY: AN ELECTRONIC COHORT STUDY OF INTERNAL MIGRATION IN WALES

A Gartner1, D Farewell1, G Greene1, L Trefan1, S Paranjothy1, D Fone1

1Division of Population Medicine, School of Medicine, Cardiff University

Introduction Alcohol-related deaths show a strong socioeconomic gradient. Selective migration, when individuals in poorer health move to more deprived areas, or vice versa, is associated with widening health inequality. However, it is not known whether recent internal migration contributes to the observed health inequality in alcohol-related deaths. This study aims to investigate whether selective migration of individuals is associated with the alcohol-related mortality gradient.Methods We constructed a record-linked electronic cohort of the total population of Wales registered with a GP on 01/01/2006 (n=3 136881) with six-years’ follow-up to 31/12/2011 in the SAIL Databank at Farr@Swansea.  The Welsh Demographic dataset provided information on age, sex, and the lower super output area (LSOA) (n = 1896) of residence at each of the 24 quarters, used to code migration. To each LSOA code we linked tenths of the Welsh Index of Multiple Deprivation 2008 and the ONS Rural-Urban definition. To code alcohol-related mortality (n=2224 of 184 247 deaths) using the standard ONS definition we linked the demographic dataset to the ONS mortality files.We estimated hazard ratios (HR) with 95% confidence intervals (95% CI) using Cox regression for the risk of alcohol-related death for deprivation fifths, adjusting for age-group, sex and rural-urban status. The first model estimated HRs allowing for changing deprivation fifths linked to each move and the second model fixed the deprivation fifth at the start of the study period. A third model considered deprivation change between high (the five most deprived tenths) and low deprivation (the five least deprived tenths).Results We found 495 subjects moved between LSOAs prior to an alcohol-related death. Of these, the majority (260) moved between high deprivation LSOAs and 75 moved between low deprivation LSOAs. Slightly more subjects moved from areas of low to high deprivation (98) than the opposite (62).In model 1 subjects living in the most deprived fifth had a higher risk of alcohol-related death (HR 3.20; 95% CI 2.77 to 3.71) compared to subjects in the least deprived fifth. In model 2 the risk was little different (HR 2.96; 95% CI 2.56 to 3.42). Formal statistical adjustment for recent migration had little effect on the inequality estimates. In model 3, there was an elevated risk of alcohol-related death for moving from low to high deprivation.Conclusion Internal migration within Wales over a six-year period does not explain the observed inequality in alcohol-related mortality.Keywords: Alcohol, migration, inequality

P16

GENDER DIFFERENCES IN WALKING (FOR LEISURE, TRANSPORT AND IN TOTAL) ACROSS ADULT LIFE: A SYSTEMATIC REVIEW

TM Pollard, JM Wagnild1Department of Anthropology, Durham University, Durham, UK

Background The possibility of gender differences in walking practices has largely been ignored, but an understanding of whether and how walking practices differ between men and women would usefully inform interventions to promote walking. The aim of this systematic review was to examine gender differences in walking for leisure, transport and in total in adults, and to assess whether gender differences in walking practices change across the adult life-course. Methods Web of Science Core Collections, PubMed and the Transportation Research International Documentation databases were searched to identify relevant studies, as well as reference lists of included studies. Papers providing quantitative data on the walking of both men and women aged 18 years and above in a high-income country and published between 1995 and 2015 were identified. Data assessed as high quality were extracted and results were synthesised using forest plots and narrative summary. Results Of the 24,498 studies retrieved by the search strategy, 36 studies were included.  Most (30) studies reported data on the prevalence of walking for a minimum duration, usually 10 minutes per week, either in total or for a specific purpose.  Of these, 6 of 6 studies combining all ages found that more women than men walked for leisure (Odds Ratios (ORs) 1.12-1.47). There was no evidence of a gender difference in overall walking for transport (ORs 0.78-1.11 for 3 studies combining all ages), but 3 of 3 studies found that more women than men walked to undertake errands (ORs 1.08-2.97).  5 of 6 studies combining all ages reported no gender difference in total walking (ORs 0.90-1.39).  6 studies reported on walking for leisure by age group; more young women than young men walked for leisure (ORs 1.54-1.99), but this difference consistently declined with age such that in the oldest age groups it had disappeared or reversed (ORs 0.63-0.96).  There was some evidence for a similar pattern in overall walking.Conclusion While we found no gender difference in participation in walking overall, we did find consistent gender differences in participation in walking for some purposes, including for leisure, and that there are gender differences in the impact of age on walking. We conclude that more research is needed to improve our understanding of how walking fits into the lives of women and men across the life-course, especially in relation to gender differences in the impact of aging on walking.Keywords: physical activity, walking, gender

P17

CHRONIC MALNUTRITION IN EARLY LIFE AND COGNITIVE OUTCOME IN MID-CHILDHOOD: EVIDENCE FROM FOUR DEVELOPING COUNTRIES 

Z ZHANG1, T CHANDOLA1

1CMIST, University of Manchester, Manchester, the UK

Background Chronic malnutrition is a major global health challenge, particularly in low- and middle-income countries (LMICs). Malnutrition, in particularly stunting in early childhood, is associated with developmental delay, including cognitive deficits and poor academic performance, often leading to reduced work capacity in adulthood. Few studies have examined the long-term influence of chronic malnutrition in early life on subsequent cognitive outcome and how recovering from early stunting can benefit cognitive development in a cross-national context.Methods Analysing data on 8062 children in Ethiopia (n=1,999), India (2,011), Peru (n=2,052) and Vietnam (n=2,000) from the Young Lives Study, a cross-national cohort study on child poverty and wellbeing in the developing world, we studied younger cohort, enrolled in 2002 at ages 6-17.9 months (round 1), followed in 2006 at 5 years old (round 2) and in 2009 at 8 years old (round 3) to explore whether duration of chronic malnutrition in early life matters for cognitive outcome and whether recovery from early stunting benefits for cognitive development.The duration of stunting was defined as having a HAZ Height-for-Age Z scores (HAZ) < -2 based on WHO standards. Chronic malnutrition was determined as the number of stunting for three rounds, ranging from 0 to 3. Recovery from stunting was defined as having a HAZ >-2 in subsequent rounds after baseline. Cognitive outcome was assessed at age 8 using the Peabody Picture Vocabulary Test (PPVT) to evaluate vocabulary skills and listening comprehension. Results Longer term of being stunted was associated with lower cognitive scores at age 8 for both girls (p<0.05) and boys (p<0.01). We found catch-up growth from early stunting significantly increases cognitive outcomes of girls who were stunted at baseline (coef=7.91, p<0.01). No significant associations were found for boys. Chronic malnutrition during childhood can be detrimental for children’s cognitive development in later life. Recovery from early stunting is possible, in particular for girls and this can lead to improvement in cognitive development. Conclusion Policy makers and program planners should consider redoubling efforts to prevent stunting and promote catch-up growth over the first few years of life, especially for girls, as a way of improving children's physical and intellectual development. Given the important role girls and women play in contributing to national and global economies, improving nutrition in early childhood of girls in developing countries is a long-term economic investment and driver of economic growth.Keywords: nutrition, cognitive, life course, developing countries

P18

IS THE RISK OF PERMANENT CHILDHOOD HEARING IMPAIRMENT HIGHER IN BABIES OF SHORTER GESTATIONAL LENGTH?  

E Butcher, C Dezateux, R Knowles1Life Course Epidemiology and Biostatistics, Great Ormond Street Institute of Child Health, UCL, London, England

Background Permanent childhood hearing impairment (PCHI) has major implications for social and cognitive development over the life-course. Early detection and management improves outcomes, providing the rationale for universal newborn hearing screening (UNHS), introduced in England from 2002. In England & Wales, one baby in every eight is admitted to specialist neonatal care, including neonatal intensive or special care baby units (NICU/SCBU), a known risk factor for PCHI, however the association with gestational length remains unclear. In England & Wales, 7.6% infants are born before 37 weeks gestation; this is higher (16.2%) among those of South Asian ethnicity. We examined PCHI risk by age 11 years in a prospective UK-wide cohort of children born before UNHS. We tested the hypothesis that lower gestational length is associated with higher PCHI risk after adjusting for NICU/SCBU admission and ethnicity.Methods PCHI risk (cumulative incidence) was based on parental report of hearing impairment and associated treatment at ages nine months, three, five, seven and 11 years for 19,518 children participating in the Millennium Cohort Study. The association of PCHI (defined by provision of hearing aids/cochlear implants, persisting hearing impairment at final report, and absence of glue ear) with gestational length was investigated using multivariable discrete-time survival analysis, adjusting for NICU/SCBU admission, ethnicity and other confounding factors, and weighting for survey design (Stata: Release 14; StataCorp LP).Results Parents reported no health problems in the first week after birth in 10,247 (52.4%), children, neonatal illness without NICU/SCBU admission in 6,781 (38.0%), and NICU/SCBU admission in 1,785 (9.6%). PCHI was ascertained in 44 children by age 11 years. PCHI risk was 1.0 per 1000 children (95% CI: 0.6-1.6) by age 9 months, rising by age five to 1.5 (1.0-2.2) and by age 11 to 2.1 (1.5-3.0). PCHI risk by age 11 was not associated with gestational length (hazard ratio (HR): 1.00, 95% CI: 0.98-1.03), but was associated with parental report of neonatal illness with or without NICU/SCBU admission (HR: 6.33; 2.27-17.63 and 2.62; 1.15-5.97, respectively) and Bangladeshi or Pakistani ethnicity (HR: 2.78; 1.06-7.31). Conclusion In this cohort, born before UNHS, PCHI risk was highest in infancy. Neonatal illness, irrespective of NICU/SCBU admission, and not gestational length, increased PCHI risk by age 11 years. Further research should explore the observed increased PCHI risk in children of Bangladeshi or Pakistani ethnicity, and the relevance for UNHS of PCHI with onset or diagnosis after infancy.ESRC-funded PhD: ES/J500185/1.Keywords: epidemiology child health

P19

FATHER INVOLVEMENT AND ATTITUDES IN EARLY CHILD-REARING AND DEPRESSIVE SYMPTOMS IN THE PRE-ADOLESCENT PERIOD IN A UK BIRTH COHORT

C Opondo1, M Redshaw1, M Quigley1

1Nuffield Department of Population Health, University of Oxford, Oxford, UK

Background Although much of the research on the influence of parenting on child development has emphasised the mother’s role, increasing evidence highlights the important role of fathers in the development, health and well-being of their children. In this study we aimed to investigate the association between father involvement in early child upbringing and depressive symptoms in the same children in their pre-teenage years.Methods Data for the study came from the Avon Longitudinal Study of Parents and Children (ALSPAC) cohort from the southwest of England. The outcome, depressive symptoms, was measured using the short Moods and Feelings Questionnaire (sMFQ) when the children were 9 and 11 years of age. The primary exposure was paternal involvement measured using scores derived from factor analysis of fathers’ report of their participation in, understanding of, and feelings about their child’s early upbringing.Results Three factors were identified in the factor analysis. Scores on factor 1 measured fathers’ emotional response to the child; scores on factor 2 measured the frequency of fathers’ involvement in domestic and childcare activities; scores on factor 3 measured fathers’ feelings of security in their role as parent and partner. Children of fathers with high scores on factors 1 and 3 had 12% (OR 0.88, 95%CI 0.81–0.97, p=0.009) and 9% (OR 0.91, 95%CI 0.82–1.00, p=0.040) respectively lower adjusted odds of reporting more depressive symptoms at 9 years. There was no evidence of a difference in depressive symptoms associated with factor 1 and factor 3 scores at 11 years. However, there was weak evidence of a 13% increase in odds of reporting more depressive symptoms associated with 1 unit increase in factor 2 scores at age 9 years (OR 1.13, 95%CI 0.99–1.29, p=0.061) and a 16% increase at 11 years (OR 1.16, 95%CI 1.01–1.34, p=0.040).Conclusion Positive psychological and emotional aspects of father involvement in children’s early upbringing, but not the quantity of direct involvement in childcare, may protect children against developing symptoms of depression in their pre-teen years.Keywords: fathers; father involvement; child-rearing; depression; ALSPAC cohort

P20

CUMULATIVE EFFECT OF ADVERSE CHILDHOOD EXPERIENCES ON AFFECTIVE SYMPTOM TRAJECTORIES IN ADULTHOOD: EVIDENCE FROM A BRITISH BIRTH COHORT

EJ Thompson1, M Richards2, D Gaysina1

1School of Psychology, University of Sussex, Brighton, UK2MRC Unit for Lifelong Health and Ageing, University College London, London, UK

Background Previous studies have shown that specific types of adverse childhood events (ACEs), such as parental divorce and parental psychopathology, pose a risk for the development of affective symptoms in adulthood (AS). However, a majority of this evidence is based on single types of retrospectively reported ACEs. This is problematic as ACEs tend to be inter-related and often co-occur.Methods We used the data from the MRC National Survey of Health and Development (NSHD). This is an ongoing longitudinal study of 5362 women and men who were born in Britain in1946. Later life AS were measures using the General Health Questionnaire (GHQ) at ages 53 y, 60-64 y and 69.Multiple imputation was implemented on each ACE predictor and a cumulative risk index was derived though summing the number of adversities experienced by each participant (0, 1, 2, 3…20) before age 16 y. The effect of cumulative ACEs on AS at each time point (53, 60-64 and 69) was examined using linear regression. Results Preliminary analyses revealed a significant association was found between cumulative ACEs and AS at ages 60-64, β(1, 2183) = .07, p = .002, and 69, β(1, 2110) = .07, p = .003, but not age 53 β(1, 2900) = .04, p = .058. Further to this growth mixture modelling will be used to model latent trajectories of AS between age 53 and 69 years and the effect of cumulative ACEs will be examined. Discussion These findings will be presented in light of the growing evidence for the negative effects of ACEs on health and wellbeing in later life. Furthermore, we will discuss how this research informs prevention for the development of psychopathology across the life course. Keywords: cumulative risk; mental health; longitudinal data analysis

P21

PATTERNS OF BREASTFEEDING AT 3 MONTHS POSTPARTUM IN THE ALBERTA PREGNANCY OUTCOMES AND NUTRITION (APRON) STUDY

M Jarman1, PJ Robson2, Y Shen3, M Madsen1, RC Bell11Dept of Agriculture Food and Nutrition Science, University of Alberta, Edmonton, Canada2Cancer Measurement, Outcomes, Research and Evaluation, Cancer Control Alberta, Alberta Health Services, Edmonton Canada3School of Public Health, University of Alberta, Edmonton, Canada12Breast feeding Patterns Prospective cohort

Background Infant feeding is often characterized in broad categories, such as exclusive breastfeeding, partial breastfeeding and formula feeding. However there is little information about detailed patterns of feeding within these groups. The purpose of this study was to examine patterns of breastfeeding from 3-day prospective feeding diaries to describe how women feed their babies.Methods The APrON study is a prospective study of women during pregnancy and their children. At 3 months postpartum women completed a prospective breastfeeding diary which collected information on the number of feeds, and duration and method of every feed over 3 days. A total of 1080 women completed the breastfeeding diaries. Results Women reported feeding their babies using combinations of up to 4 different methods (at breast, expressed breastmilk in a bottle, formula, mixed breastmilk and formula in the same bottle) in any one day. For the ongoing analyses women were categorized into 4 groups based on the number and type of feeds reported/day: 1) at breast only (n=622), 2) at breast and expressed breastmilk (n=223), 3) mixed feeding using 2 methods (n=123) and, 4) mixed feeding using 3 or 4 methods (n=112). The proportion of total feeds at breast was 80%, 75% and 64% for the mothers in the ‘at breast and expressed milk’, ‘mixed feeding using 2 methods’ and ‘mixed feeding of 3 using 4 methods’ groups, respectively. In turn the average daily duration of at breast feeds was longer in women who fed at breast only compared to those in any other method group. Women who fed ‘at breast and expressed milk’, ‘mixed feeding of 2 methods’ and ‘mixed feeding of 3 or 4 methods’ fed, on average, for 18 minutes, 17 minutes and 27 minutes less at breast per day, respectively, compared to women who fed at breast only (all P=<0.01). There were no differences in the average number of daily feeds between the 4 method groups.Women use many different combinations of methods within a day to feed their 3 month old babies. Using fewer groups to classify women ignores the ‘real life’ complexity of feeding patterns. Our data showed that those who used a combination of methods to feed their babies fed had a lower a frequency and total duration of feeding at breast. This could be important to understanding variation in maternal energy expenditure and infant growth, development and health in the future.In collaboration with the ENRICH team.Keywords: Breast feeding Patterns Prospective cohort

P22

CHILDHOOD ADVERSITY, ADULT SOCIOECONOMIC STATUS AND RISK OF WORK DISABILITY – A PROSPECTIVE COHORT STUDY

JI Halonen1, M Kivimäki1,2,3, J Vahtera4,5, J Pentti3, M Virtanen1, J Ervasti1, T Oksanen1, T Lallukka1

1Work Disability Prevention, Finnish Institute of Occupational Health, Helsinki, Finland2Department of Epidemiology and Public Health, University College London, London, UK3Department of Public Health, University of Helsinki, Helsinki, Finland4Department of Public Health, University of Turku, Turku, Finland5-, Turku University Hospital, Turku, Finland

Background Employees’ health is dependent on individual and social factors operating from across the life course. Studies have shown that negative life events during childhood or low socioeconomic status (SES) in adulthood may predict decreased labour market participation. However, the combined effects of childhood adversities and low SES in adulthood on work disability are not known. Methods We included 34384 employed Finnish Public Sector study participants who responded to questions about childhood adversities (none versus any adversity, e.g., parental divorce or financial difficulties) in 2008, and whose adult SES (based on occupational status) in 2008 was available. We categorized exposure into four groups: neither (reference), childhood adversity only, low SES only, or both. Participants were followed from 2009 until the first period of register-based work disability (sickness absence >9 days or disability pension) due to any cause, musculoskeletal or mental disorders; retirement; death; or end of follow-up (December 2011). We ran cox proportional hazard models adjusted for behavioral, health- and work-related covariates. Results When compared to those with neither exposure, hazard ratio (HR) for work disability from any cause was increased among participants with childhood adversity, with low SES, and those with both exposures. Low adult SES was a stronger predictor (HR 2.38, 95% CI=2.14 to 2.64) of disability due to musculoskeletal disorders than childhood adversity (1.11, 95% CI=1.00 to 1.23). The difference between the exposures as predictors of disability due to mental disorders was smaller (adversity: 1.40, 95% CI=1.23 to 1.59; low SES: 1.21, 95% CI=1.03 to 1.42). The highest risk was observed in those with both exposures (2.53, 95% CI=2.29 to 2.79 for musculoskeletal disability). Discussion Exposure to adversities in childhood and low SES in adulthood were associated with work disability. Exposure to both these risk factors was associated with the highest work disability risk, although this was additive rather than synergistic effect. Childhood adversity was associated with disability due to mental disorders in particular, whereas low adult SES was more strongly associated with disability due to musculoskeletal disorders. Keywords: childhood adversity; socioeconomic status; work disability

P23

POSSIBLE IMPLICATIONS OF A PERCEIVED DIVIDE BETWEEN LAY AND OFFICIAL KNOWLEDGE IN AN AREA OF COMPROMISED AND POLLUTED ENVIRONMENT, AND HIGH DEPRIVATION, IN NORTH WEST ENGLAND

P Elliott1, N Payne2

1Geography, Lancaster University, Lancaster, UK2School of Health and Related Research, University of Sheffield, Sheffield, UK Background Due to its topography and history, Halton, Merseyside, provides a valuable natural experiment to investigate the divide between lay and official knowledge about environment and deprivation, and how these may influence health choices and outcomes.  Methods To gain a cross-section of the local population, households in every street in each of six predefined areas received an invitation. Fifty-five semi-structured interviews were carried out to assess residents’ perceptions and experiences of health, environment and lifestyle. A life history approach was taken to examine health over time, and how lifestyle choices were constructed.Interviews were taped, transcribed, and analysed for themes using a Structuration Theory approach. Residents’ viewpoints were then presented to twenty local public organisation officials, whose own perceptions were also obtained. Results While local residents were aware of health messages, these were adopted out of personal choice, not in order to accord with perceived edicts. Residents expressed scepticism about how official knowledge was constructed, and dismay that official knowledge viewed ill-health as the result of lifestyle rather than environmental or socio-economic factors.There was a perception that enduring deprivation was sought to be maintained in order to attract funding which would benefit officials with vested interests who lived outside the borough and could therefore afford to be unconcerned about its problems.An illustrative quotation from a resident is: "Why kill the goose that laid the golden egg?  Why get rid of poverty, deprivation and poor health when these big pots of money keep rolling in?".Discussion Health messages, like other forms of knowledge, may be interpreted more from the presentation and perceived motivation for issuing the message than from the content of the message. Where there is a divide between lay and official knowledge, this motivation may be seen as a means of asserting power rather than either a scientifically-based or an altruistic programme of education or information.If lifestyles are viewed by officials as the prime factor influencing an area’s health profile, this view may be interpreted by lay knowledge as a means of social control meaning that the perceived root causes of socio-economic inequality need not be tackled, and that the priority of interventions is to maintain a privileged position.Keywords: environment, deprivation, health

P24

INDIVIDUAL FACTORS ASSOCIATED WITH SELF-REPORTED MEASURES OF COLD HOMES DURING WINTER

C Sartini1, P Tammes2, A Hay2, I Preston3, D Lasserson4, PH Whincup5, SG Wannamethee1, RW Morris2

1Primary care and Population Health, UCL, London, UK2School of Social and Community Medicine, University of Bristol, Bristol, UK3Household Energy Services, Centre for Sustainable Energy, Bristol, UK4Nuffield Department of Medicine, University of Oxford, Oxford, UK5Population Health Research Institute, St George's University of London, London, UK

Background An estimated 9,000 people died during the winter 2014-2015 in England and Wales from living in a cold home. Older people are susceptible to cold, but it is unclear how to identify those who particularly find it hard to keep warm in winter. The aim of this study was to identify individual factors associated with self-reported measures of cold homes among older people. Methods Data were from the British Regional Heart Study (BRHS) of older men, who were aged 74-95 when completing a questionnaire in 2014. This included four self-reported measures of cold housing during the previous winter (outcomes): (i) having difficulties in meeting the heating/fuel costs; (ii) staying in bed longer in order to stay warm; (iii) turning the heating off even when cold because of worries about the costs; (iv) keeping the living room comfortably warm. Individual data, including demographic characteristics, health and lifestyle factors were also collected. Cross-sectional associations between individual factors and measures of cold housing were analysed using logistic regression models.Results Descriptive statistics showed that (i) 327 out of 1608 (20.6%) men had difficulties in meeting the heating/fuel costs; (ii) 210 (13.3%) stayed in bed longer in order to stay warm; (iii) 157 (10.2%) turned heating off because of worries about the costs, and (iv) 54 (3.4%) could not keep comfortably warm in the living room. In full adjusted logistic models, some individual factors were independently associated with the four outcomes (p<0.05): manual social class, having more financial difficulties, feeling isolated from others, and being not married. The relationship between reporting general financial difficulties and difficulties in meeting the heating/fuel costs was particularly strong (Odds Ratio [OR] = 4.9, 95% Confidence Interval [CI] 3.9; 6.1). Also, men with mobility limitations were twice as likely to stay in bed longer in order to stay warm (OR=2.0, 95%CI 1.4; 2.9). Other individual factors, such as living in a house centrally heated and types of house insulation, as well as a proxy measure of the house energy efficiency (Energy Efficiency rating, aggregated from households within participants’ Lower Super Output Area) were not related to self-reported measures of vulnerability to cold.Conclusion Findings suggested that in older people financial difficulties and social class are key factors associated with cold housing in winter.Keywords: cold homes; social class; winter deaths

P25

FEASIBILITY CLUSTER RANDOMISED CONTROLLED TRIAL AND PROCESS EVALUATION OF AN ENVIRONMENTAL INTERVENTION IN NURSERIES AND A WEB-BASED HOME INTERVENTION TO INCREASE PHYSICAL ACTIVITY, ORAL HEALTH AND HEALTHY EATING IN CHILDREN AGED 2–4 YEARS: NAP SACC UK

RR Kipping1, R Langford1, J White2, C Metcalfe1, A Papadaki3, W Hollingworth1, L Moore4, R Campbell1, D Ward5, R Jago3, R Brockman1, S Wells 1, A Nicholson 1, J Collingwood 1

1Social & Community Medicine, University of Bristol, Bristol, England2School of Medicine, Cardiff University, Cardiff, Wales3School of Policy Studies, University of Bristol, Bristol, England4MRC/CSO Social and Public Health Sciences Unit, University of Glasgow, Glasgow, Scotland5Gillings School of Global Public Health, University of North Carolina at Chapel Hill, North Carolina, USA

Background Systematic reviews have identified the lack of intervention studies to prevent obesity in young children. Most 3 year old children in the UK attend formal childcare, and the Government plans to extend free childcare to 30 hours per week for 3 and 4 year olds; therefore these settings present an opportunity to improve health. The Nutrition and Physical Activity Self Assessment for Childcare (NAP SACC) programme aims to improve child nutrition and physical activity through changes to the nursery environment. Feasibility and acceptability have been demonstrated through Randomised Controlled Trials (RCT) in the USA. This study examined the feasibility and acceptability of adapting the NAP SACC intervention for the UK.Methods A feasibility cluster RCT in 12 nurseries with 2-4 year olds in the southwest region of England. Focus groups and interviews with Health Visitors (community children’s nurses), nursery staff and parents informed adaptation of the intervention for the UK. The intervention comprised: two staff workshops on physical activity and nutrition; Health Visitor support to review nursery practices against 80 areas of best practice, set goals and make changes; a digital media-based home component. Measures were assessed at baseline and post-intervention: zBMI, accelerometer-measured physical activity and sedentary time, diet, child quality of life, health care usage, parental and nursery staff mediators and quality of nursery environment. Fidelity and acceptability were assessed through observation and interviews analysed via thematic analysis.Results Formative work resulted in the following adaptations: inclusion of an oral health component; changes to confirm with UK guidance; specialist workshop facilitators; and development of the home component. 168 (37%) eligible children were recruited from 12 nurseries. Interviews were completed with four Health Visitors, 17 nursery staff and 20 parents. The intervention was implemented with high fidelity, with two exceptions: one nursery did not implement the intervention due to staff workload; and the digital home component was used by just 12 (14%) parents. Intervention acceptability was high. A mean of seven staff per nursery attended each workshop. The workshops and Health Visitor contact were highly valued. The mean number of goals set was eight. Nursery changes included: menu modifications, reducing portion sizes and sugary snacks, role modelling physical activity and eating, and active story telling. The trial design and methods were highly acceptable.  Descriptive analysis of the outcomes will be available by September 2017.Conclusion NAP SACC UK is feasible and acceptable with the exception of the home component; effectiveness should be tested through a full-scale RCT. Keywords: Early years; feasibility RCT; health improvement

P26

ASSOCIATIONS BETWEEN TYPOLOGIES OF NEIGHBOURHOOD ENVIRONMENTS AND ASSOCIATIONS WITH OBESITY: A CROSS-SECTIONAL STUDY

M Hobbs1,2, C Griffiths2, M Green3, J Saunders4, H Jordan5, J McKenna2

1School of Social and Health Sciences, Leeds Trinity University, Leeds, England2Carnegie, Leeds Beckett University, Leeds, England3Geography and Planning, The University of Liverpool, Liverpool, England4Carnegie, Leeds Beckett University; formerly, Rotherham Borough Council, Rotherham, England5School of Health and Related Research (ScHARR), The University of Sheffield, Sheffield, England

Background Recent research has demonstrated that neighbourhood features such as fast-food outlets and supermarkets may co-occur. However, little research has investigated the combined influences of both the built food and physical activity (PA) environments and associations with body mass index and obesity. This study aims to use latent class analysis within a large UK adult population to investigate associations between the combined environment and obesity.Methods Cross-sectional, individual-level data (n=22,889) from Wave 1 of The Yorkshire Health Study (2010-2012) were used. Body mass index (BMI) was calculated using self-reported height and weight; obesity=BMI≥30. Neighbourhood was defined as a 2km radial buffer; food outlets and physical activity facilities were sourced (2012) from Ordnance Survey Points of Interest (PoI) and categorised into ‘fast-food’, ‘large supermarkets’, ‘convenience and other food retail outlets’ and ‘physical activity facilities’. Parks were sourced from Open Street Map. Latent class analysis (LCA) was conducted on these five environmental variables. Logistic regression was then conducted to predict obesity based on the five neighbourhood types identified within LCA. Models adjusted for age, gender, ethnicity, area-level deprivation and rural or urban classification of the neighbourhood.Results A five-class solution fitted the dataset best and was interpretable. Neighbourhood typologies were labelled as; “low exposure” (19.0% of study population); “moderate exposure” (33.3%); “moderate PA, limited food” (12.2%); “saturated” (13.6%); “moderate PA, ample food” (21.2%). For associations with obesity, those within the low exposure typology were chosen as the exposure because low exposure to physical activity environments have the potential to reduce physical activity behaviours and although more debatable poorer access to the food environment may result in poorer dietary intake. Compared to the low exposure, one typology showed lower odds of obesity (“saturated”, OR=0.86 [0.75,0.99]) and one showed increased odds of obesity (“moderate exposure, OR=1.18 [1.05,1.32]. Discussion Meaningful neighbourhood typologies were derived from a range of food and physical activity measures using latent class analysis which explained differences in obesity in large UK based sample of adults. This study suggests that neighbourhoods were not wholly unhealthy or healthy, they were characterised by neighborhood features that are both health-promoting and health-constraining and this resulted in complex associations with obesity.Keywords: built environment, obesity, latent class analysis.

P27

CAN WE BETTER CAPTURE LONGITUDINAL EXPOSURE TO THE NEIGHBOURHOOD ENVIRONMENT? A LATENT CLASS GROWTH ANALYSIS OF THE OBESOGENIC ENVIRONMENT IN NEW YORK CITY, 1990-2010

N Berger1, TK Kaufman2, MDM Bader3, DM Sheehan2, SJ Mooney4, KM Neckerman5, AG Rundle2, GS Lovasi61Social and Environmental Health Research, London School of Hygiene & Tropical Medicine, London, UK2Department of Epidemiology, Columbia University Mailman School of Public Health, New York City, USA3Department of Sociology, Center on Health, Risk and Society, American University, Washington, USA4Department of Epidemiology, School of Public Health, Harborview Injury Prevention and Research Center, University of Washington, Seattle, USA5Columbia Population Research Center, New York City, USA6Department of Epidemiology and Biostatistics, Dornsife School of Public Health, Urban Health Collaborative, Drexel University, Philadelphia, USA

Background The growing availability of (non-)commercial historical datasets opens a new avenue of research on how long-term exposure to the neighbourhood environment affects health. However, traditional tools for longitudinal analysis (e.g. mixed models) are limited in their ability to operationalise long-term exposure. This study aims to summarise longitudinal exposure to the neighbourhood using latent class growth analysis (LCGA). Using the National Establishment Time-Series (NETS) 1990-2010, we analysed the trajectory of change in New York City (NYC) in the number of unhealthy food businesses – a potential indicator of an obesogenic environment.Methods The NETS is a commercial dataset providing retail business information in the United States. NYC data were acquired for the period 1990-2010. Businesses were grouped into researcher-defined categories based on Standard Industrial Classification codes and other fields such as business name. All businesses were re-geocoded to ensure accurate localisation. We defined access to BMI-unhealthy businesses (characterised as selling calorie-dense foods such as pizza and pastries) as the total number of BMI-unhealthy businesses present in each NYC census tract (n=2,167) in January of each year. We conducted LCGA in Mplus to identify census tracts with similar trajectories of BMI-unhealthy businesses. We used model fit statistics and interpretability to determine the number of classes. Using the final models, we assigned census tracts to latent classes. We predicted class membership with socio-demographic variables from the Census (population size, income, and ethnic composition) using multinomial logistic regressions and reported predicted probabilities with 95% CI. Sensitivity analyses were undertaken.Results The final models include 5 and 10 latent classes, respectively. The 5-class solution indicates an overall increase in the number of BMI-unhealthy businesses over time and shows a pattern of fanning out: the higher the value in 1990, the greater the increase over time. Classes are associated with 1990 population size, income, proportion of Black residents (all p<0.001), proportion of Hispanic residents (p=0.033), and 1990-2010 change in population size and income (p<0.001). The 10-class solution identifies two pairs of classes with similar 1990 values, but different trajectories. Differences in those trajectories are associated with population size and ethnic composition (p<0.001).Conclusion This study illustrates how LCGA contributes to the understanding of long-term exposure to the obesogenic environment. The technique can easily be applied to other aspects of the neighbourhood and to other geographies. When linked with health data, identified latent classes can be used to assess how longitudinal exposure to changing neighbourhoods affects health.Keywords: neighbourhood, methods, obesity

P28

ASSOCIATION BETWEEN INDOOR TEMPERATURE AND GENERAL HEALTH IN ENGLISH ADULTS: A CROSS-SECTIONAL STUDY

J Sutton-Klein1,2, A Moody1, J Mindell1Institute of Epidemiology & Health, University College London, London, United Kingdom2The Medical School, University of Sheffield, Sheffield, United Kingdom

Background While the association between indoor temperature and health has been indirectly investigated using variables including self-reported cold homes and the presence of insulation as proxies for temperature, research using indoor temperature directly has been lacking. We hypothesised that lower indoor temperatures would be associated with increased odds of poor general health, which would be partially explained by socioeconomic and housing variables.Methods We used data from the Health Survey for England collected between 2003 and 2014.  The analytical sample consisted of 74,735 participants aged over 16. Indoor temperature was approximated using a one-off reading of indoor air temperature in participants’ homes. Self-reported general health was used for the main outcome, dichotomised into very good/good/ fair vs. bad/very bad.Using Stata, logistic regression models were built for the data, using maximum likelihood estimation. The complex survey design and weights were taken into account in the analysis. The influence of socioeconomic and housing factors was assessed by adjusting for these variables in the models. Results The mean indoor temperature was 20.6oC (SE 0.09), which varied seasonally and regionally. Warmer indoor temperatures were associated with greater odds of poor general health. The age and sex-adjusted odds ratio (OR) of poor health for each degree increase in indoor temperature was 1.01 (95% CI 1.01-1.02 P<0.01). Lower social class and lower education level were associated with warmer temperatures and poorer health. The OR decreased after adjusting for these confounding variables. Higher income was associated with warmer temperatures and fairer health. After adjusting for both the suppressor and confounding variables, the OR for poor health for each degree increase in temperature has a statistically significant increase by 19% to 1.02 (95% CI 1.01-1.03 P<0.01).Conclusion Our analysis unexpectedly showed an association between high indoor temperature and poor health.  The relationship between socioeconomic factors and indoor temperature was complicated, with different socioeconomic variables having different directions of association with indoor temperature.  Further research is needed to understand the temporality and direction of the association.  The research is timely given the potential for data from increasingly prevalent smart thermostats.Keywords: temperature health housing

P29

RACIAL INEQUALITIES IN DENTAL SERVICE UTILISATION AMONGST MIDDLE-AGED BRAZILIAN ADULTS

HM Constante1, G Tsakos1, RG Watt1

1[1] Department of Epidemiology and Public Health, University College London, London, United Kingdom

Background Dental service utilisation is important as it helps to alleviate pain and suffering, and access to dental treatment. International evidence has shown that racial/ethnic minority groups are less likely to use dental services compared to the White majority population. In Brazil, where the concept of colour/race is used to classify the population, recent research has assessed colour/race inequalities in dental service use amongst the elderly population, but evidence is scarce for other age groups. Therefore, the overall aim of this research is to investigate the differences between colour/race groups (White versus Pardo and Black) in the use of dental service in a national sample of 35 to 44-year-old adults in Brazil taking into consideration the role of individual-level characteristics.Methods Data from 7,902 adults aged 35 to 44 years from the Brazilian National Oral Health Survey (SB Brazil 2010) was used in this study. The survey collected data using an interviewer-administrated questionnaire and clinical examinations. Dependent variables were the time since last dental visit (less than a year, or a year or more), reason for the last dental visit (prevention/check-up, pain or extraction, or treatment), and type of service used (public, or private). The main exploratory variable was self-reported colour/race (White, Pardo, or Black) measured according to the Brazilian Institute of Geography and Statistics (IBGE). Covariates were sex, level of education, family income, satisfaction with teeth/mouth, self-reported need for treatment, self-reported dental pain in the last six months, presence of decayed teeth, filled teeth, and missing teeth. Binary and multinomial logistic regression analyses were conducted weighted for complex sample design.Results No association between colour/race and time since last dental visit on the fully fitted model was observed. For the comparison between visits due to preventive/check-up and pain or extraction, the association remained significant only for Pardos compared to Whites after the adjustments (RRR 1.63; 95% CI 1.11 – 2.38). In contrast, there was no association between colour/race and visiting the dentist due to treatment. Results from the adjusted model showed statistical differences only for Blacks, and this group was 1.48 times (95%CI 1.10 – 1.97) more likely to visit the public dental service compared to Whites.Conclusion Racial inequalities in dental service utilisation were evident for middle-aged adults in Brazil. For some outcomes, the study showed that colour/race inequalities were not fully explained by the covariates investigated. Further multilevel analysis will be conducted to better understand these inequalities.Keywords: race, dental health services, inequalities

P30

DOES THE QUALITY OF SOCIAL RELATIONSHIPS BUFFER THE ASSOCIATION OF SOCIAL DISADVANTAGE WITH ALLOSTATIC LOAD? AN ANALYSIS OF ADULTS FROM THE UK HOUSEHOLD LONGITUDINAL STUDY

P Rouxel1, T Chandola2, M Benzeval31Social Science, UCL Institute of Education, London, UK2Institute for Social and Economic Research, University of Essex, Colchester, UK3CMIST and Social Statistics, University of Manchester, Manchester, UK

Background Positive support from social relationships is associated with better health, although it remains debatable whether positive social support buffers against the negative effects of social disadvantage on health. Moreover, few studies have considered both positive and negative relationship features from different networks (partner, relatives, and friends) and their association with allostatic load, a multisystem physiological dysregulation index, and none have examined whether the association of social disadvantage with higher levels of allostatic load reduces among those with more positive social support (the stress buffering role of positive social support).Methods This study examined data from 7,928 adults from the UK Household Longitudinal Study (wave 2 - 2010/11). Positive and negative social relationships were assessed within networks (partner, relatives, friends). Allostatic load (range 0-10) was measured by summing eleven risk scores across neuroendocrine, immune, metabolic, cardiovascular and anthropometric systems. We used negative binomial regression models to examine the association of positive social support and negative social interactions with allostatic load, controlling for socioeconomic/demographic, health, behavioural and personality factors.Results Positive support from all networks was associated with lower allostatic load. Similarly, negative interactions from all networks were associated with higher allostatic load. However, after adjusting for all covariates, only positive support from partners and relatives, and negative interactions with friends were associated with allostatic load. Men with degree qualifications and supportive partners had lower allostatic load (2.3; 95%CI: 2.2-2.5) than highly educated men with low partner support (2.7; 95%CI: 2.4-3.0); positive support did not buffer the association of low education with allostatic load. Unemployed adults with supportive partners had lower allostatic load (2.4; 95%CI: 1.9-2.8) compared to unemployed adults with low partner support (3.2; 95%CI: 2.4-4.0).Conclusion Positive social support from partners appears to buffer the effect of unemployment on allostatic load. However, contrary to the stress buffering hypothesis, highly educated men appear to benefit more from supportive partners than men and women with no qualifications. Keywords: Social relationships, allostatic load, socioeconomic status

P31

ACCESS TO PRIMARY CARE FOR SOCIO-ECONOMICALLY DISADVANTAGED OLDER PEOPLE IN RURAL AREAS

JA Ford1, AP Jones1, G Wong2, A Clark1, T Porter1, N Steel11Norwich Medical School, University of East Anglia, Norwich, UK2Nuffield Department of Primary Care, University of Oxford, Oxford, UK

Background Our previous realist review and qualitative research found that socio-economically disadvantaged older people in rural areas face personal, community and health care barriers to obtaining an appointment in primary care. We described these barriers with context-mechanism-outcome (CMO) configurations. Example contexts were lifelong poverty, education, previous healthcare experience, transport, mechanisms were health literacy, assertiveness and convenience, and the outcome was obtaining an appointment.  We aimed to explore these CMO configurations using Structural Equation Modelling (SEM) in a national linked dataset.Methods Individual patient data from Wave 6 of the English Longitudinal Study of Ageing (ELSA) was linked with practice data from the GP Patient Survey (GPPS). Participants in the lowest socio-economic group and living in a rural area were included. Variables from ELSA and GPPS were available for nine complete CMOs. Confirmatory factor analysis was used to generate seven latent variables for unobserved concepts such as lifelong poverty and previous healthcare experience. CMOs were analysed in one overall SEM with multiple mediation paths; contexts were treated as the exposure, mechanisms as the mediator, and the common outcome was ability to obtain an appointment. Stata and MPlus was used to estimate standardised coefficients and 95% confidence intervals using robust maximum likelihood.Results 276 patients from 178 different GP surgeries were included. We found statistically significant direct or indirect effects for two of the nine CMOs. The ease of getting through to the surgery was statistically significantly associated with being able to get an appointment (β 0.52, 0.42 to 0.61) with an indirect, statistically significantly, mediated effect through convenience (β 0.14, 0.07 to 0.21). Health care experience was not directly associated with getting an appointment (β -0.04, -0.19 to 0.12), but a statistically significant indirect effect through convenience existed (β 0.10, 0.04 to 0.16). Model fit showed mixed results (RMSEA 0.05, CFI 0.923, TLI 0.901) Analysis was limited because there were not data for all theoretical concepts. Furthermore, continuous variables were only included to obtain adequate model fit.Conclusion We found that obtaining an appointment was both directly associated with the ease of the booking system, and mediated through the mechanism of perceived convenience. We also found a mediated effect from previous health care experience to obtaining an appointment through perceived convenience.Structural equation modelling proved a useful method for exploring and quantifying realist theory. The analysis was limited by available data; therefore future research would benefit from primary data collection.Keywords: access, primary care, deprivation, ageing

P32

EXPERIENCES OF WOMEN FROM A LOWER SOCIOECONOMIC BACKGROUND WHEN USING HEALTHY EATING MOBILE APPS: A QUALITATIVE INTERVIEW STUDY. 

SJ Flaherty1, M McCarthy1, 2, A Collins3, F McAuliffe1, 4

1HRB Centre for Health and Diet Research, University College Cork, Cork, Ireland2Cork University Business School, University College Cork, Cork, Ireland3Department of Food Business & Development, University College Cork, Cork, Ireland4Obstetrics & Gynaecology, School of Medicine, University College Dublin, Dublin, Ireland

Background Mobile apps offer a potentially effective approach to support healthier food behaviours if adequately designed and informed by behaviour change theory. Individuals from a lower socioeconomic background often report unhealthier dietary patterns and consequently may benefit from a mobile app intervention supporting healthier food behaviours. However, there is limited evidence available on the use of mobile health apps in this group. Previous work suggests that a reasonable standard of health and nutrition literacy is required for effective use of existing healthy eating mobile apps but this knowledge is often low in those from a lower socioeconomic background. Consequently, it is unclear if existing mobile apps are appropriate for this population group. The aim of this study is to explore the experiences of women from a lower socioeconomic background when using healthy eating mobile apps and the individual-level and mobile-specific factors that influence their experiences. Methods A purposive sample of 15 women from a lower socioeconomic background and aged between 18 – 50 years were selected to participate. Participants completed a questionnaire assessing nutrition knowledge before using the assigned mobile apps. A total of three mobile apps were assessed in this study and were of varying quality in relation to nutrition content, behaviour change and user quality. Each participant was assigned to use two different mobile apps and used each for one week only. Assignment order was randomised. After the two-week period, semi-structured interviews were conducted with participants to discuss their experiences. Interviews were audio-recorded, transcribed verbatim, and analysed using a thematic analysis approach.    Results Preliminary analysis suggests that overall mobile app quality is adequate but there is a need to improve the customisability of mobile apps to ensure they fit users’ needs. The food lives of participants vary and mobile apps need to be flexible to reflect this variety for integration of mobile apps into everyday life. The language used in a mobile app was a reason for discontinuing use as it was not clearly understood by users or was viewed as irrelevant. Conclusion Existing mobile apps may support healthier food behaviours in women from a lower socioeconomic background but changes in design may be required. A user-centred approach is recommended where users from a lower socioeconomic background are engaged at all stages of the design process. This may improve their relevance to this population group and increase their effectiveness in supporting healthier food behaviours. Keywords: mobile apps, healthy eating, socioeconomic status

P33

WORKING UPSTREAM: EXAMINING A CENTRAL IDEA IN ADDRESSING HEALTH INEQUALITIES

NE McMahon1, M Gabbay2, J Jagosh3, CL Watkins1

1Faculty of Health and Wellbeing, University of Central Lancashire, Preston, England2Institute of Health and Society, University of Liverpool, Liverpool, England3Centre for Advancement in Realist Evaluation and Synthesis, University of Liverpool, Liverpool, England

Background Health inequalities are variations or differences in health that are systematic, socially produced and unfair. Despite an improved understanding of the causes of health inequalities, there is an overreliance on individual level interventions, often in the form of behaviour change or lifestyle interventions. It is increasingly recognised that in order to reduce health inequalities, there is a need to engage in more ‘upstream’ action. However, as action to reduce health inequalities becomes progressively more interdisciplinary, it is unclear the extent to which there is consensus as to what constitutes ‘upstream’ working to reduce health inequalities. Methods The aim of this study is to examine interdisciplinary conceptualisations of upstream action to reduce health inequalities. Conceptualisations of upstream action have been identified through (i) a review of the literature and (ii) semi-structured interviews with researchers, practitioners and members of the public actively involved in research and practice to reduce health inequalities in the North West of England. This study is being undertaken as part of an MPhil/PhD study. Interviews and data analysis are ongoing.Results There exist multiple different, but related, conceptualisations of upstream action. Examples include upstream action as any intervention focusing on prevention; ‘low agency’ or ‘population’ interventions; action on the social determinants of health; and grass roots or bottom up action that involves shifting the balance of power from individuals making decisions to individuals who are affected by decisions. There is a distinction that can be made between ‘upstream determinants’ of health and health inequalities, and ‘upstream action’, whereby people or interventions located ‘downstream’ can act to influence more upstream determinants of health inequalities. Conclusion There is a need to further unpack, and clarify, what we understand by ‘upstream action’, and the ways in which such actions produce their effects, to assist researchers and practitioners to operationalise and embed these seminal ideas into their day to day practice.Keywords: health inequalities; upstream; qualitative

P34

EXPLAINING DIFFERENCES IN CARDIOVASCULAR DISEASE MORTALITY BETWEEN LOCAL AUTHORITIES IN ENGLAND

P Bhatnagar, N Townsend1Nuffield Department of Population Health, University of Oxford, Oxford

Background Substantial inequalities in age-standardised cardiovascular disease (CVD) mortality rates exist at the local authority (LA) level within England, with particular areas having consistently higher rates. Higher deprivation is associated with higher CVD mortality, but we know little about how the demographics and environments of LAs contribute to variations in mortality rates. Our aim was to explore the extent to which demographic, behavioural and environmental factors explain differences in all ages and premature CVD mortality between LAs in England.Methods All data were sourced for each LA in England. Outcome variables were age-standardised 2012 to 2014 CVD mortality for all ages and those under 75 (premature mortality). Prevalence of ethnic and socioeconomic groups from the UK 2011 census, Public Health England data on index of multiple deprivation (IMD) score, prevalence of smoking, physical activity and obesity/overweight and Ordnance Survey environmental data on percentage of food shops, eating out shops, green/blue space, sporting facilities and health facilities were sourced. We used the Akaike Information Criterion (AIC) to assess which types of variables provided the best statistical model to explain variation in CVD mortality between LAs then used multiple linear regression to assess which variables remained associated with the outcome.Results Including health, demographic, environment and IMD variables provided the best fit for explaining variation in CVD mortality at all ages, with an adjusted R2 of 0.63. For premature CVD mortality, excluding environmental data improved the fit of the model and gave an adjusted R2 of 0.82.The percentage of Indian and Pakistani ethnic groups in LAs remained associated with all ages CVD mortality, along with higher scores for the employment domain and living environment domain of the IMD. For premature mortality, the percentage of Pakistani and Bangladeshi ethnic groups, excess weight prevalence and higher income and crime IMD scores remained associated.Conclusion Certain IMD domains and prevalence of some South Asian ethnic groups are important for explaining variation in age-standardised cardiovascular disease mortality at the LA level in England. These findings are valuable for understanding which factors to target to reduce inequalities in CVD mortality between LAs in England.Keywords: inequalities, cardiovascular disease, environments

P35

ARE PATIENT OUTCOMES IMPROVING? MAJOR AMPUTATION AND DEATH FOLLOWING LOWER LIMB REVASCULARISATION PROCEDURES IN ENGLAND

K Heikkila1,2, DC Mitchell3, IM Loftus4, DA Cromwell1,2

1Health Services Research and Policy, London School of Hygiene and Tropical Medicine, London, UK2Clinical Effectiveness Unit, Royal College of Surgeons of England, London, UK3Southmead Hospital, North Bristol NHS Trust, Bristol, UK4St. George's Vascular Institute, St. George's Vascular Institute, St. George's Healthcare NHS Trust, London, UK

Background Availability and diversity of lower limb revascularisation procedures have increased in in the past decade, co-inciding with the reconfiguration of vascular services in the United Kingdom. The aim of our study was to investigate whether these developments in care have translated to improvements in patient outcomes.Methods We used data from Hospital Episode Statistics (HES) to identify patients who underwent endovascular or surgical (endarterectomy, profundaplasty or bypass) lower limb revascularisation for infrainguinal peripheral arterial disease (PAD) in England in 2006-2013. Major lower limb amputations and deaths were ascertained from HES and Office for National Statistics mortality register. Associations of revascularisation procedures with amputation and death outcomes were investigated using Fine-Gray competing risks regression, with adjustment for patient age, sex and comorbidity score. We examined the possible impact of patient selection by stratifying our analyses by indication for revascularisation (intermittent claudication only; severe limb ischaemia without tissue loss; severe limb ischaemia with ulceration; severe limb ischaemia with gangrene).Results Over the 8-year study period the overall number of endovascular revascularisations increased and the number of surgical procedures decreased: the evidence for this trend was the clearest among patients with the most severe underlying disease (severe limb ischaemia with ulceration or gangrene). The 1-year risk of major amputation reduced from 5.9% (in 2006-07) to 5.5% (in 2012-13) following endovascular evascularization and from 10.8% (2006-07) to 7.4% (2012-13) following surgical procedures (p<0.0001). The risk of death after both types of procedures also decreased, whilst the number of comorbidities and the proportions of patients with more severe underlying disease increased.Discussion Our findings suggest that patient outcomes following lower limb revascularisation have improved during a period of centralisation and specialisation of vascular services in the United Kingdom, despite higher morbidity and an increasing proportion of patients treated for the severe end of the PAD spectrum.Keywords: routine data, revascularisation, amputation

P36

IS IT FEASIBLE TO EVALUATE CARDIAC MRI IN PATIENTS WHO ACTIVATE THE PRIMARY PERCUTANEOUS CORONARY INTERVENTION PATHWAY USING HOSPITAL EPISODE STATISTICS DATA? 

M Pufulete1, J Harris1, S Dorman2, R Brierley1, B Reeves1

1Clinical Trials and Evaluation Unit, University of Bristol, Bristol, UK2Bristol Heart Institute, University Hospitals Bristol NHS Foundation Trust, Bristol, UK

Background Cardiac magnetic resonance (CMR) imaging is a non-invasive test used to assess the structure and function of the heart. We tested the feasibility of assembling a retrospective cohort study of patients who activate the primary percutaneous coronary intervention (PPCI) pathway using data from Hospital Episode Statistics (HES) and Patient Episode Database Wales (PEDW) to: i) document the use of cardiac magnetic resonance (CMR) imaging in this population; and ii) determine whether CMR is associated with improved clinical outcomes. Methods Patients from four UK sites were recruited into the prospective cohort study.  We assembled a database by linking routinely collected hospital data for the index PPCI admission (demography, clinical, biochemistry and imaging) with HES and PEDW describing inpatient and outpatient NHS episodes in the 12 months following the index PPCI admission. We determined whether we could identify the following from HES/PEDW data: i) the index event (cohort entry); ii) CMR within 10 weeks of the index event (exposure); iii) relevant subgroups of the population (e.g. PPCI, unobstructed coronary arteries, multivessel disease, cardiac arrest, etc.); and iv) clinical outcomes.Results A total of 1670 patients were recruited prospectively into the cohort study; of these 1612 (97%) had admission data in HES/PEDW that coincided with the index event (±1 day).  Only 1227/1612 (76%) had HES/PEDW data that met the criteria for cohort entry; 1110 (91%) were identified as PPCI and 117 (9%) were identified as emergency angiography only.  The remainder (385/1612, 24%) did not meet inclusion criteria (PCI procedure code and ST-elevation myocardial infarction, STEMI, diagnostic code) or their angiography was not identified as an emergency. Hospital data documented that 187 patients had CMR but, in the HES/PEDW data (inpatient and outpatient), CMR was recorded for only 53 of these (28%).  We are currently exploring algorithms to identify specific subgroups of particular interest with respect to CMR; early results suggest that we can identify patients who have had cardiac arrest or who have unobstructed arteries but not those with multivessel disease.Conclusion It is possible to identify patients who activate the PPCI pathway from routinely collected HES/PEDW data. We conclude that it is not currently feasible to document the use of CMR in patients who activate the PPCI pathway from HES/PEDW data although the integration of the Diagnostic Imaging Dataset with inpatient and outpatient HES datasets may make it feasible soon. We are exploring the reasons for the discordance between CMR scans reported in the prospective cohort study and those identified in HES.Keywords: Hospital Episode Statistics, cardiac magnetic resonance imaging, primary percutaneous coronary intervention, cohort study

P37

BARRIERS TO AND FACILITATORS OF EFFECTIVE DIABETES SELF-MANAGEMENT AMONG PEOPLE NEWLY DIAGNOSED WITH TYPE 2 DIABETES MELLITUS (T2DM): A QUALITATIVE STUDY FROM MALAYSIA

AM Mohamed1, J Romli2, K Ismail1, K Winkley1

1Department of Psychological Medicine, Institute of Psychiatry, Psychology and Neuroscience, King’s College London, London, UK2Non_Communicable Disease Unit, Primary Care Clinic Tampoi, Johor Bahru, Malaysia

Background To determine the views and experiences of people with Type 2 diabetes (T2DM) in relation to their diabetes self-management and to understand what additional support is required to support lifestyle changes. Methods A qualitative study using semi-structured face-to-face and audio/telephone interviews. All interviews were audio-recorded, transcribed verbatim and analysed using a thematic approach. Seventeen people with newly diagnosed T2DM (less than 3 years of diagnosis) were recruited from a primary care clinic in the southern region of Malaysia (Johor Bahru).Results Qualitative analysis revealed three major barriers to diabetes self-management: (i) psychological issues, e.g. depression and anxiety, such as feeling sad about the diagnosis and worried about the future; (ii) social factors e.g. shame and stigma of T2DM, feeling ashamed have diabetes at a young age and being different from peers ; (iii) perceived barriers e.g. environment and culture, such as ineffective support from healthcare providers, beliefs and use of herbal medicine, and the importance of eating rice and feast culture. Facilitators of diabetes self-management included greater perceived self-efficacy such as being disciplined about eating well, good support from immediate family members and religiosity.Conclusion This study represents novel findings describing barriers and facilitators of effective T2DM self-management in Malaysia. It identifies specific cultural factors that are unique to the Malaysian population that have not been reported in western studies. If diabetes self-management education is to meet people’s needs within this region factors such as these need to be considered when developing new T2DM education programmes.Keywords: Type 2 Diabetes Mellitus, qualitative study, diabetes self-management, barriers

P38

DECISION REGRET IN MEN TREATED FOR LOCALISED PROSTATE CANCER: RESULTS FROM THE LIFE AFTER PROSTATE CANCER DIAGNOSIS STUDY

A Downing1, P Wright1, E Watson2, R Wagland3, L Hounsome4, H Butcher1, A Gavin5, AW Glaser1

1Leeds Institute of Cancer and Pathology, University of Leeds, Leeds, UK2Department of Applied Health and Professional Development, Oxford Brookes University, Oxford, UK3Faculty of Health Sciences, University of Southampton, Southampton, UK4National Cancer Registration & Analysis Service, Public Health England, London, UK5Northern Ireland Cancer Registry, Queens University Belfast, Belfast, UK

Background Men with localised prostate cancer have a number of treatment options. Treatments carry associated benefits and side- and late-effects.  The Life After Prostate Cancer Diagnosis study is a UK-wide survey of men 18-42 months post-diagnosis of prostate cancer. The survey included treatment questions, the Decision Regret Scale (DRS) and a single item on decision-making participation. The aim of this sub-study is to explore the association of decision regret with prostate cancer treatment and patient perception of participation in decision making. Methods The English arm of the survey achieved a 60.3% response rate (30,465 respondents). Men diagnosed with stage 1-3 disease were included in this sub-study (n=16,808).  Descriptive statistics were used to explore associations between DRS scores (0-100), self-reported treatments and perceived participation in decision-making.Results 12,600 (75.0%) men completed the DRS.  Due to the skewed nature of the data, decision regret was categorised as ‘None’ (score=0; 36.5%), ‘Low’ (score=5-20; 31.6%) and ‘High’ (score=≥25; 31.8%).  Levels of regret were lowest in men who underwent brachytherapy alone and surgery alone (25.8% and 27.5% respectively reporting ‘High’ regret).  Men who received combination therapy (e.g. radiotherapy and hormones) reported higher levels of regret compared to men having a single therapy (34.8% vs. 28.3% reporting ‘High’ regret).  28.6% of men on active surveillance reported a high level of regret. 74.2% of men said their views were definitely taken into account in treatment decisions: 22.6% of these reported high regret. 2.8% of men said their views were not taken into account; 62.8% reported high regret.Conclusion These preliminary analyses show a strong association between perceived involvement in treatment decision making and subsequent decision regret in men with prostate cancer.  Levels of regret also vary according to the type of treatment received.  Interestingly, men on active surveillance who receive no treatment report similar levels of regret to those who undergo surgery.  Further analysis will investigate the impact of patient characteristics and functional outcomes (urinary, bowel and sexual) on levels of regret.Keywords: Cancer; survey, decision making

P39

EMERGENCY ADMISSIONS TO HOSPITAL FOR OLDER PEOPLE WITH COPD DURING THE RESHAPING CARE FOR OLDER PEOPLE PROGRAMME (RCOPP): AN ECOLOGICAL STUDY

KA Levin, EM Crighton1Public Health Directorate, NHS GGC, Glasgow, UK

Background The RCOPP was a national programme aimed at enabling older people to stay independent and well at home. Three of the interventions funded by the RCOPP in Glasgow City targeted COPD. This study aims to describe emergency admissions (EAs) for those with COPD in Glasgow City during the RCOPP.Methods COPD EAs were defined using a primary diagnosis of COPD, while EAs for those with COPD were defined using any of the 6 diagnostic fields. Monthly standardised rates of emergency admission between April 2011 and March 2015 were calculated, for residents of Glasgow City aged 65 years+. Multilevel Zero-inflated Negative Binomial models for EAs nested by datazone adjusted for sex, 5-year agegroup, area-level deprivation (SIMD quintile), season, month and month squared. Relevant interventions were entered into the models, to test association with the two outcomes by time and location.Results COPD EAs first rose, from April 2011 until October 2012, then fell until March 2015. When modelled, both month (RR for month 12 relative to month 1 and 95% CI= 1.08 (0.996, 1.17)) and month squared (RR= 0.998 (0.996, 0.999)) were significant, with risk of admission falling below the baseline figures by October 2013 and continuing to reduce thereafter. EAs for COPD patients, however, had the opposite trend, first falling between April 2011 and February 2013, then rising until March 2015, although never reaching the level of April 2011. Under the model, month (RR=0.92 (0.87, 0.97)) and month squared (RR=1.002 (1.001, 1.003)) were both significant. However, this increase was only in the North East and North West sectors of Glasgow. EA for COPD patients reduced in the South, with particularly large reductions from July 2012. When included in the model, the Community Respiratory Team, in the North West, was associated with reductions in COPD EAs (RR=0.89 (0.82, 0.97)) from the point of full staffing) and increases in EAs for COPD patients (RR=1.11 (1.03, 1.20)). Two other projects were also potentially associated with increases in EAs for COPD.Conclusion COPD EAs reduced from the end of 2012, in line with all-cause EAs shown previously. The timing and geography suggest these reductions may be due to the Community Respiratory project which helps people to manage their symptoms at home. EAs for COPD patients, however, increased from around the same time. This or another RCOPP service may have raised awareness of other health concerns or health more generally, increasing EAs for these patients.Keywords: COPD, older people, emergency admissions

P40

EXPERIENCES AND ILLNESS PERCEPTIONS OF WORKING-AGE CARDIAC REHABILITATION ATTENDEES

R Nutt, G Ozakinci1School of Medicine, University of St Andrews, St Andrews, United Kingdom

Background Cardiac rehabilitation (CR) research often focuses on older individuals. With particular pressures faced by working-age individuals (<65 years), such as work and family responsibilities, it is important to consider how this group engages with cardiac rehabilitation. We present here a synthesis of qualitative literature and initial results of a longitudinal qualitative study.Methods We conducted thematic synthesis of studies (2006-2016) describing CR experiences of 18-65’s. Synthesis was gender-sensitive and guided by Leventhal’s Self-Regulation Model of Illness. Results informed development of the qualitative study.The longitudinal qualitative study recruited participants aged 18-65, following MI, from a Phase IV CR programme in Scotland. Semi-structured interviews occur at commencement and completion of the 12-week programme. Participants complete a questionnaire (Illness Perceptions Questionnaire-Revised and health behaviour questions) plus a family member is interviewed at both time points. Questionnaire data contextualises qualitative data (analysed using theoretical thematic analysis).Results Review: Nine studies were included. Heterogeneity existed in CR setting, participant numbers and gender. Thematic synthesis identified themes including illness perceptions, emotional representations and behaviour for illness control i.e. diet change. Some themes appear specific to the ‘working-age’ group. The influence of gender featured across all themes.Study: (At time of submission) Five participants and two family members interviewed, three baseline and follow-up, two baseline only. Participants are male, aged 41-61, all married and employed. Family members are female spouses. All participants had MI. Genetics or bad luck were often seen as cause of their MI, leading to limited behaviour change.  Participants and family valued CR as a place of safety and reassurance due to monitoring and advice provided by staff. Participants identified themselves as fit and active, and therefore not ‘old’, suggesting recovering function/fitness may be particularly important for this age group. Following completion of CR, participants felt they could now exert themselves without causing their body or heart harm.  Participants also described a lack of age-relevant support material that addressed issues like returning to work, family demands such as elderly parents, and exercise advice accounting for their greater pre-MI fitness.  Conclusion It is important to consider how working-age individuals experience CR, as it impacts on engagement and behaviour. Existing literature regarding this group that also includes gender is limited and heterogeneous. Our study thus far indicates male working-age CR attendees and their family value the reassurance monitoring provides and allowing them to regain confidence in their bodies. Keywords: cardiac rehabilitation, working-age, qualitative

P41

MENTAL HEALTH COMPETENCE IN ELEVEN YEAR OLDS AND ITS ASSOCIATION WITH POOR PHYSICAL HEALTH AND MENTAL WELL-BEING: FINDINGS FROM THE UK MILLENNIUM COHORT STUDY

E Rougeaux1, A Pearce1, J Deighton2, C Law1, S Hope1

1Great Ormond Street Institute of Child Health, UCL, London, United Kingdom2Evidence Based Practice Unit, UCL and the Anna Freud Centre, London, United Kingdom

Background Positive mental health (not simply absence of mental disorder) may hold potential for fostering resilience and reducing risks of poor health and well-being. However, it is challenging to define and measure. One promising conceptualisation is mental health competence (MHC), comprising age-relevant developmental tasks and abilities. To date MHC has not been investigated in UK children.   Methods We developed a measure of MHC in the UK Millennium Cohort Study (MCS) (~18000 children born 2000-2002) using seven items addressing learning skills and prosocial behaviours, reported by mothers at age 11 (n=12082). Latent class analysis (LCA) was used to identify classes of MHC, and children were assigned to the class they had the highest probability of belonging to. We investigated associations between MHC classes and child’s physical health (healthy weight, overweight or obese, based on measured heights and weights, using International Obesity Task Force cut-offs; maternal report of unintentional injuries since age 7 [none, 1, 2+]) and poor mental well-being (low self-esteem [SE] and low life-satisfaction [LS] reported by the child; maternal report of child’s emotional problems [EP], all dichotomised at the bottom decile of scores). Relative risk ratios (aRRR, 95%CI) and odds ratios (aORs, 95%CI) were used to examine three-category and binary outcomes respectively, adjusting for potential confounding. Survey weights accounted for sample design and attrition. Analyses were undertaken in Stata/SE 13.1.Results Four classes were identified (ranging from high to low MHC): “High learning skills & high prosocial behaviour” (37%), “Moderate learning skills & high prosocial behaviour” (36%), “Moderate learning skills & moderate prosocial behaviour” (19%), and “Low learning skills & moderate prosocial behaviour” (8%). Risks of 2+ injuries were raised in “Moderate learning & high prosocial behaviour” (aRRR: 1.4[95%CI:1.2-1.7]) and “Low learning skills & moderate prosocial behaviour” (aRRR: 1.4[95%CI:1.1-1.9]) as compared to “High learning & high prosocial behaviour”. Associations with MHC were absent or weak for single injury, overweight and obesity. Compared to “High learning skills & high prosocial behaviour”, odds of poor mental well-being were elevated for children from all other classes, with highest odds for “Low learning skills & moderate prosocial behaviour” (SE: aOR: 2.9[95%CI:2.3-3.6]; EP: aOR: 4.2[95%CI:3.4-5.1]; LS: aOR: 3.0[95%CI:2.4-3.7]). Conclusion Lower MHC, using a composite measure developed in a representative sample of UK children, was associated with injuries, low self-esteem and life-satisfaction, and emotional problems, but not overweight and obesity. Identifying and promoting MHC at the population-level may provide an opportunity to improve health in children and young people.Keywords: child mental health

P42

“UNSEEN INJURIES”: INVISIBILITY AND MENTAL ILLNESS IN THE ENGLISH WELFARE SYSTEM

KJ Pybus, K Pickett, C Lloyd1Department of Health Sciences, University of York, York, England

Background Research focusing on the impact of current welfare reform in England has identified that people with mental illnesses experience pronounced difficulties when accessing social security compared to those with other health conditions. This includes a disproportionate risk of being sanctioned, problematic assessments resulting from a system designed to focus on physical capabilities and a greater likelihood of being viewed as a fraudulent claimant. The impact of individual policies is useful evidence, however the existence of disadvantage across several aspects of this system may imply the presence of underlying mechanisms which need to be understood if further marginalisation is to be prevented. The perspective of individuals with mental illness has been largely omitted from this discussion and may provide useful insights into these topics.  The current study therefore aimed to explore the lived experiences of people with mental illness accessing housing and income benefits in England.Methods Participants were recruited through five organisations offering support with social circumstances to people with mental illness in Leeds, England. Semi-structured interviews were completed with adults (n=11) accessing housing and income benefits, living independently in the community.  Convenience sampling in conjunction with organisation staff was employed to recruit participants. The sample was comprised of participants with the following mental health issues: depression (n=8), anxiety (n=5), post-traumatic stress disorder (n=1), bi-polar affective disorder (n=1), psychosis (n=2), borderline personality disorder (n=1) and substance misuse (n=2) with several instances of co-morbidity. Data was interrogated using a six stage thematic analysis approach.Results Accessing housing and income benefits can be problematic for people with mental illness due to the ‘unseen’ nature of these conditions.  Barriers included being unable to provide evidence of illness to obtain financial support and the level of system flexibility in recognising the impact of mental health needs.  Alongside this, participants encountered stigma from family, friends and the wider community rooted in the perceived validity of mental illness as a reason for claiming benefits.  These themes were present across different types of mental health condition.Conclusion These findings suggest that to create a social security system designed to be effective for people with a mental illness; changes are needed to the way in which mental illness is understood, assessed and monitored in this context.  This study used a diverse but small sample in a localised setting and further research is needed to confirm these findings. Keywords: mental health, health inequalities, social inequalities

P43

THE ASSOCIATION BETWEEN DEPRESSION AND SUBSEQUENT HYPERTENSION - A SYSTEMATIC REVIEW AND META-ANALYSIS

R Prigge1, CA Jackson1

1Usher Institute of Population Health Sciences and Informatics, The University of Edinburgh, Edinburgh, UK

Background The possible causal association between depression and cardiovascular disease might be partly explained by an increased risk of hypertension. Several epidemiological studies have investigated the role of depression in the development of hypertension but the evidence is inconclusive. A previous systematic review of these studies has a number of shortcomings, including inappropriate pooling of different effect measures and lack of inclusion of all relevant studies. Additional primary studies have also been published since this earlier review was completed. Our aim was to identify, critically appraise, and synthesize evidence on the association between depression and subsequent hypertension.Methods We performed a systematic electronic search in PsycINFO, Medline, and EMBASE to identify cohort or longitudinal studies reporting on the risk of hypertension among participants with versus without depressive symptoms and/or clinical depression. We restricted our search to articles published in English. We extracted information on study characteristics, methodology, and results using customized data extraction forms and assessed study quality using the SIGN checklist for cohort studies. We used Stata 14 to perform random effects meta-analyses, to obtain summary effect estimates for the effect of depression on hypertension, pooling hazard ratios and odds ratios separately. We also separately combined studies which defined depression as a categorical or a continuous variable.Results After de-duplication, the search identified 7,402 studies. Twenty-two studies were eligible for inclusion in the review, 17 of which were included in the meta-analyses. Meta-analyses showed an increased hypertension risk among depressed versus non-depressed participants (pooled OR: 1.31, 95%CI: 1.05-1.64; pooled HR: 1.18, 95%CI: 1.02-1.36). Among studies which assessed depressive symptoms on continuous scales meta-analyses indicated an increased risk with every unit increase on the depressive symptoms scale (pooled OR: 1.06, 95%CI: 0.97-1.16; pooled HR: 1.06, 95%CI: 1.01-1.12).Discussion Our review findings provide evidence that depression may be associated with an increased risk of hypertension. However, existing studies have important limitations and the substantial heterogeneity between studies included in two of the four meta-analyses remained unexplained after performing subgroup analyses. Before concluding that depression is indeed associated with an increased risk of hypertension, future prospective studies should improve the accuracy of exposure and outcome assessment, aim to take all major confounding and effect modifying factors into account, and present effect estimates for subgroups in order to help facilitate more meaningful meta-analyses of study findings. Further research is also needed to determine whether the observed association between depression and hypertension is causal.Keywords: Depression, Hypertension, Systematic Review

P44

SELF-HARM, VIOLENCE AND PREMATURE DEATH AMONG YOUNG PERSONS WHO EXPERIENCED TRAUMA-RELATED HOSPITALISATION DURING CHILDHOOD: A NATIONAL REGISTER-BASED COHORT STUDY

RT Webb1, S Antonsen2,3, MJ Carr1, L Appleby1, CB Pedersen2,3, PLH Mok1

1Centre for Mental Health and Safety, The University of Manchester, Manchester UK2Centre for Integrated Register-based Research (CIRRAU), Aarhus University, Aarhus, Denmark 3National Centre for Register-based Research, Aarhus University, Aarhus, Denmark

Background Epidemiological research has reported strong links between trauma-related hospitalisation and future risks of fatal and nonfatal adverse outcomes. However, some important research questions remain unanswered, including association with hospitalisation occurring specifically during childhood, longer-term follow-up from mid-adolescence through the earlier stages of adulthood, assessment of self-harm versus violence risks in the same study population, and absolute risk estimation. To reduce risk in young people, clinicians and public health experts require a better knowledge of subsequent pathways for individuals who experience trauma-related hospitalisation during childhood.Methods This national cohort study examined N=1,087,672 persons born in Denmark 1977-1997 with complete linkage to national psychiatric, general hospital and crime registers. Survival analyses (© SAS Institute Inc.) was used to estimate incidence rate ratios (IRRs) for self-harm, violent criminality, interpersonal violence injury, and all-cause mortality between 15th and 35th birthdays among cohort members with and without trauma-related hospitalisation prior to 15th birthday. Accounting for competing risks, cumulative incidence percentage values were estimated to age 35. Estimates were stratified by gender and by reason for hospitalisation during childhood: self-harm, interpersonal violence or accident.Results Risk for each adverse outcome assessed was raised among young persons who experienced trauma-related hospitalisation at least once during childhood. Confounding by parental socioeconomic status, measured according to income, educational attainment and employment status, explained little of these risk elevations. Individuals hospitalised during childhood following self-harm or interpersonal violence had much higher risks for self-harm and violent criminality aged 15-35 years. Some particularly high cumulative incidence values were observed: subsequent violent offending in males hospitalised following interpersonal violence during childhood, 25.0% (95% CI 21.2-28.9); later self-harm in females hospitalised following interpersonal violence, 18.3% (95% CI 13.5-23.6) and following self-harm during childhood, 21.4% (95% CI 19.8-23.1). More frequent trauma-related hospitalisations, and hospitalisations for multiple trauma types at such an early age, conferred marked risk elevations through young adulthood.Conclusion Although not all episodes of self-harm and interpersonal violence in the community are routinely captured via hospital records, trauma-related hospitalisation during upbringing may be a clinically useful marker for familial dysfunction and childhood distress that subsequently predicts internalised and externalised destructive behaviours among youths and young adults. Comprehensive national guidelines are needed to tackle the multifaceted vulnerabilities of children hospitalised for injuries or poisonings.Healthcare, social services and educational workers must provide particularly robust support to children hospitalised following self-harm or interpersonal violence, and those who experience multiple trauma-related hospitalisations during upbringing.Keywords: childhood adversity; self-harm; violence

P45

SOCIAL AND SPATIAL MOBILITY AND SELF-REPORTED HEALTH IN OLDER-AGE: LINKAGE OF THE SCOTTISH LONGITUDINAL STUDY TO THE 1947 SCOTTISH MENTAL SURVEY

LF Forrest1,2, C Dibben1,2, Z Feng1,2, I Deary3, F Popham4

1Administrative Data Research Centre Scotland, University of Edinburgh, Edinburgh, Scotland2School of Geosciences, University of Edinburgh, Edinburgh, Scotland3Centre for Cognitive Ageing and Cognitive Epidemiology, University of Edinburgh, Edinburgh, Scotland4MRC/CSO Social and Public Health Sciences Unit, University of Glasgow, Glasgow, Scotland

Background The use of administrative datasets to create new cohorts with large sample sizes allows us to answer research questions that we previously could not.Linkage to historic datasets allows exploration of factors that may be important across the life course.There is debate within the literature as to whether social mobility inflates or constrains health inequalities. The role of geographical mobility is unknown. We were interested in exploring how spatial and social mobility might impact on health in older age using linked administrative and cohort data.Methods The 1947 Scottish Mental Survey (a 1936 birth cohort of 70,805 individuals with age 11 cognitive ability test scores) was linked to the Scottish Longitudinal Study (a semi-random sample of 5.3% of the Scottish population), and backward linked to the 1939 register to obtain parental occupation in 1939 (as a measure of social origin) and forward linked to obtain occupation from the 1991 census (social destination), as well as geographical location in 1939 and 1991. We examined the movement between three geographical areas (Edinburgh, Glasgow, Other) in Scotland. Four social mobility trajectories were derived. We modelled the relationship between social and geographic mobility and likelihood of having self-reported limiting long term illness (LLTI) at age 65, using logistic regression.Results Those who were geographically mobile to Edinburgh had the lowest rates of self-reported LLTI and those who remained resident in the Glasgow area had the highest rates. The lowest and highest rates of LLTI were found in the socially-static at the top and bottom of the social scale respectively, with intermediate rates seen in the upwardly and downwardly mobile. However neither social nor spatial mobility were significantly associated with health in later life in the fully adjusted model when highest educational qualifications and cognitive ability were included. Being female, having higher education qualifications and being in a higher social class in childhood and adulthood reduced the likelihood of poor health at age 65.Conclusion Although both social class and geographical location were associated with the likelihood of LLTI in later life, social and spatial mobility were not, when factors such as education and cognitive ability were controlled for.Keywords: social mobility, spatial mobility, health inequalities

P47

ARE OBESITY AND INFLAMMATION FROM MIDLIFE ASSOCIATED WITH PHYSICAL FATIGABILITY IN OLD AGE? FINDINGS FROM A BRITISH BIRTH COHORT STUDY

R Cooper1, M Popham1, AJ Santanasto2, NW Glynn2, D Kuh1

1MRC Unit for Lifelong Health and Ageing, UCL, London, UK2Department of Epidemiology, University of Pittsburgh, Pittsburgh, USA

Background Physical fatigability, the level of fatigue experienced while undertaking specified physical tasks, increases with age throughout adulthood.  These age-related changes, which reflect reductions in energy availability, precipitate declines in activity participation and function.  Higher levels of fatigue are also related to increased risk of disability and premature mortality so it is imperative to identify modifiable risk factors across life associated with physical fatigability.  Cross-sectional analyses suggest that obesity and inflammation may be associated with increased risk of high physical fatigability.  However, whether these inter-related factors act on the same pathway is unclear and requires further investigation in longitudinal studies.  We thus aimed to examine the associations of body mass index (BMI) and inflammatory markers from midlife with subsequent levels of perceived physical fatigability in a nationally representative sample. Methods Up to 2095 men and women from the MRC National Survey of Health and Development, a British cohort followed-up prospectively since birth in 1946, who had valid physical fatigability scores on the Pittsburgh Fatigability Scale (PFS) at age 68 years were included in analyses. Linear regression models were used to assess associations of BMI from age 43 and inflammatory markers (C-reactive protein (CRP) and interleukin-6 (IL-6)) at age 60-64 with continuous PFS scores at age 68.  Adjustments were made for sex (where no evidence of interaction was found), long-term limiting illness, physical activity and symptoms of anxiety and depression.  All analyses were performed using STATA v14.2.Results Women had higher mean PFS scores than men (16.0 (SD=9.3) vs 13.4 (SD=9.0), p<0.01), with higher scores indicating greater perceived physical fatigability.  In sex-adjusted analyses, higher BMI and higher levels of CRP and IL-6 were associated with higher PFS scores.  For example, participants with BMI≥30kg/m2 at age 43 had sex-adjusted mean PFS scores 4.7 (95% CI: 3.3-6.1) points higher than those with BMI 20-24.9kg/m2 and, those in the highest fifth of IL-6 at age 60-64 had a mean score 4.9 (95% CI: 3.5-6.3) points higher than those in the bottom fifth.  When these associations were mutually adjusted and adjusted for other covariates, higher BMI and IL-6 remained associated with higher PFS scores, whereas associations with CRP were fully attenuated. Conclusion These findings highlight the potentially important influence of inflammatory and other cardio-metabolic processes on physical fatigability.  They suggest that both BMI and inflammation from midlife may be important targets for intervention to reduce the burden of this commonly reported symptom in older populations.Keywords: ageing; birth cohorts; life course

P48

COMPARING BMI WITH SKINFOLDS TO ESTIMATE AGE AT ADIPOSITY REBOUND AND ITS ASSOCIATIONS WITH LATER CARDIO-METABOLIC RISK MARKERS

C Di Gravio1, GV Krishnaveni2, R Somashekara2, SR Veena2, K Kumaran2, M Krishna2, SC Karat2, CHD Fall11MRC Lifecourse Epidemiology Unit, University of Southampton, Southampton, UK2Epidemiology Research Unit, CSI Holdsworth Memorial Hospital, Mysore, India

Background Adiposity rebound (AR), defined as the lowest point before the second rise in body mass index (BMI), occurs between the ages of 5 and 7. Early age at AR is associated with higher risk of obesity in later life. However, BMI as a measure of adiposity has limitations: first, BMI incorporates both fat and lean mass, second, BMI is calculated from both height and weight. To identify the AR, a direct measure of fat (i.e. skinfold thickness) might be more relevant. We used data from the Mysore Parthenon Birth Cohort to compare relative merits of BMI and skinfolds in identifying AR and predicting BMI and cardio-metabolic risk factors at 13.5 years.Methods The cohort was set up in 1997 in Mysore, India, to examine the long-term effects of gestational diabetes on cardiovascular disease risk factors in the offspring. Children were followed-up annually until 5 years, and 6-monthly after that for detailed anthropometry. At 13.5 years, 545 children had measurements of cardio-metabolic risk markers. We used non-linear splines and regression analyses (STATA version 14) to characterise the subject-specific growth of BMI and skinfolds (sum of triceps and subscapular skinfolds) throughout childhood, and to assess the associations between age at AR, BMI and cardio-metabolic risk factors.Results BMI and skinfolds had similar trajectories with both reaching their minimum between 5 and 6 years. Average age at AR was similar between the two measures (5.94 years and 5.73 years respectively), with skinfold-derived AR being characterised by higher variability (standard deviation: 1.47 years and 2.18 years respectively). Later age of BMI-derived AR was associated with lower BMI (-0.89kg/m2; 95%CI:[-1.04,-0.74kg/m2]), fat mass (-1.14kg; 95%CI:[-1.36,-0.91kg]), HOMA-IR (-0.12; 95%CI:[-0.17,-0.07kg]) and blood pressure (systolic BP:-0.78; 95%CI:[-1.26,-0.31kg]; diastolic BP: -0.46; 95%CI:[-0.87,-0.04]) at 13.5 years. Similar results were obtained for skinfold-derived AR. Many of the above associations were fully explained by fat mass at 13.5 years. However, the association between skinfold-derived AR and SBP was still significant after adjusting for fat mass (-0.37; 95%CI:[-0.72,-0.01]).Conclusion BMI and skinfolds produced similar estimates of age at adiposity rebound. Associations of AR with BMI and cardio-metabolic risk factors at 13.5 years were comparable regardless of how we derived AR. AR appears to be related to later cardio-metabolic risk markers through its association with fat mass. Skinfolds, as a more direct measure of adiposity than BMI, may be a better method for estimating AR when available.Keywords: obesity, early adolescence, adiposity rebound

P49

AN ECONOMIC EVALUATION OF A COMPLEX WORKPLACE DIETARY INTERVENTION: A CLUSTER CONTROLLED TRIAL

S Fitzgerald1, A Murphy2, A Kriby2, F Geaney1, IJ Perry1

1Department of Epidemiology and Public Health, University College Cork, Cork, Ireland2Department of Economics, Cork University Business School, University College Cork, Cork, Ireland

Background The workplace is recognised as a priority environment to influence dietary behaviours and improve employee healthy. Yet, previous workplace dietary interventions have failed to combine clinical effectiveness evidence with economic costs, thus the cost-effectiveness of workplace dietary interventions remains unknown. Employing cost and outcome data from the Food Choice at Work (FCW) study, a cluster controlled trial of complex workplace dietary interventions, this study employed an economic evaluation of nutrition education, environmental dietary modification and combined workplace interventions.Methods A 9-month time horizon was assumed (length of interventions). Each of the dietary interventions (education, environment and combined) were compared to a control workplace. Firstly, a cost-benefit analysis (CBA) employed the monetary value of absenteeism to report the net benefit of the interventions compared to the control, from an employer’s perspective. Secondly, cost-effectiveness analyses (CEAs) were performed using intervention-specific clinical measures (body mass index (BMI), weight and midway waist circumference) to measure health outcomes. Thirdly, a cost-utility analysis (CUA) measured the cost-effectiveness of the interventions in terms of quality adjusted life years (QALYs). The robustness of the QALYs were assessed as the results of the CEAs and CUA were compared. Probabilistic sensitivity analysis (Monte Carlo simulation) assessed parameter uncertainty.Results The environment intervention reported the highest net benefit (€146/employee) which was associated with an average reduction of 0.7 absent days. The environment intervention also reported the lowest incremental cost-effectiveness ratios (ICERs) for BMI (€14/kg/m2), waist circumference (€3/cm) and weight (€7/kg). The CUA demonstrated similar results as the environment intervention also reported the lowest ICER of €98/QALY, followed by the education (€971/QALY) and combined interventions (€2,156/QALY). The cost-effectiveness acceptability curve (CEAC) indicated that the environment intervention had a 50% probability of being cost-effective when compared to the control at a ceiling ratio of €45,000/QALY. However, as demonstrated on the CEAC, no decision uncertainty surrounded the cost-effectiveness of the education or combined interventions, the control had a higher probability of being cost-effective.Conclusion Although demonstrated over a short timeframe, environmental dietary modification alone, offers a potentially cost-effective approach for improving employee health and generating positive net benefit for employers. While environmental dietary modification strategies are potentially sustainable and important interventions for obesity prevention, future research should include long-term outcomes to determine if improvements in outcomes persist.Keywords: Economic evaluation, workplace dietary interventions, cost-effectiveness

P50

THE PREVALENCE AND DETERMINANTS OF POLYPHARMACY: DATA FROM THE BRITISH 1946 BIRTH COHORT

M J Rawle1, A G Moore1, M Richards1, D Kuh1

1Medical Research Council Centre for Lifelong Health and Ageing, University College London, London, England

Background Polypharmacy, a growing phenomenon within the British population, has been linked with increased falls, reduced functional status and higher all-cause mortality in later life. However the risk profile for individual medications is not universal, with cardiovascular medications in particular posing a high risk. Prior research has found that greater socioeconomic disadvantage is associated with higher levels of polypharmacy but studies rarely control for disease burden or distinguish between cardiological and non-cardiological polypharmacy. The aim of this study was to describe the development of polypharmacy and its composition in a British birth cohort in its seventh decade and to investigate socioeconomic and gender differences independent of disease burden.Methods Medication data from the Medical Research Council National Survey for Health and Development (NSHD), the oldest British birth cohort, were analysed to determine the prevalence and composition of polypharmacy at age 69 and its change from ages 60-64. Multinomial regression was used to test associations between gender, education and occupational social class and total, cardiological and non-cardiological polypharmacy controlling for the number of diagnosed diseases.Results At age 69, 22.8% of individuals were taking more than 5 medications. There was an increase in the use of 5 to 8 medications (+2.3%) and over 9 medications (+0.8%) between ages 60 to 64 and 69. The greatest increases were found for cardiovascular (+13.4%) and gastrointestinal medications (+7.3%). Men experienced greater cardiological polypharmacy, women greater non-cardiological polypharmacy. Higher levels of education were associated with lower levels of both types of polypharmacy independent of disease burden, with strongest effects seen for over five cardiological medications (RRR 0.3, 95% CI 0.2,0.5 p<0.001 for advanced secondary qualifications compared with no qualifications); there was no additional effect of occupational class.Conclusion Polypharmacy, particularly cardiological polypharmacy, increased over the seventh decade and was associated with lower educational attainment. While this study could not assess the appropriateness of the polypharmacy observed, it provided understanding of its genesis and the possible benefits of targeted interventions to reduce potential harm caused by adverse drug events. Further study of the consequences of different types of polypharmacy should take into account these educational differences.Keywords: Polypharmacy Epidemiology Birth-Cohort

P51

LIFE COURSE SOCIOECONOMIC POSITION EFFECTS ON INFLAMMATION IN OLDER ADULTS: COMPENSATING FOR MISSING DATA IN THE ENGLISH LONGITUDINAL STUDY OF AGEING 

G Chatzi, T Chandola, A Cernat, N Shlomo1Social Statistics, University of Manchester, Manchester, United Kingdom

Background Greater cumulative life-course exposure to socioeconomic disadvantage is associated with higher levels of inflammatory biomarkers, C-Reactive protein and fibrinogen which are associated with increased cardiovascular disease risk. However, in most existing studies, researchers often use complete case data for analyses and ignore the impact of missing data on inference for blood based biomarker data. The English Longitudinal Study of Ageing (ELSA) has considerable missing data but the impact of missingness on inference is seldom examined. This paper aims primarily to examine whether the levels of adulthood inflammatory biomarkers of C-Reactive protein and fibrinogen can be explained by life course socioeconomic position. We evaluate the typologies of missing data under Missing Completely at Random and Missing at Random mechanisms and methods for compensating for missing data under these mechanisms.Methods This paper uses cross-sectional data from Wave 2 of ELSA (2004) which includes 9,432 men and women aged over 52 living in England. However, only 6,000 people had data for C-Reactive protein and fibrinogen. Logistic Regression modelling is implemented to identify predictors of missingness in ELSA health examination and blood collection. We use multiple linear regression modelling for analysing the association between socioeconomic position and C-Reactive protein and fibrinogen after accounting for different mechanisms of non-response using the following methods: complete case analysis (listwise deletion), inverse probability weighting and multiple imputations.Results Participants who refused to respond to the health examination were more likely to be renters, White, single, living in London or with poor assessed health. Those who refused to give a blood sample were more likely to be older female, living East of England or London, had cancer/cardiovascular disease/stroke and poorer self-assessed health. Complete case analysis showed that people with lower education level [0.14(CI0.04-0.25) and working in lower supervisory position [0.17(CI0.04-0.30)] were more likely to have higher C-Reactive protein levels. While people in the highest wealth quintile were less likely to have higher C-Reactive protein [-0.26(CI-0.36,-0.14)] and fibrinogen levels [-0.02(CI-0.04,-0.004)]. These associations remained similar in inverse probability weighting and multiple imputations although there was some variation in the estimates from the different methods for compensating for missing data.Conclusion While associations between socioeconomic position and inflammatory markers were similar across different approaches for compensating for missing data, there were differences in the estimated coefficients suggesting that it is important to account for missing biomarker data for statistical inference.Keywords: social inequalities, biomarkers, non-response

P52

EARLY LIFE FACTORS ASSOCIATED WITH LIFE COURSE TRAJECTORIES OF RESTING HEART RATE

CJ O'Hare1, D Kuh1, R Hardy1

1MRC Unit for Lifelong Health and Ageing, University College London, London, UK

Background Elevated Resting Heart Rate (RHR) is an established cardiovascular risk factor and higher RHR even in childhood has been associated with higher adult mortality rates. We are the first to describe trajectories of age-related changes in RHR in the same individuals followed from childhood to later life. We hypothesised that indicators of better health, development and socioeconomic advantage in early childhood would be reflected in lower RHR trajectories, independent of smoking status.Methods Data were from the Medical Research Council National Survey of Health and Development (maximum n=4779) followed from birth in 1946. RHR was measured on eight occasions between 6-69years (three in childhood and five in adulthood). Information on early life factors was ascertained prospectively. Multilevel models were used to estimate life course mean RHR trajectory by sex, and to investigate associations with childhood socioeconomic position (SEP), birth weight (BW), growth from 2-6years (conditional change in Body Mass Index (BMI) and height), duration of breast feeding (BF), and markers of neurodevelopment (age at first walking independently and cognitive ability at eight years).Results Mean estimated RHR decreased with increasing age and plateaued from early adulthood tending to remain higher in women. In sex-adjusted analyses higher BW and conditional BMI change were associated with lower RHR at 6years and across the life course (-0.57bpm (95% confidence interval: -0.95;-0.18) per 1kg higher BW and -0.23bpm (-0.42;-0.05) per 1kg/m2 greater change in BMI); however neither influenced the rate of change with age in RHR. Neither SEP nor BF duration was associated with childhood RHR but associations with RHR trajectory emerged in adulthood such that by 60years RHR was 1.28bpm (0.10;2.16) higher in participants from a disadvantaged versus advantaged SEP and -1.13bpm (-2.01;-0.24) lower for BF >=8months versus never. A later age at walking independently was associated with a higher RHR at 6years (1.33bpm (0.22;2.44) for >=18months versus <12months) but lower RHR in adulthood (-1.09bpm (-2.48;0.30) at 60 years). Childhood cognitive ability was not associated with RHR. In a model including all early life factors, estimates were typically only slightly attenuated from sex-adjusted models except that adjustment for conditional growth substantially reduced the effect of BW. Associations were independent of life course smoking status.Conclusion Early life is a key period in determining future RHR trajectories with effects of potentially modifiable factors persisting into the seventh decade of life.Keywords: Life course epidemiology; cardiovascular disease; early life risk factors

P53

BIRTH AND CHILDHOOD FACTORS AND LATE LIFE CEREBROVASCULAR DISEASE: AN ANALYSIS OF 3 LONGITUDINAL COHORT STUDIES

EV Backhouse1, SD Shenkin2, A McIntosh1, I Deary3, M Bastin1, S de Rooij4, T Roseboom4, JM Wardlaw1

1Centre for Clinical Brain Sciences, University of Edinburgh, Edinburgh, Scotland2Geriatric Medicine, University of Edinburgh, Edinburgh, Scotland3Centre for Cognitive Ageing and Cognitive Epidemiology, University of Edinburgh, Edinburgh, Scotland4Center for Repoductive Medicine, University of Amsterdam, Amsterdam, The Netherlands

Background Cerebral small vessel disease (cSVD) is a major cause of stroke and dementia. Midlife vascular disease and adult socioeconomic status (SES) are established risk factors. Less is known about the effect of factors earlier in life. A recent meta-analysis found that lower levels of childhood IQ, childhood SES and education increased the risk of cSVD in later life but was unclear if these relationships persist after adjustment for vascular risk factors or adult SES.Methods We analysed birth parameters including birth and placental weight (grams), measures of childhood SES such as father’s occupation (manual/non-manual), toilet (outdoors; number of people sharing), childhood IQ and premorbid adult IQ using the National Adult Reading Test (NART) and education (years) from participants from 3 cohort studies: the Dutch famine birth cohort (n=118), the Lothian Birth Cohort 1936 (LBC1936,n=685) and the Simpson cohort (n=110). We analysed cSVD features individually, then summed into a total “cSVD score” (range 1-4) and imaging evidence of infarcts. Data were adjusted for vascular risk factors and adult SES, analysed separately for each cohort and meta-analysed, adjusted for vascular risk factors and adult SES.Results Across the 3 cohorts increasing birth weight was associated with lower cSVD score (OR 0.999 95% CI 0.998-0.999,p=0.02). In the Simpson cohort increasing placental weight was associated lower cSVD score (OR 0.995 95% CI 0.991-0.999,p=0.01), fewer white matter hyperintensities (WMH) (OR 0.995 95% CI 0.99-0.999), cerebral microbleeds (CMBs) (OR 0.995 95% CI 0.991-0.999,p=0.01) and infarcts (OR 0.99 95% CI 0.98-0.998,p=0.02). Higher childhood IQ was associated with fewer lacunes (all studies; OR 0.98 95% CI 0.97-0.99 P=0.04) and infarcts (LBC36; OR 0.99 95% CI 0.97- 0.99,p=0.04). Higher NART score was associated with fewer infarcts (all studies; OR 0.97 95% CI 0.95-0.99,p<0.01). Having an outdoor toilet and more people sharing a toilet in childhood were associated with more infarcts (Simpson cohort; OR 13.31 95% CI 1.52-116.38, p=0.02; OR 1.18 95% CI 1.001-1.4 p=0.049). Lower education was associated with more CMBs (all studies; OR 1.78 95% CI 1.04-3.04,p=0.04).Discussion Birth parameters including birth and placental weight may influence risk of cSVD in later life. Childhood factors such as IQ, education and SES may influence some cSVD features and risk of infarcts but it was not possible in these data to determine whether they contribute independently to WMH or total cSVD or not. The effects sizes and potential impact of these findings suggest that larger samples are needed to robustly test these associations.Keywords: Cerebrovascular disease, epidemiology

P54

CHANGES IN PHYSICAL AND MENTAL HEALTH FUNCTIONING DURING RETIREMENT TRANSITION: A REGISTER-LINKAGE FOLLOW-UP STUDY

M Manty1, A Kouvonen2,3,4, T Lallukka1, 5, J Lahti1, E Lahelma1, O Rahkonen1

1Department of Public Health, University of Helsinki, Helsinki, Finland2Department of Social Research, University of Helsinki, Helsinki, Finland3, SWPS University of Social Sciences and Humanities in Wroclaw, Wroclaw, Poland4Centre for Public Health, Queen’s University Belfast, Belfast, UK5, Finnish Institute of Occupational Health, Helsinki, Finland

Background The impact of retirement on health-related functioning, an important indicator of individual´s ability to function in everyday life, is poorly understood. Thus the aim of this study was to examine the association between transition to retirement, and changes in physical and mental health functioning among Finnish municipal employees.Methods Follow-upsurvey data were collected among ageing employees of the City of Helsinki, Finland, at three Phases: 1 (2000-2002), 2 (2007) and 3 (2012). Physical and mental health functioning were measured using the Short-Form 36 questionnaire at each Phase. Retirees between Phases 1 and 3 were identified from the national registers of the Finnish Centre for Pensions: Full-time statutory retirement (n=1464), part-time retirement not due to health reasons (n=404) and disability retirement due to health reasons (n=462). Generalized estimating equations were used to study the associations between transition to retirement and changes in health functioning. We also examined whether changes in functioning vary by gender, occupational status, or different health- and life-style factors.Results Statutory and part-time retirement were not associated with changes in physical health functioning during retirement transition process when adjusting for gender and age (β 0.1, 95% CI -0.3 to 0.5 and -1.0, -1.8 to 0.1, respectively), whereas clear decline was observed among disability retirees (-4.3, -5.4 to -3.2). Higher occupational class before retirement and sedentary lifestyle were associated with greater decline in physical health functioning. Mental health functioning improved during the retirement transition process among statutory and part-time retirees (1.9, 1.4 to 2.4 and 2.0, 1.0 to 3.0, respectively) while no association was observed among disability retirees (-0.3, -1.7 to 1.0).Conclusion Disability retirement was associated with decrease in physical health functioning, and statutory retirement with a slight improvement in mental health functioning. Evidence on changes in physical and mental functioning during retirement transition process may provide useful information for interventions to promote healthy ageing.Keywords: Retirement; Ageing; Disability; Health; Functioning

P55

PAIN AND ROUTES OF EXIT OUT OF PAID EMPLOYMENT AMONG BRITISH CIVIL SERVANTS: A FOLLOW-UP STUDY 1985-2013

TL Lallukka1, 2, MM Mänty2, CC Cooper3, MF Fleischmann4, AK Kouvonen5, 6, 7, JH Head4, JIH Halonen1

1Finnish Institute of Occupational Health, Helsinki, Finland2Department of Public Health, University of Helsinki, Helsinki, Finland3MRC Lifecourse Epidemiology Unit, Universities of Southampton and Oxford, UK4Institute of Epidemiology and Health, University College London, London, United Kingdom5Department of Social Research, University of Helsinki, Helsinki, Finland6SWPS University of Social Sciences and Humanities in Wroclaw, Wroclaw, Poland7Administrative Data Research Centre - Northern Ireland, Centre for Public Health, Queen's University Belfast, Belfast, UK

Background Pain is a risk factor for work disability, however, routes of exit out of paid employment have not been examined in detail. Thus, the aim of this study was to examine the contribution of repeated exposure to pain and subsequent transitions out of paid employment. A further aim was to examine, whether the associations vary by sex, age, occupational class, job demands, job control and body mass index.Methods We included participants of the Whitehall II study cohort (n=8445, 69% men, aged 35-55 at baseline), who had provided at least 2 measurements of back pain between phases 1 and 3 (1985-1994). People with pain at 1 point in time, and pain at 2-3 time points were compared to people with no pain at any phase. Exit from paid employment was observed between 1995-2013 (phases 4-11). Routes of exit were 1) health-related (long-term sick or retirement on health grounds), 2) unemployment, 3) other early exit, 4) retirement not related to health. Those remaining working served as a reference group. Sex, age, parental and own socioeconomic position, job demands, job control, and body mass index were controlled for. Repeated measures logistic regression models were fitted.Results Altogether 10% exited the employment due to health-related reasons, 2% due to unemployment and 6.5% due to other reasons. Pain contributed to the transitions out of paid employment due to health reasons. After full adjustments, reporting pain at one time point (26%) was not associated with exit due to health reasons, while reporting repeated pain was associated with such exit (18%, OR 1.52, 95% CI 1.16-2.00), as compared to those who did not report pain during phases 1-3 (56%). Associations were somewhat stronger among middle or lower class employees and non-existent among high class employees, but otherwise differences e.g. by age, working conditions or obesity were small. The risk of exit due to other routes than health-related did not vary between people with or without pain.Conclusion These results highlight the need for early detection of repeated pain, to prevent the risk of health-related early exit out of paid employment. As the risk varies between different occupational groups and somewhat by working conditions, this emphasizes the importance of identification of high risk groups and their modifiable risk factors, such as working conditions.Keywords: pain; employment transitions; prospective

P56

DO WORKING CONDITIONS ALTER TO ACCOMMODATE OLDER WORKERS’ CHANGING NEEDS WHEN THEIR HEALTH DECLINES: A 10-YEAR FOLLOW-UP OF THE ENGLISH LONGITUDINAL STUDY OF AGEING

MS Fleischmann1, J Head1

1Epidemiology & Public Health, University College London, London, United Kingdom

Background Due to transformations in pension policies, older workers, even those with poor health, are expected to extend their working lives. In this context, favourable working conditions are frequently discussed as modifiable characteristics that might help to their extended working lives. We investigate whether working conditions actually change if older workers’ physical or mental health is declining.Methods Data on older men and women, aged 50-70 and observed at least twice, from six waves of the English Longitudinal Study of Ageing (ELSA) were used (n=2,958). We observed both the onset of a chronic disease, i.e. diagnose/incident of diabetes, arthritis, stroke, heart problem, asthma, high blood pressure, cancer, or lung disease, and changes in mental health (CESD-R scale for depression), as well as changes in working conditions. Working conditions encompassed physical job demands, decision authority, social support, job security, and working hours. We used fixed effects models to analyse whether changes in older persons’ health could be related to alterations in working conditions.Results Preliminary results show that after adjusting for individuals’ age, the onset of a chronic disease was related to lower decision authority (b= -0.16 [-0.25,-0.07]) and lower social support (b= -0.16 [-0.25,-0.08]). These associations also held after adjusting for individuals’ mental health. Furthermore, a decline in mental health were related to more unfavourable working conditions, specifically job security (b= -0.01 [-0.2,-0.004]), social support (b= -0.05 [-0.07, -0.03]), and decision authority (b= -0.04 [-0.07, -0.3]). The changes in working conditions are rather small but significant. Robustness checks are done to account for possible endogeneity.Conclusion Working conditions do not seem to alter in order to accommodate older workers’ changing needs when their health declines. Rather, older workers with declining mental or physical health report deterioration in their working conditions. Future researchers, policy advisors, and employers might want to consider scrutinizing the group of older workers with declining health and their special workplace requirements when discussing possibilities for extended working lives.Keywords: working conditions, older workers, chronic and mental health

P57

PHYSICAL FITNESS OFFSETS COGNITIVE DYSFUNCTION IN DEMENTIA

D. Cadar1, A. Hollamby1, E.J. Davelaar2

1Department of Behavioural Science and Health, University College London, London, UK2Department of Psychological Sciences, Birkbeck College, University of London, UK

Background Physical fitness has been associated with improved cognition in older age. We explored the associations between physical fitness and cognitive performance in healthy elderly and in individuals with dementia, taking into account the time since dementia diagnosis.Methods Thirty elderly with dementia and 40 healthy controls completed a battery of standardised cognitive tests: Mini-Mental State Exam, Verbal Fluency, Prospective and Retrospective Memory Questionnaire, Clock Drawing and California Verbal Learning Test. Participants were grouped into high versus low levels of physical fitness (PF) based on their Physical Fitness Questionnaire scores.A 2x2 between-subject multivariate analysis of covariance was used to evaluate the associations between PF levels and cognitive performance, adjusting for age, gender, education, occupation, head injury, internet use, brain training, and past levels of exercise. This was followed by a discriminant analysis to obtain a latent score of global cognitive functioning for each individual.Results Healthy controls showed higher performance scores than individuals with dementia [Pillai’s Trace = .61, F(10,49) = 7.75, p < .001]. PF was associated with better cognitive performance [Pillai’s Trace = .32, F(10,49) = 2.29, p < .05] and this interacted with health status [Pillai’s Trace = .30, F(10,49) = 2.12, p < .05]. A significant discriminant analysis (Wilks λ = .15, Chi-square = 119.46, df = 30, canonical correlation = .87, p < .001) extracted two discriminant functions that accounted for 95.4% of the variance, with the first function, reflecting global cognitive function, accounting for 83%. Reclassification using the canonical variables was 72.9% correct. Predicted values from the first discriminant function revealed that the interaction between health status and physical fitness was due to the latter improving global cognitive function in the dementia group [t(25) = 2.90, p < .01], but not in the healthy group (p > .10). A regression analysis howed that although global functioning decreased with the number of years since dementia diagnosis [unstandardised regression coefficients (b) = -.25, 95% CI (-.47, -.04), p < .05], physical fitness was still associated with improved cognitive functioning [b = 1.54, 95% CI (.49, 2.59)], with no interaction (p > .05).Conclusion Physical fitness was associated with improved cognitive functioning, particularly for individuals with dementia. This benefit was present independent of exercise history, age, or duration of dementia. These findings provide empirical support for the development of fitness programmes for dementia patients to offset the cognitive deterioration associated with the condition.Keywords: Physical fitness, cognitive dysfunction, Dementia

P58

BARRIERS AND FACILITATORS TO IMPLEMENTATION OF DIET AND PHYSICAL ACTIVITY INTERVENTIONS IN SCHOOLS-A DEDIPAC (DETERMINANTS OF DIET AND PHYSICAL ACTIVITY) QUALITATIVE STUDY

CB Hayes1, MP O'Shea1, K Horodyska2, A Luszczynska2, LJ Langøien3, G Roos4, S Muellman5, CR Pischke5, I De Bourdeaudhui6, J Brug7

1Public Health and Primary Care, Trinity College Dublin, Dublin, Ireland2Psychology, SWPS University of Social Sciences and Humanities, Wroclaw, Poland 3Physical Education, Norwegian School of Sports Science,, Oslo, Norway4Consumption Research Norway , SIFO, Oslo and Akershus University College of Applied Sciences, Oslo, Norway5Leibniz Institute for Prevention Research and Epidemiology, BIPS, Bremen, Germany6Movement and Sports Sciences, Ghent University, Ghent, Belgium

7Social and Behavioural Sciences, VU University Medical Center (VUmc), Amsterdam, Netherlands

Background This case study was undertaken in Ireland as part of the European DEterminants of DIet and Physical Activity (DEDIPAC) Knowledge Hub. Two national interventions were chosen based on predetermined selection criteria: a Healthy Eating Programme (HEP) to encourage primary schoolchildren to consume more fruit and vegetables, and a Travel to School Programme, (TSP) to promote sustainable modes of transport, car-pooling and public transport use in primary and secondary schools. The HEP is EU and government funded, the TSP entirely government funded. TSP adopts a flexible approach where schools can set their travel targets. School coordinators (teachers) cascade both programmes to classroom teachers.Methods Seven of eight schools invited to participate based on predetermined criteria took part in the study. Face-to-face interviews (n=15) were conducted with teachers, project managers and key stakeholders using a topic guide developed by the international DEDIPAC team and informed by a prior systematic umbrella review of conditions influencing implementation. Data were coded in NVIVO using a common categorization matrix and thematic analysis carried out using parameters of the RE-AIM (Reach, Effectiveness, Adoption, Implementation, Maintenance) evaluation framework. Results Good working relationships were critical to adoption, successful implementation and sustainability in line with findings from case studies in other EU countries. Organisational and leadership ability of coordinators was key to successful delivery. Incentives and rewards acted as motivators to engage children’s interest, which motivated teacher and parent involvement. Particular challenges faced by the TSP included a lack of funding security and timetable constraints within secondary schools. HEP was based on well-funded external research with clearly defined core components and has been frequently externally evaluated. TSP core components were broad rather than specific, implementation was flexible and there was a lack of agreement among stakeholders on how targets were set and the accuracy of these.Conclusion Good relationships, organisational and leadership ability, and secure funding were key conditions for implementation, sustainability and dissemination of promising public health interventions. The findings have informed the DEDIPAC-KH Pan European Toolbox set up for researchers and practitioners who want to develop, evaluate or implement multicomponent interventions on physical activity, sedentary behaviour or dietary behaviour. Keywords: Diet, physical activity, schools, qualitative

P59

LIFELAB SOUTHAMPTON: IMPROVING SCIENCE LITERACY AS A TOOL FOR INCREASING HEALTH LITERACY IN TEENAGERS – A PILOT CLUSTER-RANDOMISED CONTROLLED TRIAL

KS Woods-Townsend1,2, H Aiston3, L Bagust1, H Davey1, D Lovelock1, A Christodoulou1, JB Griffiths1,4, MM Grace1, KM Godfrey2,5,6, MA Hanson2,6 , HM Inskip2,5

1 Southampton Education School, University of Southampton, Southampton, UK2 NIHR Southampton Biomedical Research Centre, University of Southampton and University Hospital Southampton NHS Foundati, Southampton, UK3 Science, St Anne’s Catholic School, Southampton, UK4 Mathematics and Science Learning Centre, University of Southampton, Southampton, UK5 MRC Lifecourse Epidemiology Unit, University of Southampton, Southampton, UK6 Human Development and Health Academic Unit, University of Southampton, Southampton, UK

Background Behavioural risk factors are the largest contributor to the non-communicable disease burden, and those of parents can affect prenatal and infant development with lasting impact on children’s long-term health. Adolescence offers a window of opportunity during which improvements in health behaviours would not only benefit long-term health of individuals, but also enable them to be better prepared for parenthood and pass better health prospects to their children. We have developed an educational intervention, LifeLab, based around a purpose-built laboratory in University Hospital Southampton with support from teachers, to engage adolescents in understanding effects of their health behaviours for themselves and their future children.Aims To assess whether engaging adolescents with the science behind health messages, thus improving their science literacy, increases their health literacy and hence their health behaviours.Methods In a pilot study, in preparation for a large cluster randomised trial of LifeLab, we recruited six schools. Three were randomised to the LifeLab intervention and three to control, with 392 students completing online questionnaires at baseline and 12 months follow up. Summary statistics were used to examine differences between groups. The categorical outcome variables were dichotomised and Poisson regression with robust variance used to obtain prevalence rate ratios (PRRs) for the outcome in relation to the intervention, adjusted for baseline values, sex and Index of Deprivation Affecting Children (IDACI) score.Results 12 months post intervention, intervention students had greater understanding than control students of the influences of health behaviours on their long term health and that of their children. Compared with control students those in the intervention were more likely to agree that nutrition starts to affect our future health early in life (PRR 1.87 (95%CI 1.42-2.45) and that the food a father eats before having a baby could affect the health of his children (PRR 4.05 (95%CI: 2.34-7.01)), but no more likely to agree that it was important to eat healthy food now (PRR 1.19 95%CI: 0.79-1.79)).  The students in the intervention groups took similar amounts of exercise and their diets were comparable to those in the control group.Discussion It is possible to change students’ scientific awareness and health literacy as measured 12 months after the LifeLab intervention, but this does not necessarily translate into behaviour change. Interventions require more than knowledge acquisition in order to motivate and sustain behaviour change.Keywords: Education, Behaviour Change, Health Literacy

P60

HOW CAN MENTORING PROGRAMMES FOR YOUNG PEOPLE IN SECONDARY SCHOOLS IN THE UNITED KINGDOM BE CLASSIFIED? DEVELOPING A TYPOLOGY USING QUALITATIVE METHODS

H Busse1, R Campbell1, R Kipping1

1School of Social and Community Medicine, University of Bristol, Bristol, United Kingdom

Background Despite a lack of robust evidence of effectiveness, mentoring programmes are commonplace in various settings and contexts in the United Kingdom (UK). Due to their potential to influence health and educational outcomes, it is of public health interest to obtain a better understanding of the types of mentoring programmes currently available to comprehend what ‘mentoring’ means and to aid the evaluation of such programmes. The aim of this study was to develop a typology of currently active mentoring programmes that provide formal mentoring for young people in UK secondary schools.Methods Eight websites were searched to retrieve details of UK organisations that provide mentoring programmes for young people. Maximum variation sampling based on country and the type of mentoring programme was used to include a variety of different programmes. Programme managers from purposefully selected organisations were invited to take part in semi-structured telephone interviews to obtain a thorough account of their mentoring programme(s). Interviews were facilitated using a topic guide and were audiotaped and transcribed verbatim. Thematic data analysis occurred iteratively to data collection and was facilitated using NVivo10 software. A framework matrix was established to compare programmes (cases) with categories derived from the analysis (codes) to aid the development of a typology.Results Of 29 invited programme managers, 23 agreed to take part (79% response rate) and described a total of 28 mentoring programmes. The typology drawn from this work differentiates mentoring programmes by three overarching categories: mentoring programmes’ overall aim and target group; type of mentor and mentoring programme setting. These categories each have a range of sub-categories. Based on different combinations of these sub-categories, 12 ‘mentoring models’ were identified within two broad groupings of ‘personal and developmental’ and ‘educational and employability’ mentoring programmes. Conclusion Although mentoring programmes are heterogeneous, it is possible to identify key characteristics and distinguish between different models. Using semi-structured telephone interviews allowed for a thorough investigation of differences between mentoring programmes that was grounded in participants’ accounts of their programmes. The typology enables mentoring programmes to be categorised into one of 12 ‘mentoring models’. A future study is needed to test the typology’s generalisability in the UK. Such a typology can help us to understand what is being delivered, for whom, and how, which is a necessary precursor to any public health evaluation.Keywords: young people, intervention, mentoring programmes

P61

MEDIA REPRESENTATIONS OF SUGAR CONSUMPTION AND SUGAR-SWEETENED BEVERAGE TAX IN UK NEWSPAPERS: IMPLICATIONS FOR PUBLIC HEALTH POLICY

CH Buckton1, L Hyseni2, C Patterson1, SV Katikireddi1, F Lloyd-Williams2, A Elliott-Green2, S Capewell2, S Hilton1

1MRC/CSO Social and Public Health Sciences Unit, University of Glasgow, Glasgow, UK2Department of Public Health Policy, University of Liverpool, Liverpool, UK

Background Modifiable risk factors such as unhealthy diet, including excess sugar consumption, and physical inactivity are common causes of overweight and obesity, which are in turn risk factors for type-2 diabetes and other non-communicable diseases. Sugar-sweetened beverages (SSBs) have been identified as a key target for fiscal policy interventions designed to reduce sugar consumption, particularly in young people. Research shows that the media play a powerful role in forming public perceptions, and thus likely acceptance, of such public health policies. This study assessed how the UK print media presented the debate around the issue of sugar consumption, SSBs and the UK Government’s planned soft drinks industry levy.Methods Quantitative and qualitative content analysis of articles regarding sugar, SSBs and taxation published in a diverse sample of 11 UK national newspapers from 1 April 2015-30 November 2016. Articles were identified by a systematic search of the Nexis database. A coding frame was piloted with a randomised 10% subsample, and revised to include additional emergent codes. Two researchers double-coded the 10% subsample to ensure consistency in the definition and application of codes. Cohen’s kappa coefficient was used to measure inter-rater agreement. All remaining articles were coded by one researcher. Qualitative data were analysed thematically, following the principle of constant comparison, with attention paid to contradictory data.Results The database search returned 3,127 articles, of which 1,495 were manually excluded due to insufficient relevance, producing a final sample of 1,632 relevant articles. None of the articles presented a positive slant on sugar or SSB consumption, whereas representations of SSB taxation were more heterogeneous. The debate initially framed high sugar consumption, particularly SSBs, as problematic, especially for young people. A high proportion of articles framed the problem as being driven by failures of industry, such as the formulation of “unhealthy” products and advertising and marketing aimed at young people. Discussion of potential solutions centred on the role of industrial responsibility, the need for government intervention to curb sugar consumption and the role that taxation could play.Conclusion SSBs received substantial media attention in mainstream UK national newspapers during 2015 and 2016. Public health media advocacy was prominent throughout, with a growing consensus that SSBs are bad for health, government intervention is required and taxation may be an important policy measure. Our findings suggest that the reporting of the SSB policy debate may have helped shape the public health policy agenda on sugar consumption.Keywords: Public health policy, Media content analysis, Sugar tax

P62

THE CHALLENGES OF USING SOCIAL THEORY TO UNDERPIN DIETARY INTERVENTIONS

S Chambers1MRC/CSO Social and Public Health Sciences Unit, University of Glasgow, Glasgow, UK

Background Behavioural approaches to designing dietary interventions can over-emphasise the role of reasoned, individual decision-making, neglecting contextual factors, resulting in sub-optimal interventions. This presentation critically analyses the utility of social theory (structuration theory) in the design of dietary interventions.Methods This study uses the example of designing, implementing and evaluating a school-based dietary intervention underpinned by structuration theory.  Structuration theory describes the interplay of societal structures and individual agency that result in social practices and patterns. Qualitative data collected with children (n=124), parents (n=17) and teachers (n=8) were analysed identifying key structures (meanings and normative rules, and resources) as a framework to understand dietary practices.  Identified rules and resources provided a basis on which to design an intervention.  A process evaluation included interviews with school staff (n=4) and baseline and follow-up data collected from children (n=137).  No feedback was returned from parents. The process evaluation examined adherence, fidelity, and acceptability, and provided an indication of effect.Results Key rules and resources identified in qualitative work were valuing food cooked by family members, cooking inexperience, food misconceptions and rules, lack of food vocabulary, home food provision, school meal and drink provision, curriculum, teacher training, school funding, and national legislation. These findings were translated into an intervention that provided water bottles and water, classroom-based teaching around the curriculum that incorporated the development of cooking skills, a food vocabulary, homework exercises, and teacher training.  Areas that could not be addressed included home food provision, school meal provision, school funding and national legislation. The intervention was acceptable to children and largely acceptable to teachers, but timing pressures meant not all exercises could be covered.  Improvements were reported around children’s drinks.Conclusion Intervention components were largely limited to impacting individual level agency. For university-based research teams, structural components remain difficult to modify. Interventions seeking to impact at multiple levels must work with influential stakeholders working at structural levels, who can impact on long term processes. Social theory can help identify structural and individual level opportunities through which to focus dietary interventions, but the small scale approaches that predominate must be re-thought to increase intervention impact. Keywords: dietary interventions; social theory; process evaluation

P63

DIET@NET: DEVELOPMENT OF THE NUTRITOOLS WEBSITE FOR DIETARY ASSESSMENT

JE Cade1, M Warthon-Medina1, J Hooson1, N Hancock1

1Nutritional Epidemiology Group, School of Food Science and Nutrition, University of Leeds, Leeds, UK

Background Selection of validated dietary assessment tools (DAT) is challenging. Researchers are often unable to identify available tools or gain access to them. Development of new tools is difficult, due to lack of appropriate guidance; inability to identify which foods to include and the need to link these to food composition tables. The DIET@NET (Dietary Assessment tools network) partnership aims to create a dietary assessment website to provide guidelines for selection of tools, with access to previously validated questionnaires. In addition, to create a novel platform for creation of new food questionnaires.Methods Development of the Nutritools website was divided into 3 strands. 1) Creation of Best Practice Guidelines (BPG) for dietary assessment; generated with a Delphi method to generate consensus amongst expert views. 2) Creation of an interactive DAT e-library, with tools being identified through a systematic review of systematic reviews. 3) Creation of an online platform to create new questionnaires.Results The Delphi consultation generated 43 BPG and a summary of the strengths and weaknesses of the dietary assessment methods. The interactive BPG enables researchers to choose the most appropriate dietary assessment tool. The systematic review of systematic reviews resulted in identification of 62 tools validated in UK populations. Detailed information on these tools is provided in the DAT e-library. Visual representation of this data through two interactive plots (bubble and summary plots) allows researchers to compare between the DATs. Only 9 of the tools date from 2010 onwards; many tools identified are no longer available or need updating. Existing validated food questionnaires have been transformed and updated from paper-based to web-based using the novel FQC. Foods are mapped to the latest food composition database. In addition, new online food questionnaires can be created through guided food selection with database mapping. New tools will support online data entry and analysis. At present, the FQC only provides UK dietary assessment tools and databases but it has the capability to allow for international databases and tools to be added.Conclusion The Nutritools website, www.nutritools.org  provides a central resource for researchers undertaking studies which require dietary measurement. By providing guidance and access to validated DATs, the quality, consistency and comparability of dietary assessment in public health and epidemiology can be improved. The Food Questionnaire Creator, is a unique feature of the site which will encourage a more standardised approach to dietary assessment.Keywords: Dietary Assessment Website; Dietary Assessment Tools; Guidelines; Nutritional Epidemiology

P64

COMPARISON OF SOCIOECONOMIC DIFFERENCES IN HEALTH USING OBJECTIVE AND SELF-REPORTED MEASURES OF THE SAME CONDITION: EVIDENCE FROM THE HEALTH SURVEY FOR ENGLAND

JS Mindell1, S Morris2, A Conolly2, S Scholes1

1Epidemiology & Public Health, University College London, London, United Kingdom2Health, National Centre for Social Research, London, United Kingdom

Background Studies quantifying health inequalities typically rely on self-report data. Systematic differences in reporting error (participants denying, minimising, or being unaware of their condition) could result in inaccurate estimates of inequalities. Using Health Survey for England data (2012-2014), we compared the relationship between socioeconomic status (SES) and health using objective and contemporaneous self-reported measures of the same condition.Methods Our study focused on four outcomes: obesity (n=19,993), raised blood pressure (n=14,398), diabetes (n=12,248), and hearing (n=5,334). Obesity (BMI≥30kg/m2) using interviewer-measured height and weight was compared with self-reported height and weight. Blood pressure (BP) measurements (raised BP: systolic/diastolic BP ≥140/90mmHg) were compared with high BP reported as a longlasting illness/condition. Values of glycated haemoglobin (diabetes:≥6.5%) were compared with reported diabetes including hyperglycaemia as a longlasting illness/condition. Using an audiometric screening test, hearing loss was defined as impairment in the better hearing ear to the level of ≥35dBHL at 3kHz; self-reported hearing difficulty was a perceived hearing difficulty, or report of current hearing-aid use. Equivalised household income was our chosen indicator of SES. Logistic regression (using Stata) was used to examine the statistical significance of SES differences. We present the age-adjusted Odds Ratios (AOR) with 95% Confidence Intervals (CI) separately for the objective and self-report measures comparing participants in the lowest income quintile with those in the highest.Results Except for hearing, self-reported estimates were lower than those using objective data. For example, 25.3% (95%CI: 24.2%-26.3%) of men and 24.1% (95%CI: 23.2%-25.1%) of women were obese using interviewer-measured BMI; 18.9% (95%CI: 18.0%-19.9%) of men and 18.1% (95%CI: 17.2%-18.9%) of women were obese based on self-report.Age-adjusted odds of interviewer-measured obesity (AOR: 1.64; 95%CI: 1.44-1.87, p<0.001) and self-reported obesity (1.69; 1.47-1.95, p<0.001) were significantly higher for participants in the lowest income quintile (reference: highest quintile). Findings for diabetes were similar. We found that the magnitude of inequalities for raised BP and for hearing varied by whether self-reported or objective data was used. AORs for each raised BP measure did not attain significance among men. AORs for women were higher for objectively-measured raised BP (1.53; 1.19-1.97, p<0.001) than for self-report (1.37, 0.97-1.94, p=0.075). AORs for low income men were higher using the audiometric screening test (1.81; 1.14-2.89, p=0.012) but not for self-reported hearing problems (0.92, 0.63-1.34, p=0.657). A similar but weaker pattern was found for women.Conclusion Reliance on self-reported data, particularly for asymptomatic conditions such as raised BP, might underestimate socioeconomic differences in health.Keywords: socioeconomic inequalities survey data

P65

SOCIOECONOMIC DIFFERENCES IN HEARING AMONG MIDDLE-AGED AND OLDER ADULTS IN THE HEALTH SURVEY FOR ENGLAND

S Scholes1, J Biddulph1, A Davis2, JS Mindell11Epidemiology & Public Health, University College London, London, UK2UCL Ear Institute, University College London, London, UK

Background Hearing loss impacts on physical and social functioning. Using audiometry data collected in 2014, we estimated the current prevalence of hearing loss in a nationally-representative sample of adults aged 45+ years. We also estimated current hearing aid use among persons with hearing loss.Methods Cross-sectional analysis of the Health Survey for England 2014 (n=3292 participants aged 45+ years with valid screening audiometry data; n=769 with hearing loss). Using Stata, we estimated the prevalence of: (1) hearing loss (defined at ≥35dBHL at 3 kHz in the better hearing ear), and (2) current hearing aid use (among persons with hearing loss). Differences in these outcomes were examined by groups stratified by demography, duration of work-related noise exposure, cardiovascular disease (CVD) risk factors, and socioeconomic status (SES: equivalised household income, education, and area-based Index of Multiple Deprivation). Using sex-specific logistic regression modelling, we evaluated the associations between SES and hearing after adjustment for potential confounders. Results are presented as fully-adjusted Odds Ratios (OR) with 95% Confidence Intervals (95% CIs).Results Hearing loss was higher for men (26%) than women (20%, P<0.001); increased monotonically with age; and was higher for men exposed to work-related noise for ≥5years. SES differences in the odds of hearing loss remained significant in fully-adjusted models for men but not for women. For example, the odds of hearing loss were almost twice as high for men in the lowest versus the highest income tertile (OR: 1.77; 95%CI: 1.15-2.74).Among persons with hearing loss, the prevalence of current hearing aid use was similar for men (30%) and women (27%, P=0.533); increased with the severity of hearing loss; and increased monotonically with age but remained below 40% even for persons aged ≥75 years. Fully-adjusted ORs for current hearing aid use were lower for persons in the lower SES groups among men but not among women, but SES differences did not attain statistical significance. For example, the odds of current hearing aid use for men in the lowest versus the highest income tertile were 0.55 (95% CI: 0.27-1.11).Conclusion Hearing loss is another potential source of socioeconomic inequalities in health, especially in men. Among persons aged ≥45years, more than one in four men and one in five women had hearing loss severe enough to benefit from hearing aid use. However, there is significant unmet need: fewer than one–third of adults with hearing loss currently used a hearing aid.Keywords: socioeconomic inequalities hearing

P66

BARRIERS TO ACCESSING PRIMARY CARE FOR SOCIO-ECONOMICALLY DISADVANTAGED OLDER PEOPLE LIVING IN RURAL AREAS: A QUALITATIVE STUDY

RT Turley1, JF Ford1, TP Porter1, TS Shakespeare1, GW Wong2, AJ Jones1, NS Steel11Faculty of medicine and health sciences, University of East Anglia, Norwich, Norfolk2Nuffield department of primary care health sciences, University of Oxford, Oxford, Oxfordshire

Background More healthcare is needed for an ageing population. Within a health care system that has infinite demands and finite resources there is a need to direct those resources available towards those most in need. Previous research has found that older people, those in rural areas, and socio-economically disadvantaged groups are at particular risk of poor access and often have higher health need. We aim to explore the barriers to accessing primary care for socio-economically disadvantaged older people in rural areas. Methods Semi structured interviews were conducted with fifteen older people over the age of 65 living in a rural area and receiving financial support. Recruitment took place through multiple methods within the community using posters, local radio, hand delivered invitations and invitations dropped in pharmacy bags for home delivery. Participants were purposively sampled for specific characteristics. Four focus groups were held with rural health professionals from primary care, recruited via the Clinical Research Network.Thematic analysis was used to identify barriers to primary care access.Results Multiple barriers were found to impair access to health care of socio-economically disadvantaged older people in rural areas.From the patient’s perspective barriers included attitudes, expectations, experiences and resources. These included engaged telephone lines, availability of appointments, receptionists, transport options, social networks and personal skills. Patients alluded to unwritten rules or a social contract, for example not to bother the doctor in return for additional goodwill when they became unwell. The mismatch between expectations and subsequent experiences led some patients to feel unwelcome or marginalised.From the health professionals’ perspective, barriers included rising demands and expectations, necessitating service reorganisation to overcome significant financial and workforce challenges. Changes included fewer home visits, telephone consultations, triaging calls and appointment system modification. Conclusion Multiple barriers to accessing primary care exist for this group. As primary care is re-organised to reduce costs, commissioners and practitioners must not lose sight of the perceived social contract and models of care that form the basis of how many older people interact with the service.Access to services is a key policy area to address inequalities. One-size-fits-all policies may improve access for the whole population leading to increased pressure on the system and counterproductively worse access for marginalised groups. A targeted approach is needed which appreciates and addresses key barriers. Keywords: Barriers Older primary-care

P67

THE RELATIONSHIP BETWEEN SOCIO-ECONOMIC CLASS AND WORK-RELATED MENTAL ILL-HEALTH

L Hussey1, A Money1, M Gittins2, R Agius1

1Centre for Occupational & Environmental Health, The University of Manchester, Manchester, UK2Centre for Biostatistics, The University of Manchester, Manchester, UK

Background Inequalities in health are often described showing an inverse relationship with social economic group and mortality and morbidity. However, work-related mental ill-health (WRMIH) reported by GPs show highest rates amongst higher socio-economic groups, particularly ‘Lower managerial & professional’ and ‘Intermediate’ occupations. Similar results are shown with the self-reported WRMIH collected by the Self-reported Work-related Illness (SWI) survey as part of the Labour Force Survey (LFS).  The Health & Occupation Research network (THOR) collects data on WRMIH from GPs, occupational physicians (OPs) and psychiatrists. This study aimed to analyse the psychosocial factors associated with the WRMIH reported in different socio-economic groups.Methods With each case of WRMIH, participating physicians report information including occupation and the psychological stressor considered to be associated with the condition. Occupational data were coded using the National Statistics Socio-Economic Classification (NS-SEC). A multi-level logistic regression calculated odds ratios (ORs) (adjusted by age and gender) reporting the likelihood of the stressor occurring within each socio-economic group compared to the highest reference group ‘Large employers & higher managerial occupations’. Results Results showed that the socio-economic groups were exposed to significantly different workplace stressors. Cases reported from the highest group were more likely to be associated with workload and workplace change with ORs falling as socio-economic status decreased, e.g. for workload, ORs in the lowest NS-SEC group were 0.51 (GPs), 0.14 (OPs) and 0.16 (psychiatrists). Lower NS-SEC groups were more likely to have problems associated with bullying and interpersonal relationships, e.g. for bullying, ORs in the lowest group were 2.60 (GPs), 1.76 (OPs) and 1.60 (psychiatrists). Conclusion WRMIH rates may be greatest amongst higher socio-economic groups as they are more likely to have problems associated with a heavy workload and changes in the workplace such as staff shortages and organisational restructure. NS-SEC groups with highest rates such as the ‘middling’ lower managerial and intermediate occupations may find themselves caught in the middle and experience the “worst of both worlds”, i.e. they feel responsibility for the success of the workplace, have demands from those in higher managerial positions, but also have to deal with issues surrounding the maintenance of interpersonal relationships with other workers, clients, patients or pupils. Information from this study may help develop workplace interventions targeted to prevent psychosocial factors affecting different sections of the workforce.Keywords: Health inequalities, work-related mental ill-health, psychological stressors

P68

RELATIONSHIP BETWEEN SOCIOECONOMIC STATUS AND GASTROINTESTINAL INFECTIONS IN DEVELOPED COUNTRIES: A SYSTEMATIC REVIEW AND META-ANALYSIS

NL Adams1,2,3, TC Rose1,2, J Hawker1,3, M Violato1,4, SJ O'Brien1,2, B Barr1,2, VJK Howard2, M Whitehead1,2, R Harris3, DC Taylor-Robinson1,2

1NIHR Health Protection Research Unit in Gastrointestinal Infections, Health Protection Research Unit, Liverpool, UK2Department of Public Health and Policy, University of Liverpool, Liverpool, UK3National Infection Service, Public Health England, London, UK4Health Economics Research Centre, University of Oxford, Oxford, UK

Background The public health impact of gastrointestinal (GI) infections is substantial, with around a quarter of individuals experiencing an episode of infectious gastroenteritis each year. Yet relatively little is known about the social patterning of these infections. Studies investigating the association between socioeconomic status (SES) and risk of GI infections have produced conflicting results, with some reporting greater risk amongst lower SES and other observing the opposite effect. This systematic review and meta-analysis aimed to assess the association between SES and risk of GI infections, and explore possible sources of heterogeneity in effect estimates reported in the literature.Methods MEDLINE, Scopus, Web of Science and grey literature were searched from 1980 to October 2015 for studies reporting a quantitative association between GI infections and SES in a representative population sample from a member-country of the Organisation for Economic Co-operation and Development.  Quality assessment was conducted using the Liverpool University Quality Assessment Tool. Harvest plots were created for comparison where heterogeneity between studies was high, stratified by age, SES measurement, GI infection measurement and pathogen type. Meta-analysis was conducted on a subset of data. To explore sources of heterogeneity, meta-regression and stratified meta-analyses were performed on the basis of country, age, pathogen type, GI infection measurement and SES measurement. The protocol was registered on PROSPERO: CRD42015027231.Results In total, 6021 studies were identified; 102 met the inclusion criteria. Overall risk of GI infection for low versus high SES was 1·06 (95%CI 0·95–1·19). For children, risk was higher for those of low SES versus high (RR 1·51, 95% CI 1·26–1·83), but there was no association for adults (RR 0·83, 95% CI 0·61–1·14). Results were similar when sensitivity analyses were performed on the basis of study quality. Age explained a small proportion of the overall heterogeneity.Discussion We quantify, for the first time, the association between SES and risk of GI infection in developed countries and show that disadvantaged children, but not adults, appear to be at greater risk of GI infection compared to their more advantaged counterparts. Increased risk may relate to differential exposures, vulnerability or healthcare-seeking behaviours by SES. It is possible that factors that could not be adjusted for may explain the high residual heterogeneity. Strategies to improve childhood socioeconomic conditions are likely to reduce the burden of GI illness. Gaining greater insight into this relationship will help to inform policies to reduce the health inequalities identified.Keywords: Gastrointestinal infections; socioeconomic inequalities; systematic reviews

P69

EXAMINING STRATEGIES TO INCREASE KNOWLEDGE MOBILISATION BETWEEN PUBLIC HEALTH ENGLAND AND KEY STAKEHOLDERS: A MIXED METHODS STUDY

R Johnson1, G.J. Melendez-Torres1, G Currie2, P Bradley3, O Oyebode1

1Warwick Medical School, University of Warwick, Coventry, UK2Warwick Business School, University of Warwick, Coventry, UK3Chief Knowledge Officer's Directorate, Public Health England, Cambridge, UK

Background Public Health England (PHE) is an executive agency, sponsored by the Department of Health, which aims to protect and improve the nation’s health and wellbeing, and reduce health inequalities. PHE has a number of responsibilities relating to the collection, curation and sharing of research, data and other knowledge relevant to public health. The organisations key stakeholders include local authorities and clinical commissioning groups. PHE requested that an academic partner support the organisation to develop the organisation’s knowledge mobilisation function.Methods We conducted a sequential mixed methods study. 1. We performed a rapid evidence review to identify strategies which improve knowledge mobilisation; 2. We held a workshop for PHE staff during which we prioritised identified strategies using Delphi methods; 3. We conducted semi-structured interviews and a focus group with a range of PHE staff to discuss the relevance of identified strategies in context; 4. We integrated findings from all three sources using the Pillar Integration Process, a technique for analytical integration of mixed data.Results We identified 13 relevant reviews/meta-reviews. 16 PHE staff attended the workshop. 18 PHE staff were involved in indepth qualitative work (8 semi-structured interviews, 1 focus group of 10 people). Of strategies identified in the literature, workshop participants agreed that some were already working well at PHE, and this was echoed in the qualitative findings. Existing strengths of the organisation are that it is large and trusted, with established local networks. Short term priorities included gaining better understanding of stakeholder needs and future challenges in order make best use of social marketing, tailoring and targeting, and also to reflect narratives that are of particular interest to both immediate and downstream users of PHE evidence (e.g. the Director of Public Health who directly accesses PHE evidence, but also the councillor who ultimately makes decisions based on this). A longer term priority was to develop methods of measuring and evaluating the use of PHE knowledge products. The importance of flexibility in approaches, harmonising rather than homogenising, was a strong theme arising from the qualitative work. Integrated findings highlight the legacy of multiple previous entities having come together to form PHE and the burgeoning identity of PHE as a knowledge brokering organisation.Conclusion We have identified some priority actions for both the short and long-term to improve mobilisation of knowledge from PHE to key stakeholders. This project will inform the development of a knowledge mobilisation strategy at PHE.Keywords: knowledge mobilisation; dissemination, public health

P70

ARE SEPSIS AWARENESS AND ANTIMICROBIAL STEWARDSHIP COMPETING GOALS?A CONTENT ANALYSIS OF THE FRAMING OF SEPSIS AND ANTIMICROBIAL RESISTANCE IN THE POPULAR NEWS MEDIA

LJ Rush1, S Hilton1, LM McDaid1, C Patterson1

1CSO/MRC Social and Public Health Sciences Unit, University of Glasgow, Glasgow, UK

Background Antimicrobial resistance has emerged as one of the greatest threats to population health of modern times.  A review on antimicrobial resistance in 2016 by the UK government predicted that infections caused by resistant organisms could be responsible for ten million deaths annually by 2040.  A key driver of resistance is inappropriate use of antibiotics within human healthcare in managing minor illnesses that would resolve spontaneously without drug treatment.  This has been referred to as ‘just in case’ prescribing. The Chief Medical Officer, Dame Sally Davies, has called for a reduction in unnecessary prescribing as part of improved antimicrobial stewardship however; there is evidence that antimicrobial resistance remains overlooked by the public and by health professionals as an important health risk. Antimicrobial stewardship is complicated by growing recognition that serious bacterial infections can initially present in the same way as mild viral illnesses.  Recently, there has been a drive to improve recognition and management of sepsis, a serious consequence of untreated bacterial infection that is associated with high mortality. Simultaneously, there has been a large volume of high profile stories about sepsis in the media.  This study aims to explore similarities and differences in how antimicrobial resistance and sepsis are framed in the popular print news.Methods Quantitative analysis of the manifest content of 297 articles about sepsis and 163 articles about antimicrobial resistance published in 11 UK national newspapers on all available dates until 31st December 2016, identified via a systematic search on the Nexis database.  A coding frame was developed through a priori knowledge and close reading of 200 articles using the constant comparative method.  A random sample of 10% of these articles were coded by two reviewers to ensure consistency in the application of codes.  The remaining articles were coded by a single reviewer and analysed using SPSS.Results Articles about sepsis were more likely to identify individuals who had suffered serious adverse health effects, often infants or young children, and to criticise the actions of individual health professionals.  These factors have been identified as ‘media triggers’ that can increase public interest in a news story.Conclusion Exposure to news stories about sepsis has the potential to alter public awareness and perception of risks associated with minor illness.  This may impact on expectations of receiving antimicrobial treatment, with implications for antimicrobial stewardship.Keywords: 'antimicrobial resistance' 'sepsis' 'media'

P71

IMPLEMENTATION OF STEP-DOWN INTERMEDIATE CARE IN GLASGOW CITY: A QUALITATIVE STUDY EXPLORING BARRIERS AND ENABLERS

KA Levin, EM Crighton1Public Health Directorate, Glasgow, UK

Background The aim of Step-down Intermediate Care (IC) is to provide a short-term care environment for older people ready for discharge from acute hospital but requiring a period of assessment and/or rehabilitation. There are a number of models of IC in the UK. Glasgow City’s model of IC has gone through several iterations and stages of development. This study aims to examine the implementation of IC in Glasgow City, to identify enabling factors and barriers.Methods The study used multiple qualitative methodologies: document review, semi structured interviews, focus groups and attendance at IC development and care home review meetings. The documentary analysis included reports and meeting minutes. Nine key members of staff were interviewed and three focus groups were run; the first included IC social work staff from Glasgow City's three sectors, the second included rehab staff from two sectors, and the third included care home staff from three IC units. Framework analysis was used to identify common themes. Results Perceived enablers common to all staffing groups included: buy-in from a range of stakeholders, strong leadership and a risk management system in place, good relationship, trust and communication between agencies, the role of targets and in particular the 72 hour discharge target, training of staff, changing perception of risk and risk aversion among practitioners, the right infrastructure and staffing in place, an accommodation based strategy for patients discharged from IC, the right context of political priorities, funding, and a wider model of care for older people, and ongoing adaptation of the model in discussion with frontline staff. Perceived barriers included: differing perception of IC aims, the use of separate technologies by agencies for recording information, transition of patients from hospital to IC, inappropriate referrals to IC, and variation in health and social care systems between sectors and hospitals. Additionally, perceptions differed by staffing group. While social workers noted a need for continuous education of acute staff, and placement issues on discharge from IC, rehab staff found that care homes often lacked appropriate facilities. Both social work and rehab staff noted the benefits of being attached to one unit, while care home staff described the importance of continuity of acute staff.Conclusion The proposed benefits of IC were understood and supported anecdotally by staff. The development of IC in Glasgow gives an insight into enablers of and barriers to implementation of the service, highlighting further opportunities for improvements to the model.Keywords: health services research, older people, intermediate care

P72

UNDERSTANDING WHY UK GPS ARE LEAVING GENERAL PRACTICE - A SYSTEMATIC REVIEW OF QUALITATIVE RESEARCH

L Long1, A Sansom2, A Aylward3, S Robinson1, R Anderson1, E Fletcher2, J Welsman4, S Dean5, J Campbell21Evidence Synthesis & Modelling for Health Improvement (ESMI), University of Exeter Medical School, Exeter, UK2Primary Care Research Group, University of Exeter Medical School, Exeter, UK3Patient and Public Involvement representative, University of Exeter Medical School, Exeter , UK4Centre for Biomedical Modelling and Analysis, University of Exeter, Exeter, UK5Psychology Applied to Health (PAtH), University of Exeter Medical School, Exeter, UK

Background UK GPs are quitting direct patient care. The research presented here is part of a wider mixed methods study (ReGROUP) which focuses on retention of the experienced GP workforce, and on supporting the return to work of GPs following a career break. We undertook a qualitative synthesis of the evidence to identify factors that affect GPs’ retention in the workforce and that affect GPs’ decisions to quit direct patient care, take career breaks from general practice, and return to general practice after a career breakMethods Five UK interview-based studies were identified and quality assessed using the adapted “Wallace tool”. A thematic synthesis was performed using NVivo software and a detailed explanatory model was constructed to provide an overview of the key contexts and factors affecting UK GPs in practice. Applicability of the explanatory model was verified by patient representatives in a patient and public involvement (PPI) workshop as well as by stakeholders from the wider research team.Results Three central dynamics key to understanding UK GP quitting behavior emerged – factors associated with low job satisfaction, high job satisfaction, and those linked to the doctor-patient relationship. The importance of contextual influence, especially the changing nature of clinical practice, was noted. GPs with high job satisfaction describe feeling supported by good practice relationships, while GPs with poor job satisfaction report feeling overworked and unsupported. Combined with changing relationships with patients and interfaces with secondary care, and the gradual sense of loss of control over large parts of the job, many GPs report a reduction in job satisfaction. Once job satisfaction has become negatively impacted, the combined pressures of increased patient demand and workload together with other stress factors has left many feeling unsupported and vulnerable to burnout and ill health, and, ultimately, to the decision to leave general practice.Conclusion Job satisfaction appears to be a key indicator of whether a GP will successfully adapt and remain in practice, or will become overwhelmed by external influences and pressures and leave the profession. Many GPs report that job satisfaction directly relates to the quality of the doctor-patient relationship, and is dependent on the time available for GPs to spend with their patients.Keywords: primary care review

P73

AVOIDABLE EMERGENCY ADMISSIONS IN THE REPUBLIC OF IRELAND: ANALYSIS OF REGIONAL DETERMINANTS

B LynchUCC, AP FitzgeraldUCC, JP BrowneUCC

1Epidemiology and Public Health, UCC, Cork, Ireland2Epidemiology and Public Health, UCC, Cork, Ireland3Epidemiology and Public Health, UCC, Cork, Ireland

Background Many emergency admissions are deemed to be avoidable. There is a potential for several conditions, including angina, diabetes and chronic obstructive pulmonary disease (COPD), to be more appropriately treated in community and urgent care settings.  The rate of emergency admissions for these conditions can be influenced by a number of factors. These include factors directly linked to health services, such as access to general practice, ambulance resources, emergency departments and hospitals. Other factors beyond the control of health services, such as deprivation and rurality, may also play a role. Methods Directly age-standardised emergency admission rates were constructed at a county level for a series of urgent conditions for 2013-2015.Based on methods adopted by a UK study, we examined the role of service level factors on these rates (block 2: availability and accessibility of services) having removed the effect of non-modifiable by service factors (block 1: population, geography and health factors).A multivariate regression using forward stepwise regression within each block was used, beginning with variables from block 1. We then examined the association between the residuals from the first model and block 2 service level variables. Factors were included in the final model if the p-value was <0.05. Results There were 228,435 emergency admissions for selected urgent conditions in 2013-2015. They accounted for 24.5% of all emergency admissions. Results from the first stepwise regression identified one significant non-modifiable by services factor, the regional unemployment rate. This accounted for 30% of the observed variation (R²=30%). Preliminary analysis indicates that factors concerning availability and access to primary care services and hospitals accounted for an additional 47% of the remaining variation (total R²= 63%).  The factors identified as significant were: the percentage of total emergency admissions to hospital with a length of stay of exactly 1 day; the number of general practice nurses per 100,000 population; and, the percentage of general practitioners that had effective access to social workers.Conclusion Having allowed for non-modifiable by service factors, almost half of the variation in emergency admissions for urgent conditions was explained by the availability of primary care and hospital resources. Therefore, increased access to, and use of, community-based primary care may reduce the number of inappropriate admissions.  Keywords: Emergency care, urgent conditions, avoidable admissions

P74

THE IMPACT OF RECONFIGURING ACUTE HOSPITAL SERVICES ON HOSPITAL-TREATED SELF-HARM: A BEFORE-AND-AFTER STUDY

E Griffin1, C Murphy1, IJ Perry2, B Lynch2, E Arensman1,2, P Corcoran2,3

1National Self-Harm Registry Ireland, National Suicide Research Foundation, Cork, Ireland2Department of Epidemiology and Public Health, University College Cork, Cork, Ireland3Department of Obstetrics and Gynaecology, University College Cork, Cork, Ireland

Background Health services in Ireland and the United Kingdom have developed strategies to reduce the number of acute hospitals. This has involved the centralisation of services to centres of excellence along with the reconfiguration of smaller hospitals to urgent care centres – with reduced emergency department (ED) hours. However, the evidence base for improved patient outcomes is limited. We aimed to assess the impact of the reconfiguration of a hospital group in Ireland in terms of the burden of hospital-treated self-harm on each hospital and the clinical management of individual self-harm patients. Methods The study was conducted in three Mid-Western regional hospitals in Ireland. The reconfiguration in April 2009 involved two hospitals (B and C) reducing the operation of their EDs while services at a third hospital (A) remained unchanged. As part of the National Self-Harm Registry Ireland, data were recorded relating to all self-harm presentations during the period January 2004 to April 2014. We used Poisson regression analysis to assess changes in the hospital burden and clinical management of self-harm.Results During the study period there were 9,223 self-harm presentations to the EDs of the three hospitals. Hospital A received the majority (75%), with Hospitals B and C receiving 14% and 11% of presentations, respectively. The reconfiguration was associated with a marked increase in the rate of self-harm presentations at Hospital A, from a rate of 46.0 to 65.2 per month [+19.2 (95% CI 16.2, 22.4)]. This increase was approximately equivalent to the decreases at Hospitals B [-7.7 (-8.4, -6.8)] and C [-9.4 (-9.9, -8.4)]. Despite this large increase in presentations, there was only a small increase in admissions into Hospital A [+3.4 per month (1.4, 5.5)].Conclusion The cumulative decrease in self-harm presentations at the hospitals reconfigured to have reduced ED services was of similar magnitude to the increase at the other hospital’s ED. This suggests that such reconfiguration does not reduce self-harm presentations to hospital EDs but shifts the burden of cases to other hospitals. The disproportionately small impact on hospital admissions suggests that the management of self-harm patients may change in response to increased numbers presenting. This is the first known study to quantify the impact of acute hospital reconfiguration on patterns of self-harm attendances, highlighting changes in the management of self-harm specifically. Patient outcomes following hospital reconfiguration should be an ongoing research priority.Keywords: self-harm, acute service reconfiguration, health services research

P75

A COST-EFFECTIVENESS ANALYSIS OF IMPROVING ACCESS TO PSYCHOLOGICAL THERAPIES IN BRITISH SIGN LANGUAGE (BSL)

GE Shields1, KD Rogers2, A Young2, D Buck1, LM Davies1

1Centre for Health Economics, University of Manchester and MAHSC, Manchester, UK2Division of Nursing, Midwifery and Social Work, University of Manchester and MAHSC, Manchester, UK

Background Deaf BSL users can access standard Improving Access to Psychological Therapies (standard-IAPT) services with/without interpreter assistance in the UK. Some areas use specialist BSL-IAPT services, delivered by qualified IAPT workers who are fluent in BSL, addressing cultural and linguistic needs of Deaf users. This NIHR HS&DR funded study (Ref: 12/136/79) assessed the cost-effectiveness of BSL-IAPT, versus standard-IAPT services, for Deaf BSL users.Methods A combined decision tree (initial IAPT intervention period) and Markov model (longer-term costs and health benefits) used a NHS and social care perspective to estimate the incremental cost/quality-adjusted life-year (QALY) gained and probability that BSL-IAPT is cost-effective. The probability of cost-effectiveness was estimated for willingness to pay thresholds from £0 to £30,000 to gain one QALY. The model structure was developed from published literature and extensive discussion with the research team and advisory group. Both included Deaf people. Data sources for clinical effectiveness, direct costs and utilities included systematic reviews, IAPT databases and new studies with Deaf BSL users. The timeframe was 5 years with costs and QALYs discounted. Probabilistic and deterministic sensitivity analyses assessed parameter and structural uncertainty.Results BSL-IAPT is associated with a net saving of £240 (95th percentiles -£2,303, £935) and little difference in QALYs (net QALY 0.001; 95th percentiles -0.009, 0.013). In both cases the 95th percentiles cross zero, indicating uncertainty about the results. However, cost-effectiveness acceptability analysis indicated probability above 50% that BSL-IAPT is cost effective, even if decision makers are not prepared to pay to gain one QALY. Most sensitivity analyses also indicated that BSL-IAPT was likely to be cost effective. Exceptions were if it was assumed that (i) there was no additional cost for interpreters in Standard-IAPT cost; (ii) BSL-IAPT provides more high intensity care than currently, though the latter is very uncertain due to limited data.Conclusion The primary analysis indicated that BSL-IAPT may be more cost effective than Standard IAPT. Limited data and reliance on a small sample of IAPT data means that the robustness of the results are uncertain. Nevertheless, this is the first economic evaluation of a healthcare for depression and anxiety interventions for Deaf BSL users. It provides initial information to support healthcare commissioners achieve increased access to effective and efficient psychological therapies for depression in Deaf people. The study also highlights the need for new research to prospectively collect robust data.Keywords: Economic evaluation, Deaf, depression

P76

THE DETERMINANTS OF SUBJECTIVE WELLBEING: AN ANALYSIS OF A HEALTH AND WELLBEING SURVEY IN SOUTHEAST ENGLAND

AM Lagnado1, K Gilchrist2, A Memon3

1Undergraduate Medical School, Brighton and Sussex Medical School, Brighton , UK2Directorate of Public Health , Brighton & Hove City Council, Brighton , UK3Epidemiology and Public Health Medicine , Brighton and Sussex Medical School, Brighton , UK

Background The concept of wellbeing is now increasingly used as one of the key measures of societal progress, along with the traditional methods that are based on economic activity. Subjective wellbeing (SWB) is a construct by which national wellbeing can be measured—this can inform development of health and social policy. The objective of this study was to determine the association between sociodemographic/personal factors and low subjective wellbeing.Methods Data from the health and wellbeing survey conducted in Brighton and Hove in 2012 (n = 2,035) was analysed. The survey included the Office of National Statistics verified measure of SWB, which consisted of four questions regarding life satisfaction, fulfilment, happiness and anxiety. Low SWB was the outcome measure, the threshold of which was determined according to the Faculty of Public Health outcome framework. The survey also included a range of population measures, sixteen of which were chosen as explanatory variables. The analysis included descriptive statistics and multivariate logistic regression, using the SPSS statistical programme.  Results In the multivariate analysis, poor general health (self-reported) was strongly associated with low SWB [dissatisfaction with life (adjusted OR=3.9, 95% CI, 2.7-5.6); unfulfilled (3.4, 2.3-4.8); unhappiness (3.0, 2.1-4.2); anxiety (2.4, 1.7-3.3)]. Other factors found to be significantly associated with low SWB included: illness and disability, low social capital, lack of physical exercise, a history of self-harm, not owning a home, not being in a relationship and being middle aged. On the other hand, unemployment, deprivation and poor education were not associated with SWB. Conclusion This study demonstrates that an individuals SWB is likely to be affected by a number of sociodemographic/personal factors. The limitations of this study include the extent of external validity, the lack of causality and potential selection and information bias. These findings are relevant to the design and delivery of policy aimed at improving the perception of wellbeing in individuals, and the general population. Keywords: Determinants, Wellbeing, Quantitative

P77

LEVELS OF PSYCHOLOGICAL DISTRESS AND PREDICTORS OF DISTRESS IN FAMILY CARERS OF PATIENTS WITH CANCER AT END OF LIFE

GE Grande1, C Rowland1, B van den Berg2, E Batistatou3, B Hanratty4

1Division of Nursing, Midwifery & Social Care, University of Manchester, Manchester, UK2Faculty of Economics & Business, University of Groningen, Groningen, Netherlands3Institution of Population Health, University of Manchester, Manchester, UK4Institute of Health & Society, Newcastle University, Newcastle, UK

Background Family carers provide substantial support for patients at end-of-life. It is important to understand how caregiving impacts on carers to guide appropriate interventions to improve carer wellbeing. The aims of this study were to investigate levels of psychological distress and predictors of distress during end of life care giving in a national sample of family carers of people with cancer.Methods Four-month post-bereavement postal survey of a national census sample of relatives reporting a death from cancer 1-16th May 2015. Retrospective data collected included carer demographics, carers’ psychological distress (GHQ-12), care giving hours and tasks, out of pocket expenses, support from informal and formal care, other demands on carers’ time (work, other caregiving responsibilities, voluntary work), opportunities for respite, patient symptoms and activities of daily living (ADL). Exploratory univariate analyses were used to describe the data and inform multivariate analysis. Results Surveys were completed by 1504 (28.5%) of 5271 carers. Carers’ median GHQ distress score was 7 (IQR 4-9), where a score >=4 indicates ‘caseness’ for psychological distress. Univariate analysis results at p< 0.05 indicate that increased hours of caregiving, other caring responsibilities and the patient's worsening symptoms and reduced ADL increased distress. Formal support, hours of volunteering and respite were associated with reduced distress. Carer age, sex, work situation and level of deprivation also related to distress. Multivariate analysis indicates that the total hours of care giving, patients’ psychological symptoms and the carer being female was related to increased distress, whilst formal service provision was related to reduced distress. The final model explained 19% of variance in distress.  Conclusion A considerable majority of family carers suffer clinically significant levels of psychological distress during end of life care giving. Objective care burden in the form of total hours of caregiving is associated with increased distress. Being female and caring for a patient with psychological symptoms appears to increase distress, whereas support from formal care services can ameliorate distress. Whilst the final model explains a modest amount of variance in carer distress, it indicates that reduction in objective care burden and support from services can have an important, positive impact.  The study was funded by Dimbleby Cancer Care; NIHR CLAHRC Greater ManchesterKeywords: family carers, psychological distress, end of life, objective burden

P78

MEASURING MATERNAL MOOD

E Savage-McGlynn1, M Redshaw1, C Martin2

1National Perinatal Epidemiology Unit, University of Oxford, Oxford, UK2Faculty of Society and Health, Buckinghamshire New University, Uxbridge, UK

Background Mood disorders are common among women during the perinatal period, impacting their well-being and that of their families. Prevalence estimates range from 8 to upwards of 40%. Variable detection of cases has been attributed to different measures used, timing of assessment, women not disclosing their feelings, and health professionals not asking questions in meaningful ways. Different measures have been used along the maternity care pathway, with varying reliability and specificity to detect cases and commonly employ graded multiple response format.Listening to what women say about completing such measures,there is a clear need to develop alternative methods of assessing maternal mood in the perinatal period.Objective To develop a measure of maternal mood that provides novel answering and scoring formats, accounts for the spectrum of emotions and symptoms experienced by women in the perinatal period, and correlates with clinical diagnostic measures.Methods A literature review was conducted to assess current understanding of diagnostic criteria for perinatal mental health conditions. Based on previous research with adjective checklists and women’s free text responses to national maternity surveys in 2010 and 2014, scoping of the measure was undertaken. A list of 24 adjectives (12 positive items, 12 negative items) was determined for a prototype measure in which women could choose the adjectives to describe how they had been feeling in the last seven days. Cognitive interviews were conducted with 12 women who had recently given birth, and positive feedback endorsed the content, verified item selection and face validity of the scale.Results The checklist was administered in a survey of maternal and child health to which 551 new mothers responded. Exploratory and Confirmatory factor analyses were conducted to explore underlying factor structure. Two models resulted: a two-factor solution (1. positive mood, 2. negative mood) and a four-factor solution: (1. positive mood, 2. negative mood, agitation; 3.anhedonia, low energy; 4. positive life orientation). Analyses were undertaken for validation and to explore associations with other screening measures to support its use.Conclusion This novel method of reporting feelings and mood in an engaging format will facilitate research in the perinatal field and allow more opportunities for conversations about mood and mental health with health care professionals. As a tool that is psychometrically robust, time-efficient, and which may afford greater insight on the emotional state of the women cared for, the perinatal mood checklist is an effective addition to  measures currently available.Keywords: mental health, maternal, measurement

P79

VALIDATION OF TWO SECONDARY SOURCES OF FOOD ENVIRONMENT DATA AGAINST STREET AUDITS IN ENGLAND

E Wilkins1, M Morris2, D Radley1, C Griffiths1

1Carnegie, Leeds Beckett University, Leeds, UK2Leeds Institute for Data Analytics, School of Medicine, University of Leeds, Leeds, UK

Background Public Health England recently published a tool to help local authorities monitor the density of fast food outlets, with restrictions on access to these ‘unhealthy’ outlets being promoted in the fight against obesity. Secondary data containing the locations of food outlets present valuable resources to guide and evaluate these interventions. However, evidence appraising these data sources is limited. This study therefore seeks to validate two sources of secondary food environment data (SFED): Ordnance Survey Points of Interest (POI) data and food hygiene data from the Food Standards Agency (FSA), against street audits. These data sources are commonly used in research, with the former also being used in the Public Health England tool.Methods Audits were conducted across 54 Lower Super Output Areas in England. All streets within each Lower Super Output Area were covered to identify the name and street address of all food outlets therein. Audit identified outlets were matched to outlets in the SFED to identify true positives (TP: outlets in both the SFED and the audits), false positives (FP: outlets in the SFED only) and false negatives (FN: outlets in the audits only). Agreement was assessed using positive predictive values (PPV: TP/(TP+FP)) and sensitivities (TP/(TP+FN)). Confidence intervals were calculated in Excel using the Agresti-Coull method.Results Overall, the street audits identified 1188 food outlets, compared to 1102 and 1098 for the POI and FSA data respectively. Sensitivity and PPV were significantly higher for FSA data (sensitivity: 0.80, CI: 0.77-0.82; PPV: 0.86, CI: 0.84-0.88) than for the POI data (sensitivity: 0.73, CI: 0.71-0.76; PPV: 0.79, CI: 0.77-0.81). Both datasets had ‘good’ agreement with street audits according to the Paquet classifications (‘good’ defined as PPV and sensitivities between 0.71 and 0.90).Conclusion This study provides new evidence for the validity of SFED commonly used in research and emergently used by policymakers. Agreement between the SFED and street audits is sufficiently good to provide local authorities with confidence in using tools and research based on these SFED. Whilst FSA data has statistically significantly higher agreement with street audits than POI, the magnitude of the difference is relatively small. POI also has other advantages (e.g. more detailed outlet classifications and better spatial accuracy). Thus, POI is still a useful and recommended source of food environment data.Keywords: Foodscape, obesogenic environment, secondary data

P80

DOES SELF-RATED HEALTH MEASURE THE SAME CONCEPT ACROSS COUNTRIES? INSIGHTS FROM A COMPARISON OF OLDER ADULTS IN ENGLAND AND JAPAN

B D Williams1, T Chandola1, N Cable2

1Cathie Marsh Institute for Social Research, University of Manchester, Manchester, UK2Institute of Epidemiology and Health Care, University College London, London, UK12older adults, comparative research, self-rated health Background Self-Rated Health (SRH) is predictive of morbidity and mortality, correlates well with objective measurements of physical function and is simple to use in multidisciplinary surveys. It could be a useful way of comparing health policies in different countries. However, it may not be comparable between countries which may wish to contrast health policies, for example Britain and Japan, because of linguistic, cultural or health differences. We aimed to test for differences in the association between SRH and physical function (grip strength), mental health (depression) and cardiovascular risk (smoking and BMI) between older adults in Japan and England. Methods Data were used from the English Longitudinal Study of Ageing (ELSA; 2004, 2008 and 2012) and the Japanese Study of Aging and Retirement (JSTAR; 2007, 2009 and 2011), giving n=10, 174 ELSA participants and n=4279 JSTAR participants, all aged 50 and above. Multilevel binary logistic regression was used to test whether participants’ country of residence was associated with odds of fair or poor SRH and whether the country of residence would moderate associations between SRH and grip strength, depression, smoking or BMI.Results Japanese women (15.6%) and men (14.1%) were less likely to report fair/poor SRH than English women (23.6%) and men (24.1%). After adjusting for covariates these differences remained for men (Odds Ratio [OR] for Japanese men 0.64, 95% CI 0.74-0.85) but not for women (OR for Japanese women 1.02, 95% CI 0.47-1.34). Grip strength (OR’s women 0.86-0.89; men 0.91-0.94), depression (OR’s women 4.40-6.82; men 5.25-9.26), BMI (OR’s women 1.11-1.16; men 1.10-1.15) and smoking status (OR’s ≥ 20 cigarettes per day women 3.07-7.22; men 2.13-4.54) were associated with fair/poor SRH. No interactions were found between country and grip strength (OR’s women 0.95-1.03; men 0.99-1.05) or depression (OR’s women 0.63-1.39; men 0.50-1.22) but were found for BMI (OR’s women 0.89-0.98; men 0.87-0.97) and smoking (OR’s ≥ 20 cigarettes per day women 0.12-0.34; men 0.20-0.55). The interaction between country and BMI reduced when the analysis was restricted to those with a BMI less than 30 (OR’s women 0.93-1.07; men 0.90-1.05).Conclusion Our findings agree with previous research that SRH captures general physical and mental health similarly across countries. We may need more caution comparing SRH across countries when considering other aspects of health. We find that cardiovascular risk has different associations with SRH in England and Japan, possibly reflecting differences in cultural norms and different stages in their obesity and tobacco epidemics.  Keywords: older adults, comparative research, self-rated health

P81

FORECASTING TRENDS IN DISABILITY IN ENGLAND AND WALES TO 2030: A MODELLING STUDY

M Guzmán-Castillo1, S Ahmadi-Abhari2, P Bandosz1,3, S Capewell1, A Steptoe2, A Singh-Manoux2,4, M Kivamaki2, MJ Shipley2, EJ Brunner2, M O'Flaherty1

1Department of Public Health and Policy, University of Liverpool, Liverpool, United Kingdom2Department Epidemiology and Public Health, University College London, London, United Kingdom3Department of Prevention and Medical Education, Medical University of Gdansk, Gdansk, Poland4Inserm, U1018, Centre for Research in Epidemiology and Population Health, Villejuif, France

Background Reliably estimating the future burden of disability in our ageing societies is crucial. However, previous forecasts failed to consider the potentially significant impact of trends in disease incidence, both up and down. Our aim is thus to forecast the future disability burden in England and Wales to 2030, while taking into account ongoing trends in CVD and dementia incidence. Methods We developed and validated the IMPACT-Better Ageing Model. Using evidence-based age-sex- and year-specific transition probabilities, this probabilistic model tracks the England and Wales population aged 35-100 years through ten health states (notably CVD, cognitive impairment, disability, dementia and death).  We projected continuing declines in dementia incidence, CVD incidence and mortality rates through to 2030 (based on ELSA analysis and consistent with cohorts elsewhere). We then estimated future disability prevalence, distinguishing four types of disability: CVD-related, dementia-related, CVD & dementia-related and Non-CVD/Non-dementia disability. Results England and Wales will continue its transition towards a much older population structure. Between 2015 and 2030, the number of people aged over 65 years will increase by approximately 33.2% (95% uncertainty interval 32.2%–34.1%) while the very old population (over 85) will increase by some 68.3% (63.2%–73.0%) The standardised prevalence of disability among adults aged over 65 years will stay constant at around 21.4% (21.2%–21.6%). However, the number of people living with disability will rise by 40.1% from 2.3 million to 3.2 million. In 2030, Non-CVD/Non-dementia related disability will account for approximately two-thirds of disabled people aged 65-84, (the absolute numbers increasing in approximately 52% from 2015). Dementia-related disability will also increase by some 51% (44.4%−58.8%). Conversely, CVD-related disability will decline by approximately 27% (22.2%−30.0%). Among the very old aged over 85 years, 40% of the cases of disability in 2030 will be attributable to dementia, representing a 113.4% (102.6%−123.6%) increase since 2015. Non-CVD/Non-dementia disability will rise even more, by approximately 159.2% (148.4%−169.1% ). However, CVD-related disability cases, will increase by just 2.8% (-1.4%−7.0%).Conclusion The number of older people with care needs will expand by approximately one third by 2030, mainly reflecting population ageing rather than higher prevalence rates of disability. This will pose a substantial societal challenge and increase the need for cost-effective long-term care in all its forms: institutional, home-based and informal.Future research on the potential benefits of effective prevention strategies might therefore concentrate on the shared determinants of these non-communicable diseases.Keywords: Disability, Dementia, Cardiovascular diseases, Forecast, Modelling study

P82

AN ILLUSTRATION OF THE ANALYTICAL CHALLENGES DUE TO MATHEMATICAL COUPLING IN HEALTH GEOGRAPHY RESEARCH

L Berrie1,2, PD Norman3, PD Baxter1,2, MS Gilthorpe1,2

1Leeds Institute for Data Analytics, University of Leeds, Leeds, UK2Division of Epidemiology and Biostatistics, University of Leeds, Leeds, UK3School of Geography, University of Leeds, Leeds, UK

Background It has been acknowledged that the use of ratio variables is problematic in regression analysis where ratios comprise common components and are present in both the dependent and independent constituents of the model, e.g. regression analyses of ratio variables with common population denominators. However, such ratio variables are ubiquitous in health research and their resultant mathematical coupling (MC) has not been investigated extensively in relation to studies in health geography, where common population denominators are frequently encountered. It is common, for instance, that area level measures for health outcomes are considered in relation to area levels of mortality and/or indicators of social deprivation, where the common denominator is the area population. Our study seeks to illustrate this issue and examines the implications of this form of MC from a causal inference perspective.Methods We examine the impact of MC amongst ratio variables in regression analyses using simulated data based on the correlation structure and distribution of variables derived from the UK census. Specifically, we consider the proportion of limiting long term illness (LLTI) in relation to mortality rates and the Townsend Material Deprivation Score (constructed from percentages of the population of an area that are experiencing pre-defined properties). Simulations are constructed under the null hypothesis, i.e. there is no impact of an area measure of deprivation on the relationship between mortality and proportion of limiting long term illness. A causal framework is introduced utilising directed acyclic graphs (DAGs) to assess variable relationships and to suggest analytical strategies that mitigate any problems arising due to MC.Results We show that artefactual relationships arise in the regression analyses of composite proportions due to MC: area measures of deprivation appear to influence the relationship between LLTI and mortality when no such influence is present in the simulated data. A DAG aids comprehension of this issue from a causal inference perspective and, depending upon the exact nature of the MC present, the DAG can also point to alternative analytical strategies that are discussed.ConclusionMathematical coupling of ratio variables has been recognised and reported on in the past, yet its problems remain pervasive. By setting the problem within a causal framework, we provide a means by which the issue might be more readily identified. Furthermore, using DAGs can help direct alternative analytical strategies to remove bias due to MC from future health research. Keywords: mathematical coupling, ratio variables, health geography

P83

EXPLAINING THE FALL IN CORONARY MORTALITY IN ARGENTINA BETWEEN 1995 AND 2010: IMPACT MODELLING ANALYSIS

J Vicens1, G Perman1, P Bandosz2, M Guzman2, D Ferrante3, H Schargrodsky1, V Aliperti1, P Pramparo, M O'Flaherty2, S Figar1

1Department of Internal Medicine and Cardiology, Hospital Italiano de Buenos Aires, Buenos Aires, Argentina2Public Health and Policy, University of Liverpool, Liverpool, UK3Ministry of Health, City of Buenos Aires Government, Buenos Aires, Argentina

Background Coronary heart disease (CHD) mortality has declined substantially in Argentina in the last decades. However, the contributory factors remain unclear because major CHD risk factor levels have paradoxically increased, notably physical inactivity, obesity and diabetes.our objectiv is to quantify the contributions of prevention and treatment to the coronary heart disease mortality trends in Argentina between 1995 and 2010.Methods We used the previously validated IMPACT model to analyse mortality trends between 1995 and 2010 in the Argentinean population aged 25 years and over. This model integrates data on changes in population size, CHD mortality, risk factors, and uptake of evidence-based cardiac treatments. Main data sources included official vital statistics, national health surveys, and published papers in the scientific literature. If none of these were available, we used expert judgement. Relative risks and regression coefficients came from the published meta-analyses. The difference between observed and expected CHD deaths in 2010 was then partitioned among treatment benefits and risk factor changes. We also performed probabilistic sensitivity analyses to quantify the potential effects of parameter uncertainty.Results From 1995 to 2010, age-adjusted CHD mortality rates in Argentina fell by 24.1%, resulting in 8,500 fewer CHD deaths in 2010 had mortality rates remained unchanged. Improvements in medical and surgical treatments were associated with 53.8% (95% uncertainty interval, 46.2% - 61.8%) of the total mortality decrease. Major contributions came from heart failure therapies (15.1%), antihypertensive medication (11.9%), and secondary prevention following acute myocardial infraction (8.2%).Risk factor changes accounted for 40.4% (9.5% to 65.6%) of the mortality fall. Decreases in systolic blood pressure accounted for 34.6%; in cholesterol levels, 19.9%; and in smoking prevalence, 6.7%. However, rises in diabetes, BMI and physical inactivity negated some of these benefits, potentially increasing mortality by 9.4%, 6.4% and 5.0%, respectively. Approximately, 6% of the deaths prevented or postponed could not be explained in our model.Results Approximately 54% of the CHD mortality fall in Argentina between 1995 and 2010 was attributable to increases in evidence-based medical treatments, and 40% to falls in population risk factors. However, the substantial contributions from falls in blood pressure, cholesterol and smoking were offset by adverse trends in diabetes, obesity and physical inactivity. Our results highlight the potential for further improvements in cardiovascular Keywords: Coronary heart disease prevention modelling

P84

DEFINING A PRIMARY COMPOSITE OUTCOME FROM HOSPITAL EPISODE STATISTICS DATA TO TEST THE BENEFIT OF CARDIAC MAGNETIC RESONANCE (CMR) IMAGING AFTER PRIMARY PERCUTANEOUS CORONARY INTERVENTION (PPCI)

M Pufulete1, J Harris1, S Dorman2, R Brierley1, B Reeves1

1Clinical Trials and Evaluation Unit, University of Bristol, UK2Bristol Heart Institute, University Hospitals Bristol NHS Foundation Trust, UK

Background We researched the feasibility of using routinely collected data to establish a registry to document the use of cardiac magnetic resonance imaging (CMR) in patients with suspected heart attack who activate the primary percutaneous coronary intervention (PPCI) pathway.  To compare outcomes between patients having CMR or not, we aimed to define a primary composite outcome, representing clinically important changes in management resulting from CMR, using data from Hospital Episode Statistics (HES) or the Patient Episode Database Wales (PEDW).  Clinically important changes in management, and patient subgroups to which these related, were defined by formal consensus.Methods Patients at four UK sites were prospectively consented.  We assembled a database by linking hospital-collected data about the index PPCI admission (demography, clinical, biochemistry and imaging) and CMR ‘exposure’ within 10 weeks with HES/PEDW data for the following 12 months.  An experienced medical coder identified ICD10 diagnostic codes and OPCS procedure codes that the pre-defined clinically important changes in management achieved by CMR were hypothesised to influence.  Episodes in HES/PEDW inpatient, outpatient and accident and emergency datasets with these codes were identified.  This process was applied to key patient subgroups: i) PPCI; ii) multivessel coronary disease; iii) out-of-hospital cardiac arrest; iv) coronary angiogram showing unobstructed arteries.Results Of the 1670 patients consented, 1476 (88%) had PPCI; 682 (41%) had multivessel disease, and 194 (12%) had unobstructed arteries.   189 (11%) had CMR within 10 weeks of the index admission.  1612 (97%) had their index event identified in HES/PEDW and 985 of these (61%) had a full year of follow-up available.  Focusing on important changes by patient subgroup avoids creating a composite outcome based on diverse changes.  Frequencies of relevant health episodes will be reported for patient subgroups and by CMR exposure.  Comparing frequencies is straightforward but identifying better targeting of care to patients’ needs without a change in frequency, e.g. due to better diagnostic accuracy of CMR, is complex. Discussion Clinical events relating to important changes in management resulting from CMR can be identified in HES but CMR may not alter the frequency of these events. This case study demonstrates the promise of using routinely collected data to evaluate changes in practice. However, it is challenging to identify subtle changes in management such as better targeting of care without changing its frequency.  In future, imaging investigations and changes in management should be identified more accurately by linking to the Diagnostic Imaging Dataset and prescribed medications. Keywords: Cohort study, cardiac magnetic resonance imaging, primary percutaneous coronary intervention, defining a primary composite outcome

P85

EXCLUSION OF COMMUNITY-LED INITIATIVES BY PUBLICATION BIAS: EVIDENCE FROM A SYSTEMATIC SCOPING REVIEW OF COMMUNITY ENGAGEMENT IN THE UK

AM Bagnall1, J South1, J Trigwell1, K Kinsella1

1Centre for Health Promotion Research, Leeds Beckett University, Leeds, UK

Background A recent systematic review found that initiatives with high levels of community engagement may produce more positive health outcomes than those with lower engagement. Systematic reviews in this area risk publication bias because (i) literature on community based health initiatives is widely dispersed and poorly indexed; and (ii) professionally-led (top down) interventions are more likely than community-led (bottom up) interventions to be formally evaluated and published.An opportunity to examine the gap between research and practice arose in a systematic scoping review commissioned by the UK National Institute of Health & Care Excellence (NICE), of current practice in community engagement.Methods We searched specialised research registers and websites; literature searches and citations from recent relevant systematic reviews; and direct calls for evidence via networks of community practitioners and groups. Records were screened independently by more than one reviewer, and included if published after 2000, relevant to the UK and evaluated or described community engagement in public health. Included records were coded for type, level and extent of community engagement, indicators of disadvantage, issues and outcomes.Results 316 articles were included, 72% were research or evaluation studies (7% were randomised controlled trials; most were mixed methods or qualitative studies).  26% were found through website searches and the calls for evidence. The issues addressed most frequently were social capital or social cohesion (41%) and community wellbeing (35%). Indicators of health inequality observed most frequently were socioeconomic (39%) and “other” (39%), including people with disabilities; refugees and asylum seekers; mental health service users. Only 33 initiatives reported a high extent of community engagement; a comparatively high proportion were in the non-research literature (20% of non-research articles, compared to 8% of research articles). This may indicate a gap between organisations which usually write and publish research articles, and organisations which fully involve community members, and/or may indicate challenges in the evaluation or publication process of high community engagement initiatives.Conclusion Using conventional systematic review methods to examine community based approaches risks overlooking community-led “bottom up” initiatives, which may have the highest potential to reduce health inequalities. Reviewers should therefore make every effort to find reports of such initiatives, and consider broadening their definition of “evidence”.Keywords: systematic review methods; community engagement; health inequalities

P86

ARE PROCESS EVALUATION MEASURES RELATED TO INTERVENTION OUTCOMES IN THE PACE-UP PRIMARY CARE PEDOMETER-BASED WALKING TRIAL?

C Furness1, E Howard1, E Limb1, D G Cook1, S Kerry2, C Wahlich1, C Victor3, U Ekeland4, S Iiffe5, M Ussher1, M Ussher1, P Whincup1, J Fox-Rushby3, J Ibison1, S DeWilde1, T Harris1

1Population Health Research Institute, St. George's University of London, London, England2Pragmatic Clinical Trial Unit, Queen Mary's University of London, London, England3Gerontology and Health Services Research Unit, Brunel University, London, England4Department of Sports Medicine, Norwegian School of Sports Science, Oslo, Norway5Research Department of Primary Care and Population Health, University College London, London, England

Background PACE-UP trial results demonstrated positive effects of a pedometer-based walking intervention on objective physical activity (PA) outcomes at 3 and 12 months in 45-75 year old primary care patients, in both postal and nurse-supported trial arms compared to controls.  We explored associations between intervention implementation measures and change in PA outcomes. Methods In accordance with the MRC guidance and framework (2014), the methods were selected through a key function model. Three quantitative aspects of the process evaluation relating directly to PA outcomes at 12 months were identified to assess intervention implementation: nurse session attendance (dose); PA diary completion (fidelity); and pedometer use (fidelity).  These were considered as independent variables in the multi-level models estimating the effectiveness of the intervention on PA outcomes (changes in step-counts and time in moderate-to-vigorous PA (MVPA) levels in ≥ 10 minute bouts).Results Dose: participants attending all 3 nurse sessions increased their step-count at 3 months by 961  steps more than those attending 0-2 sessions (95% CI 401 – 1520, p=0.001). Minutes of MVPA were also significantly increased by 64 (36, 92) at 3 months and by 28 (1, 54) at 12 months. Fidelity: both postal and nurse groups showed strong positive associations of diary return on step-count and minutes of MVPA at 3 months compared with those who didn’t return the diary: postal steps 1458 (854, 20161), nurse steps 873 (190, 1555), postal MVPA 64 (33, 94), nurse MVPA 47 (17, 75). These differences had decreased by 12 months, and only the postal group effects remained statistically significant: steps 1114 (538, 1689), MVPA 47 (17, 75). Regular pedometer use in the postal group was associated with higher step counts at 3 and 12 months: 1029 (383, 1675) and 606 (22, 1990) respectively. Regular pedometer use was not associated with PA outcomes in the nurse group.Discussion Process evaluation measures showed significant associations with most PA outcomes at 3 and 12 months. These were stronger for the postal than the nurse group for diary and pedometer use.  We cannot infer causality from these results, but the strong associations between nurse appointments, diary return, pedometer use and PA outcomes suggests that they were important factors in enabling the trial changes observed. We have shown the MRC framework to be an effective tool for process evaluation of intervention implementation.Keywords: Process evaluation; randomised controlled trial; Physical Activity

P87

A PROCESS EVALUATION OF THE IMPLEMENTATION OF ASSIST IN SCOTLAND

F Dobbie1, L Bauld1, R Purves1, J McKell1, N Dougall2, J White3, R Campbell4, A Amos5, L Moore6

1Faculty of Health Sciences and Sport, University of Stirling, Stirling, Scotland 2School of Health and Social Care, Edinburgh Napier University, Edinburgh, Scotland 3college of biomedical and life sciences, Cardiff University, Cardiff, Wales 4School of Social and Community Medicine, University of Bristol, Bristol, England 5School of Molecular, Genetic and Population Health Sciences, University of Edinburgh, Edinburgh,, Scotland 6Institute of Health and Wellbeing, University of Glasgow, Glasgow, Scotland

Background ASSSIT (A Stop Smoking in Schools Trial) is a peer led smoking prevention programme that encourages the dissemination of non-smoking norms. Students (aged 11-13) are nominated by their peers to become peer supporters. ASSIST is an evidence based programme with results from a large cluster randomised trial showing a reduction in smoking prevalence. However, these findings are now 13 years out of date and adolescent smoking prevalence has continued to decline. In 2014 ASSIST was piloted in Scotland. This presentation will present key findings from the Scottish evaluation offering points for consideration for the future delivery of ASSIST and further research areas.Methods Mixed method study with a range of stakeholders using qualitative (school staff, trainers, students, policy and commissioning leads n=101) and quantitative methods (a before and after student survey across 20 secondary schools in Scotland (n=2166, at follow-up).Results Feedback was overwhelmingly positive regarding the wider benefits of taking part in ASSIST for peer supporters (i.e. personal and communication skills) but also for the school and communities. Findings showed less certainty regarding the extent of message diffusion and any impact this may have had on adolescent smoking. Student survey results showed no significant change in self-reported smoking prevalence with 1.6% of pupils (n=33) reporting that they smoked one or more cigarettes per week increasing slightly to 1.8% (n=38) at follow-up. The student survey also indicated that conversation recall was low at 9% (n=145) Conclusion ASSIST is a well delivered, popular programme with additional benefit for students, their wider social network, school and community. Yet, there is uncertainty regarding the extent of message diffusion. Further research is needed to update the existing evidence base.Keywords: smoking prevention, adolescent, peer intervention

P88

DEVELOPING ROBUST METHODS FOR A LARGE SCALE, MULTI-SITE QUALITATIVE POLICY EVALUATION

C Guell1, 2, N Unwin2, 3, TA Samuels3, 4, L Bishop4, MM Murphy4

1European Centre for Environment and Human Health, University of Exeter, Truro, United Kingdom2MRC Epidemiology Unit and Centre for Diet and Activity Research (CEDAR), University of Cambridge, Cambridge, United Kingdom3Chronic Disease Research Centre, University of the West Indies, Bridgetown, Barbados4Faculty of Medical Sciences, University of the West Indies, Cave Hill Campus, Barbados

Background Non-communicable diseases (NCDs) are the leading health burden in all but the poorest countries, and there is an increasing interest in macro-level policy responses to tackle their upstream determinants. There is also an increasing use of qualitative methods for evaluation research, in particular when evaluating multi-component and multi-level initiatives. However, methodological opportunities offered by qualitative research, such as its strength to capture interconnections, complexities and inconsistencies, are limited by challenges, such as its small scale and in-depth rather than broad reach. We report methodological considerations, challenges and solutions.Methods We conducted a qualitative evaluation study of seven Caribbean countries in 2015 to assess the progress made in NCD policy measures. This comprised policy document analysis, and 76 semi-structured interviews with 80 stakeholders inside and outside government. Interviews were conducted by six interviewers organised in regional teams, and analysed by an expanded team under the guidance of the authors. Data collection and analysis protocols for this relatively large-scale project were developed iteratively in workshops.Results A first consideration was to purposively sample across settings, sectors and professional roles. This was a challenge of scale, as stakeholders were initially drawn from key informants and existing networks, and then cascaded by eliciting further recommendations to cover relevant sectors (government ministries, private, civil society), organisations within these sectors, and roles (technical, executive, elected). Some stakeholders were recommended because it was perceived as ‘politically’ important to include them, even if information elicited was less ‘rich’ in terms of relevant technical expertise or topic insight. Second, it was challenging to analyse and synthesise a large qualitative dataset across similar but distinct settings. Eleven researchers coded and categorised the data pragmatically according to the WHO NCD Action Plan, and this was guided by the Multiple Streams policy evaluation framework and realist evaluation principles to compare across contexts and themes. Later, the authors expanded the analysis to capture more inductive insights. Finally, there was an ethical challenge to secure anonymity of the participants as well as settings – e.g. not attributing policy shortcomings to individuals, organisations or even countries and their governments – while retaining relevant insights for each setting and political context.Conclusion Qualitative policy evaluation requires careful consideration and adaptation of standard research methods, the use of clear theoretical frameworks, and transparent interrogation of limitations. Its increasing popularity and use, and its detailed description and discussion among the research community, should enable the development of robust processes.Keywords: methods development, policy evaluation, stakeholder interviews

P89

HEALTH CARE PROFESSIONAL’S EXPERIENCES OF LIFESTYLE MANAGEMENT IN OVERWEIGHT AND OBESE PREGNANT WOMEN: A QUALITATIVE STUDY

C Flannery*1, 2, S McHugh2, L Kenny3, 4, M O’Riordan3, FM McAuliffe6, C Bradley5, PM Kearney2,M Byrne1.

1Health Behaviour Change Research Group, National University of Ireland, Galway, Galway, Ireland2Department of Epidemiology and Public Health, University College Cork, Cork, Ireland3Department Obstetrics and Gynaecology, University College Cork, Cork, Ireland4The Irish Centre forFetal and Neonatal Translational Research (INFANT), University College Cork, Cork, Ireland5Department of General Practice, University College Cork, Cork, Ireland6Obstetrics and Gynaecology, School of Medicine, University College Dublin, National Maternity Hospital,Dublin, Ireland

Background Obesity during pregnancy is associated with a number of complications including gestational diabetes mellitus (GDM). Currently, little is known about guidelines in clinical practice and the challenges faced by heath care professionals (HCPs). The aim of this study was to understand the perceptions, approach and challenges faced by midwives, obstetricians and general practitioners who provide antenatal care to women who are overweight and obese during pregnancy with the view to informing the development of an antenatal lifestyle intervention.Methods Semi-structured interviews were conducted with a purposive sample of health care professionals (HCPs) from Cork University Maternity Hospital (CUMH) (n=10) and with a sample of General Practitioners (GPs) working in primary care in the region (n=7). Data was collected until data saturation occurred. Interviews were digitally recorded and transcribed into NVivo V.10 software. Thematic analysis is on-going.Results Preliminary results identified 'knowledge of weight management’ and ‘antenatal services’ as key issues.  A lack of knowledge was evident involving risks, complications and initiating a conversation around overweight and obesity in pregnancy. Variation exists around what is considered appropriate weight gain and whether HCPs were following any particular guidelines. HCPs expressed concern about the dramatic increase in the number of pregnant women who are overweight and obese and how weight perception has changed in society.  Large ‘caseloads’ meant that lifestyle factors were not routinely discussed with the women and furthermore, a lack of communication is very evident between HCPs in the hospital and GPs in terms of the services provided.Conclusion HCPs expressed challenges when communicating with their patients about weight management in pregnancy. By ensuring midwives and other HCPs have the knowledge, skills and opportunity to discuss weight and lifestyle factors with pregnant women, the women, in turn, may be more motivated to maintain healthy behaviour’s during pregnancy.Keywords: Overweight, Obesity, Pregnancy, Gestational weight gain, General Practitioners, Health care professionals, Qualitative, Antenatal, Obstetrics

P90

SOCIO-ECONOMIC VARIATION IN CHILD BMI TRAJECTORY FROM INFANCY TO ADOLESCENCE IN THREE CONTEMPORARY EUROPEAN CHILD COHORTS

C McCrory1, S Leahy1, A.I. Ribeiro2, S Fraga2, H Barros2, R Layte3

1The Irish Longitudinal Study on Ageing (TILDA), Trinity College Dublin, Dublin, Ireland2EPIUNIT, Institute of Public Health University of Porto, Porto, Portugal3Department of Sociology, Trinity College Dublin, Dublin, Ireland

Background Rates of overweight and obesity have been shown to vary across socio-economic groups (SEG) from at least the age of three years but little is known about whether SEG differentials vary after adipose rebound and into adolescence and whether these trajectories differ by national context. This study examines socio-economic differentials in children’s body mass trajectories in infancy and extending across childhood into early adolescence in three contemporary child cohort studies. Methods Data on body mass index (BMI) measured on at least three occasions between birth and adolescence was obtained from four prospective cohort studies – Generation 21 (G21 - Portugal), Growing Up in Ireland (GUI - Ireland) (infant and child cohorts), and the Millennium Cohort Study (MCS - UK) – involving a total sample of 44,136 children. SEG differentials in children’s BMI trajectories was modelled by maternal educational level (primary, secondary, tertiary) using hierarchical models with fixed and random components for each cohort study.Results Child BMI growth trajectory was greater for children of lower educated mothers but only from three years of age. In G21, the educational differential emerged by 4 years of age and increased from 0.25boys [CI95=0.14, 0.38] and 0.44girls [CI95=0.30, 0.58] to 0.45boys [CI95=0.25, 0.64] and 0.70girls [CI95=0.48, 0.92] by 7 years of age. In GUI, the mean difference in BMI between polarised educational groups increased from 0.21boys [CI95=0.08, 0.35] and 0.35girls at 3 years of age [CI95=0.21, 0.49] to 0.92boys [CI95=0.63, 1.21] and 1.40girls [CI95=1.09, 1.71] by 13 years of age. In MCS, the educational differential was first observed at 5 years of age and increased from 0.14boys [CI95=0.06, 0.23] and 0.19girls [CI95=0.10, 0.28] to 0.66boys [CI95=0.49, 0.83] and 0.61girls [CI95=0.42, 0.79] by 11 years of age. Conclusion Socio-economic factors are strongly implicated in the aetiology of childhood obesity. This study shows that the socio-economic differentials emerge in early childhood and widen over time providing important policy evidence about the timing of potential policy interventions designed to eliminate the adverse life course health effects associated with early emerging adiposity. Keywords: health inequalities; body mass index; growth curves

P91

MECHANISMS OF ACTION IN GROUP-BASED INTERVENTIONS (MAGI) STUDY: A FRAMEWORK OF CHANGE PROCESSES IN GROUP-BASED HEALTH INTERVENTIONS

AJ Borek1, JR Smith1, C Abraham1, CJ Greaves1, S Morgan-Trimmer1, F Gillison2, M Jones3, M Tarrant1, R McCabe1

1Medical School, University of Exeter, Exeter, UK2Department for Health, University of Bath, Bath, UK3Health and Social Sciences, University of the West of England , Bristol, UK

Background Groups are often used to promote personal and health-related psychological and behavioural change. Although there is considerable literature on group dynamics and intra- and inter-personal change processes in groups, this knowledge is dispersed across different disciplines and rarely used in the design, delivery and evaluation of group-based health interventions. The aim of the Mechanisms of Action in Group-based Interventions (MAGI) study was to identify and enhance understanding of change processes in group-based health interventions by (1) developing a conceptual framework of change processes in groups, (2) identifying examples of these processes and practical strategies for facilitating them, and (3) exploring potential relationships between group processes and intervention engagement and outcomes in three group-based weight loss interventions.Methods Qualitative methods were used to develop the conceptual framework and identify examples of change processes and practical strategies. This involved (1) reviewing literature on theories of group dynamics and change in groups, qualitative studies, taxonomies of behaviour change techniques, and assessment tools to measure group processes; (2) reviewing and coding content of intervention manuals and 38 transcripts of group session recordings from three studies of group-based weight loss interventions; and (3) consultations with researchers, practitioners, facilitators and participants involved with group-based interventions. Further 24 transcripts of group sessions from one of the weight loss studies were coded using the framework and analysed to explore associations with intervention engagement and outcomes. Results Key concepts and processes were synthesised into a framework of change processes in group-based health interventions. Processes were categorised into: (1) group dynamic and development processes, (2) inter-personal change processes, and (3) intra-personal change processes in groups. Examples of practical strategies to facilitate and manage these processes were identified and categorised into: (1) group design elements, (2) group set-up tasks, (3) group facilitation strategies, and (4) group closure tasks. Work to identify examples of how group interaction and group processes might link with participant engagement and outcomes is still in progress but will be presented.Discussion The presented framework integrates a large body of literature on change processes in groups, and provides examples of practical strategies that can be used to instigate and facilitate change processes in group-based health interventions. It provides a practical tool for researchers and practitioners that can be used to design and deliver better group-based health interventions, train group facilitators, and guide evaluations of group-based interventions, with a view to optimising intervention engagement and outcomes.Keywords: group dynamics, behaviour change, weight loss

P92

LIFESTYLE INTERVENTIONS FOR THE TREATMENT OF OVERWEIGHT/OBESE ADOLESCENTS - COCHRANE REVIEW

LA-K Al Khudairy1, EL Loveman2, JC Colquitt2, EM Mead3, RJ Johnson1, HR Fraser1, OJ Joan Olajide3, MM Murphy1, RV Velho1, CO O'Malley4, LA Azevedo3, LE Ells3, MIM Metzendorf5, KR Rees1

1Division of Health Sciences, Warwick Medical School, Coventry, UK2Effective Evidence LLP, Waterlooville, UK3Health and Social Care Institute, Teeside University, Middlesbrough, UK4School of Medicine, Pharmacy and Health, Durham University Queen's Campus, Stockton-on-Tees, UK5Cochrane Metabolic and Endocrine Disorders Group, Medical Faculty of Heinrich-heine-University Dusseldorf, Dusseldorf, Germany

Background The prevalence of overweight and obese adolescents has increased worldwide, presenting a global public health crisis. This review assessed the efficacy of diet, exercise and behavioural interventions for the treatment of overweight/obesity in adolescents (12-17yrs).MethodsA systematic literature search (up to July 2016) with no language restrictions was performed in CENTRAL, MEDLINE, EMBASE, PsycINFO, CINAHL, LILACS, and the trial registers ClinicalTrials.gov and ICTRP Search Portal. Search terms included obesity, diet, exercise and adolescent. References of identified studies and systematic reviews were checked. Authors of included studies were contacted for missed studies. Two reviewers evaluated studies independently at all stages. Eligibility: RCTs that observed participants for ≥ six months, overweight/obese (investigator-assessed) adolescents (mean age 12-17yrs), interventions with a primary aim to treat overweight/obesity with any form of dietary, exercise and / or behavioural therapy delivered as a single or multi component intervention, any setting and any delivery method. Comparators were no treatment/wait list control, usual care or an alternative concomitant therapy providing it is delivered in the intervention arm. Primary outcomes were changes in BMI/weight measured at baseline and at ≥ 6 months. Data that could be meta-analysed were expressed as mean differences (MD) with 95% CI otherwise data were reported narratively. Heterogeneity, risk of bias and quality of evidence were assessed.  ResultsThe searches generated 16106 records (duplicates removed). Title and abstract screen excluded 15422 records. 736 records were assessed for eligibility. 105 trials were included: 50 ongoing; 11 awaiting classification, 44 qualitative synthesis and 39 quantitative synthesis. This review included 4682 participants. BMI change at the longest follow-up was MD -1.18 (95% CI -1.67 to -0.69); P < 0.00001; 2774 participants; 28 trials; low quality evidence. Most studies were multi-disciplinary interventions (BMI MD -1.18, 95% CI -1.75 to -0.61, P < 0.0001; 2293 participants; 22 studies), some studies were diet alone (BMI MD -0.62, 95% CI -1.29 to 0.06; P = 0.07; 277 participants; 3 studies) or exercise alone interventions (BMI MD -1.73, 95% CI -3.12 to -0.34, P = 0.01; 229 participants; 4 studies). Studies undertaken in schools showed a lower effect compered to community and health care settings.ConclusionInterventions that involve a combination of diet, exercise and behavioural components appear to be an effective treatment option for overweight/obese adolescents. Effects of lifestyle interventions were maintained at 18-24 months follow-up. Results should be interpreted with caution as the evidence was rated as low quality for inconsistency and publication bias.Keywords: adolescent obesity lifestyle

P93

WHAT ARE THE VIEWS OF OVERWEIGHT AND OBESE ADOLESCENTS (12-17YRS) ATTENDING LIFESTYLE TREATMENT INTERVENTIONS: A QUALITATIVE SYSTEMATIC REVIEW

HM Jones, L Al-Khudairy, GJ Melendez-Torres, O Oyebode1Warwick Medical School, University of Warwick, Coventry, UK

Background A third of all children in England are overweight or obese. Physical and psychosocial effects of obesity in adulthood are now seen in children and adolescents. Current NICE guidance recommends that obese adolescents attend a family-based multi-component weight management service. Cochrane reviews have shown that lifestyle programmes can be effective. However, weight management interventions recruit <2% of the childhood population. For those that do attend weight management interventions, attrition is high. Understanding the views of adolescents is necessary for planning and development of future interventions.The goals of this systematic review include:• To explore the views of overweight or obese adolescents aged 12-17 years who have attended a lifestyle treatment intervention.• To identify what adolescents deem as successful and unsuccessful components of an intervention to inform future practice, as well as research.Methods This review examined studies of overweight or obese adolescents who have attended a lifestyle treatment intervention. Interventions were single or multi-component and contained one or more diet, physical activity and behavioural elements. Only studies that collected and analysed data qualitatively were included. There were no language restrictions.Published literature was identified by searching the following databases: MEDLINE, EMBASE, Web of Science, PsycINFO, ASSIA and CINAHL. Reference lists of included studies were screened.Titles and abstracts of identified records were assessed independently by two reviewers. Full texts of all potentially relevant papers were retrieved. Two reviewers assessed each full text and reasons for exclusion were recorded. Differences in opinions were resolved by a third reviewer.Methodological quality was assessed by two reviewers independently as low, medium or high in terms of trustworthiness and reliability of findings. This was done using criteria developed by the EPPI-Centre. Full texts will be opened in NVivo and analysed using thematic synthesis, as described by Thomas and Harden (2008). The certainty of review findings will be assessed using the Confidence in the Evidence from Reviews of Qualitative Research (CERQual) approach.Results 19,230 records were identified for title and abstract screening. 284 full-texts were screened, with 30 included in the final sample. Conclusion This review is due for completion in June 2017. Findings will inform development of an adolescent weight management intervention in Coventry, West Midlands.To date, no systematic review has focused on adolescent views of weight management interventions. The views of adolescents should be taken as a starting point when planning future initiatives.Keywords: Obesity, adolescent, systematic review

P94

ASSOCIATIONS BETWEEN MATERNAL PREGNANCY, SOCIAL AND LIFESTYLE CHARACTERISTICS AND OFFSPRING BLOOD PRESSURE AT AGE 4/5 IN WHITE BRITISH AND PAKISTANI ORIGIN PARTICIPANTS IN THE BORN IN BRADFORD STUDY

J West1, G Santorelli1, P Collings1, J Wright1, D Lawlor2, 3

1Born in Bradford, Bradford Institute for Health Research, Bradford, UK2School of Social and Community Medicine, University of Bristol, Bristol, UK32MRC Integrative Epidemiology Unit , University of Bristol, Bristol, UK

Background South Asians have an increased risk of stroke and coronary heart disease (CHD) compared to white European populations. High blood pressure (BP) is an important risk factor for CHD and higher diastolic BP has previously been reported in both South Asian adults and children. Our aim was to examine whether maternal pregnancy, social and lifestyle characteristics are associated with BP at age 4/5 and contribute to these differences.Methods Born in Bradford is a prospective study of children born to 12,453 mothers between 2007 and 2010 in Bradford, UK. All mothers completed an oral glucose tolerance test in pregnancy and provided detailed social and lifestyle data. We present data from 2129 White British and 2528 Pakistani mother/offspring pairs from the Born in Bradford study. Associations of BMI, gestational diabetes (GD), fasting and post load glucose, maternal hypertension (HDP), smoking in pregnancy and maternal education with offspring systolic and diastolic BP at age 4/5 were examined. We adjusted our results for sex and age at BP measurement.Results Pakistani children had lower systolic (mean difference -0.16 95% CI -0.79, 0.47) but higher diastolic (mean difference 1.38 95% CI 0.74, 2.03) compared to White British children. In Pakistani children maternal BMI and HDP were strongly associated with higher systolic and diastolic BP but the effect was minimal and weak in White British children. Associations of maternal glucose and BP were consistent with the null hypothesis in both groups. Smoking and education were not associated with BP in either group.Conclusion Ethnic differences in systolic and diastolic BP are present in children at age 4/5. Consistent with findings in adults, Pakistani origin children have on average lower systolic and higher diastolic BP compared to White British children. Maternal BMI and HDP influence BP in Pakistani children but have minimal effect in White British children. These results suggest a role for some maternal pregnancy and lifestyle exposures in ethnic differences in systolic and diastolic BP at age 4/5.Keywords: Ethnic differences; child blood pressure, maternal influences

P95

A Cassidy1, O Molaodi1, M Green1, L Moore1, S Harding2

1MRC CSO Social and Public Health Sciences Unit, University of Glasgow, Glasgow, UK2Diabetes & Nutritional Sciences Division, King's College London, London, UK

ETHNIC VARIATIONS IN CLUSTERING OF ADOLESCENT HEALTH RISK BEHAVIOURS: LATENT CLASS ANALYSIS

Background There is evidence from existing literature of ethnic variations and clustering in adolescent health risk behaviours. However, it is not known whether there are ethnic variations in health risk behaviour clustering. We aimed to investigate ethnic variations in the clustering of health risk behaviours using the Determinants of young Adults Social wellbeing and Health study (DASH).Methods Latent class analysis and multinomial logistic regression were used to investigate adolescent health risk behaviour clustering and ethnic variations in clustering in the MRC DASH study. Data were collected from 51 schools across 8 London boroughs. In 2005-06, 4785 pupils were followed up at 14-16 years old. Age, gender, ethnicity, substance use (SU) behaviours (current tobacco and alcohol, and lifetime illicit drug use), daily fruit and vegetables (FV), weekly physical activity (PA), and body mass index (BMI) were recorded.Results A model with four latent classes was selected (entropy: 0.757). Classes could be characterized as high-SU/high-PA (n=232), high-SU/low-PA (n=811), low-SU/PA/high-FV (n=1471), and low-SU/low-PA/FV (n=2260).Using the low-SU/PA/high-FV class as reference males were more likely to be in the high-SU/high-PA class (OR: 1.89; 95% CI: 1.39-2.57), and less likely to be in the high-SU/low-PA class (OR: 0.56; 95% CI: 0.47-0.67) than females.Compared to White UK, Black Caribbean adolescents were less likely to be in the high-SU/high-PA (OR: 0.48; 95% CI: 0.30-0.76), or high-SU/low-PA (OR: 0.34; 95% CI: 0.26-0.45) classes. Black Africans were less likely to be in the high-SU/high-PA (OR: 0.18; 95% CI: 0.10-0.31), high-SU/low-PA (OR: 0.11; 95% CI: 0.08-0.15), or low-SU/low-PA/FV (OR: 0.55; 95% CI: 0.44-0.69) classes. Indians were less likely to be in the high-SU/high-PA (OR: 0.36; 95% CI: 0.19-0.71), or high-SU/low-PA (OR: 0.17; 95% CI: 0.11-0.26) classes. Pakistanis/Bangladeshis were less likely to be in the high-SU/high-PA (OR: 0.40; 95% CI: 0.23-0.68), high-SU/low-PA (OR: 0.08; 95% CI: 0.05-0.13), or low-SU/low-PA/FV (OR: 0.65; 95% CI: 0.50-0.85) classes.Conclusion Latent class analysis is a valuable method for investigating ethnic variations in adolescent lifestyles. Ethnic minority adolescents tend to be in classes which are characterized by less unhealthy behaviour; patterns also vary between ethnic minority groups. Research needs to investigate risk and protective factors that may explain these ethnic variations to identify likely intervention targets and to inform public health policy.Keywords: Ethnicity Adolescence Behaviours

P96

ETHNIC DENSITY EFFECTS FOR ADULT MENTAL HEALTH: SYSTEMATIC REVIEW AND META-ANALYSIS OF INTERNATIONAL STUDIES

L Becares1, M Dewey2, J Das-Munshi21Social Statistics, University of Manchester, Manchester, England2Department of Health Services & Population Research, King’s College London, Institute of Psychiatry, Psychology & Neuroscience, London, England

Background Despite increased ethnic diversity in more economically developed countries it is unclear whether residential concentration of ethnic minority people (ethnic density) is detrimental or protective for mental health. This is the first systematic review and meta-analysis covering the international literature to assess ethnic density associations with mental health outcomes.Methods We systematically searched Medline, PsychInfo, Sociological Abstracts, Web of Science from inception to March 31st, 2016, for observational studies using search terms for the following study outcomes: Depression, anxiety and the common mental disorders, suicide, suicidality, psychotic experiences and psychosis. We obtained additional data from study authors. We conducted random effects meta-analysis taking into account clustering of estimates within datasets. Meta-regression assessed heterogeneity in studies due to ethnicity, country, generation and area-level deprivation. This review is registered with PROSPERO CRD42014012992.Results Out of 2288 studies identified 41 met inclusion criteria. 12 studies provided data for meta-analysis and meta-regression. In the meta-analyses, we found a large reduction in relative odds of psychotic experiences (OR: 0.82 (95% CI: 0.76-0.89) (nine estimates)) and suicidal ideation (OR: 0.88 (95% CI: 0.79-0.98) (ten estimates)) for each 10 percentage-point increase in own ethnic density. For common mental disorders, depression and anxiety, associations were indicative of protective effects of own ethnic density however results were not statistically significant. Conclusion The findings support consistent protective ethnic density associations across countries and racial/ethnic minority populations as well as mental health outcomes. This is indicative of the importance of the social environment in patterning detrimental mental health outcomes in marginalized and excluded population groups. Keywords: mental health, ethnicity, ethnic density

P97

THE “PROBLEM” OF ROMA HEALTH AND WELLBEING: A CRITICAL ANALYSIS OF EUROPEAN POLICY PERSPECTIVES

LC Orton1, S Sheard1

1Public Health & Policy, University of Liverpool, Liverpool, UK

Background Roma are the largest minority ethnic group in Europe. They have faced ongoing discrimination and persecution, with reports of large and widening social and health disparities between Roma and non-Roma. Most analyses of the policies that aim to improve the situation for Europe’s Roma compare the pros and cons of proposed solutions. They do not reveal the assumptions that shape how the “problem” of Roma health and wellbeing is understood in policy proposals and how this influences the solutions that are – or are not - put forward. The aim of this study is to explore the conditions that have allowed certain representations of the “problem” of Roma health-and-wellbeing to acquire dominance/suppression during the 1980–2015 period of economic transition and successive eastward EU expansion. Methods Five oral history interviews were conducted with European policy specialists to provide accounts of the development of European policy relevant to Roma health and wellbeing and to map key documents produced by EU decision-making institutions (the Parliament, Council and Commission). Further documents were located through catalogue searches. All retrieved items were mapped/categorised before five policies relating to pivotal shifts in the way Roma health and wellbeing has been understood as a policy problem were selected. Analyses drew on Bacchi’s “What’s the problem represented to be?” post-structural analysis framework  to explore: 1) underlying frameworks of thought that shape problem representations; 2) the conditions that allowed shifts in problem representations to occur; and 3) how these problem representations shaped the different solutions that have been proposed. Results Since the 1980’s policies produced by the main EU institutions have situated the concern for Roma health and wellbeing within a wider drive for economic growth. A perceived low productivity among Europe’s Roma populations is attributed to their low educational attainment and subsequently reduced employment opportunities. Inadequate housing/living conditions and low uptake of health care services are seen to compound the “problem” leaving Roma prone to poor health that further limits their productivity. The proposed solution is to remove the barriers that are perceived to prevent Roma from accessing mainstream health, education, employment and housing. Conclusion With a perceived increase in the size and mobility of its Roma populations, recent EU policy has framed “Roma health” as a barrier to economic development. Policy solutions frame improvements in Roma health as a means to improve overall productivity of the region, rather than as an end in itself.Keywords: Policy, Roma, health inequalities

P98

ASSESSING CHANGE IN PHYSICAL ACTIVITY LEVELS IN ADOLESCENT ETHNIC GROUPS

P Bhatnagar, N Townsend1Nuffield Department of Population Health, University of Oxford, Oxford, UK

Background Previous research has established that physical activity declines during childhood, but we have inadequate information on how these changes differ by ethnic group. Determining if inequalities exist between ethnic groups is vital for understanding whether different intervention strategies are necessary. We aimed to determine how physical activity has changed over time within different UK ethnic groups.Methods We used Understanding Society, a nationally representative UK-based longitudinal survey to track changes in physical activity over time in different ethnic groups. For children aged 10 to 15, we explored the frequency of participating in sports and mode of travel to school. These factors were measured in four of the available six waves (1, 2 4 and 6), between 2009 and 2015. Children were asked how frequently they participate in sports per week. For mode of travel, children reported whether they walked, cycled, used public transport, were driven or used a combination of these. When including children who were followed up, the sample size was too low to stratify by age or gender. We used linear regression to assess if ethnicity was associated with change in sports participation and mode of transport to school.Results At baseline, low sports participation (6%) and inactive commuting (54%) was highest in Pakistani children; differences were significant between ethnic groups for mode of travel to school (p <0.00).  Frequency of sports participation declined for all ethnic groups between waves 1 and 2, 1 and 4 and 1 and 6. The majority of children did not increase or decrease the overall activity level of their mode of transport to school between any of the waves. Most groups continued to use an active method, though between waves 1 and 4, 50% of Pakistani children retained an inactive method of commuting to school. Despite some differences, linear regression showed that ethnicity was not significantly associated with change in frequency of sports participation or change in mode of transport between any of the waves. Conclusion Interventions aimed at preventing the decline of physical activity in children from the general population may be appropriate for ethnic minority groups. As active travel to school remains low in some ethnic minorities, targeted interventions to initially change mode of commuting in ethnic minority children is still needed.Keywords: physical activity, ethnicity, adolescents