Comparing Results from the England and Wales, Scotland and Northern Ireland Longitudinal Studies: Health and Mortality as a case study Census Microdata.

Comparing Results from the Comparing Results from the England and Wales, Scotland and England and Wales, Scotland and

Northern Ireland Longitudinal Studies:Northern Ireland Longitudinal Studies:

Health and Mortality as a case studyHealth and Mortality as a case study

Census Microdata ConferenceCensus Microdata Conference11stst September 2008 September 2008

Harriet Young, Emily GrundyHarriet Young, Emily GrundyLondon School of Hygiene and Tropical MedicineLondon School of Hygiene and Tropical Medicine

Paul BoylePaul BoyleUniversity of St AndrewsUniversity of St Andrews

Dermot O’ReillyDermot O’ReillyQueens University BelfastQueens University Belfast

BackgroundBackground Mortality patterns Mortality patterns

differ between differ between constituent countries constituent countries of the UK – Scotland of the UK – Scotland has highest mortalityhas highest mortality

100.67 - 171.34171.34 - 195.47195.47 - 216.8216.8 - 253.42253.42 - 428.1

Standardised mortality rates by Local Authority across the UK

Std. Rate per 100,000 population

From Rosato and O’Reilly 2004

Self-reported health Self-reported health patterns also vary patterns also vary between countries – between countries – high levels of poor high levels of poor self-reported health in self-reported health in Northern Ireland and Northern Ireland and WalesWales

2643.9 - 45124512 - 5400.65400.6 - 5961.35961.3 - 7256.87256.8 - 12708.8

Std. Rate per 100,000 population

From Rosato and O’Reilly 2004

Standardised rates of poor general health by Local Authority across the UK

Research AimsResearch Aims

1. Analyse socio-economic and socio-1. Analyse socio-economic and socio-demographic variation in reporting poor demographic variation in reporting poor health in 2001 in England and Wales, health in 2001 in England and Wales, Scotland and Northern IrelandScotland and Northern Ireland

2. Analyse associations between 2. Analyse associations between reporting of health status in 2001 and reporting of health status in 2001 and subsequent mortality (taking account of subsequent mortality (taking account of socio-economic and socio-demographic socio-economic and socio-demographic factors)factors)

DatasetsDatasets

ONS Longitudinal Study – England and Wales (ONS LS)ONS Longitudinal Study – England and Wales (ONS LS)Scottish Longitudinal Study (SLS)Scottish Longitudinal Study (SLS)Northern Ireland Longitudinal Study (NILS)Northern Ireland Longitudinal Study (NILS)

Record linkage studiesRecord linkage studies Samples drawn from census data, based on a number of Samples drawn from census data, based on a number of

birthdays in the yearbirthdays in the year After initial census starting point, subsequent census After initial census starting point, subsequent census

points linked in points linked in Vital events data linked in (birth to sample mother, death Vital events data linked in (birth to sample mother, death

of spouse, death & other) of spouse, death & other) Samples maintained by addition of new births and Samples maintained by addition of new births and

immigrantsimmigrants

DatasetsDatasets

Census Census data fromdata from

% of % of population population in samplein sample

# of # of birthdays birthdays selectedselected

Sample Sample size in size in 20012001

ONS LSONS LS 1971 - 1971 - 20012001

11 44 Approx Approx 500,000500,000

DatasetsDatasets

Census Census data fromdata from

% of % of population population in samplein sample

# of # of birthdays birthdays selectedselected

Sample Sample size in size in 20012001

ONS LSONS LS 1971 - 1971 - 20012001

11 44 Approx Approx 500,000500,000

SLSSLS 1991, 1991, 20012001

5.35.3 2020 Approx Approx 274,000274,000

DatasetsDatasets

Census Census data fromdata from

% of % of population population in samplein sample

# of # of birthdays birthdays selectedselected

Sample Sample size in size in 20012001

ONS LSONS LS 1971 - 1971 - 20012001

11 44 Approx Approx 500,000500,000

SLSSLS 1991, 1991, 20012001

5.35.3 2020 Approx Approx 274,000274,000

NIILSNIILS 20012001 2828 104104 Approx Approx 500,000500,000

| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | |1971 1976 1981 1986 1991 1996 2001 2006 2011

Birth ofChild

Birth ofChild Census Emigration Re-entry

Death ofSpouse Census Cancer Death

Example timeline for an individual Example timeline for an individual in the ONS LSin the ONS LS

Developmental AimDevelopmental Aim

To demonstrate how parallel and if To demonstrate how parallel and if possible combined analyses of the 3 possible combined analyses of the 3 datasets can give UK wide resultsdatasets can give UK wide results Consider issues that could undermine cross-Consider issues that could undermine cross-

country analysis: address and document country analysis: address and document Negotiate procedures facilitating UK wide Negotiate procedures facilitating UK wide

analysis and a user guide for others analysis and a user guide for others

ConfidentialityConfidentiality

Access to individual level data only in Access to individual level data only in relevant Statistical Office safe setting.relevant Statistical Office safe setting.

Outputs from safe setting governed by Outputs from safe setting governed by disclosure guidelines.disclosure guidelines.

Variations in data access between the Variations in data access between the studies reflect studies reflect differences in sampling fractionsdifferences in sampling fractions legal differences.legal differences.

Analysis strategyAnalysis strategy

1. Parallel analysis using individual level 1. Parallel analysis using individual level data in relevant safe settingdata in relevant safe setting

2. Parallel analysis using aggregated 2. Parallel analysis using aggregated datasets (cleared to researcher if possible)datasets (cleared to researcher if possible)

3. Combined analysis3. Combined analysis Combined individual level data not possibleCombined individual level data not possible combined aggregated datacombined aggregated data

Sample populationSample population Population aged 35 and overPopulation aged 35 and over Present at 2001 CensusPresent at 2001 Census Follow up to end 2005Follow up to end 2005

Analyses separate by genderAnalyses separate by gender Analyses separate by broad age group: 35-54, Analyses separate by broad age group: 35-54,

55-74, 75+55-74, 75+

This presentation: individual country analysis for This presentation: individual country analysis for England and Wales, and Northern IrelandEngland and Wales, and Northern Ireland

Outcome variablesOutcome variables Self-rated health Self-rated health

““Over the last twelve months would you say your Over the last twelve months would you say your health has on the whole been: health has on the whole been:

Good ? Fairly good ? Not good ?”Good ? Fairly good ? Not good ?” Limiting long term illness Limiting long term illness

““Do you have any long term illness, health problem or Do you have any long term illness, health problem or disability which limits your daily activities or the work disability which limits your daily activities or the work you can do ? (include problems which are due to old you can do ? (include problems which are due to old age)”age)”

Yes, NoYes, No MortalityMortality

2001 Census to end 20052001 Census to end 2005

Explanatory VariablesExplanatory Variables

Self rated health and limiting long term illness Self rated health and limiting long term illness (for Aim #2)(for Aim #2)

Age, gender, marital statusAge, gender, marital status Education, NS-SECEducation, NS-SEC Housing tenure, car accessHousing tenure, car access Region (for ONS LS only)Region (for ONS LS only)

1. Variation in health reporting1. Variation in health reportingPrevalence of limiting long term illness by age, gender and country,

ONS LS and NILS 2001.

0

10

20

30

40

50

60

70

80

90

100

35-54 55-74 75+ 35-54 55-74 75+

Men Women

per

cen

t England

Wales

Northern Ireland

Odds ratios from logistic regression showing age and Odds ratios from logistic regression showing age and gender differences in reporting a limiting long term gender differences in reporting a limiting long term

illness by country, ONS LS and NILS, 2001.illness by country, ONS LS and NILS, 2001.

* p<0.05 ** p<0.01 *** p<0.001

Models controlled for age group and gender only

EnglandEngland WalesWales Northern IrelandNorthern Ireland

OROR P valueP value OROR P valueP value OROR P valueP value

Gender (reference: men)Gender (reference: men)

WomenWomen 1.031.03 ****** 0.940.94 1.071.07 ******

Age group (reference: 35-54)Age group (reference: 35-54)

55-7455-74 3.103.10 ****** 3.593.59 ****** 3.213.21 ******

75+75+ 9.429.42 ****** 8.798.79 ****** 8.688.68 ******

Odds ratios from logistic regression showing the association of Odds ratios from logistic regression showing the association of socio-economic factors and limiting long term illness bysocio-economic factors and limiting long term illness by

country for men aged 35-54. ONS LS and NILS, 2001.country for men aged 35-54. ONS LS and NILS, 2001. England & WalesEngland & Wales Northern IrelandNorthern Ireland

Odds RatioOdds Ratio P valueP value Odds RatioOdds Ratio P valueP value

Housing tenure (Housing tenure (referencereference: owner occupier): owner occupier)

Private rentPrivate rent 1.511.51 ****** 2.032.03 ******

Social rentSocial rent 2.482.48 ****** 2.402.40 ******

Education (Education (referencereference: upper secondary or degree level): upper secondary or degree level)

Lower sec’ryLower sec’ry 1.061.06 ** 1.361.36 ******

NoneNone 1.481.48 ****** 1.991.99 ******

Car access (Car access (referencereference : has car) : has car)

No carNo car 1.901.90 ****** 1.401.40 ******

* p<0.05 ** p<0.01 *** p<0.001

Models controlled for age, marital status, tenure, car access, education, social class, region for England and Wales

2. Association of morbidity and mortality2. Association of morbidity and mortality

Hazard ratios of mortality for those with an LLTI compared with no LLTI by country for women, comparing crude and adjusted models,

NILS and ONS LS 2001-2005

0

0.5

1

1.5

2

2.5

3

3.5

Age adjusted Fully adjusted Age adjusted Fully adjusted

England and Wales Northern Ireland

Haz

ard

rat

io

Crude models controlled for age

Adjusted models controlled for age, marital status, housing tenure, car access, education, social class, and region for England and Wales

Crude models controlled for age

Adjusted models controlled for age, marital status, housing tenure, car access, education, social class, and region for England and Wales

Hazard ratios of mortality for those with an LLTI compared with no LLTI by country for women, comparing crude and adjusted models,

NILS and ONS LS 2001-2005

0

0.5

1

1.5

2

2.5

3

3.5

Age adjusted Fully adjusted Age adjusted Fully adjusted

England and Wales Northern Ireland

Haz

ard

rat

io

Hazard ratios of mortality for those with an LLTI comapred with no LLTI by age and country, fully adjusted models for women, ONS LS

and NILS 2001-05

0

1

2

3

4

5

6

35-54 55-74 75+ 35-54 55-74 75+

England and Wales Northern Ireland

Haz

ard

Rat

io

Adjusted models controlled for age, marital status, housing tenure, car access, education, social class, and region for England and Wales

DiscussionDiscussion

Results to date indicate:Results to date indicate: Are differences between England, Wales Are differences between England, Wales

and Northern Ireland in gender and age and Northern Ireland in gender and age effects on reporting poor health.effects on reporting poor health.

In all areas, reporting a limiting long term In all areas, reporting a limiting long term illness and reporting poor self-rated health illness and reporting poor self-rated health or an illness are associated with mortality.or an illness are associated with mortality.

Limitations: dataset comparabilityLimitations: dataset comparability

Variable differences e.g. educationVariable differences e.g. education Differences in question order in Census Differences in question order in Census

form between Scotland and other form between Scotland and other countriescountries

Differences in methods of event data Differences in methods of event data collection e.g. on emigrations: NILS, ONS collection e.g. on emigrations: NILS, ONS LSLS

Small sample size for WalesSmall sample size for Wales

Future activitiesFuture activities

Analysis of data from the SLSAnalysis of data from the SLS Creation of aggregated (weighted count) Creation of aggregated (weighted count)

datasets for each countrydatasets for each country Appending aggregated datasets and Appending aggregated datasets and

carrying out combined country analysescarrying out combined country analyses Development of tools to aid others to carry Development of tools to aid others to carry

out analyses on these datasets in futureout analyses on these datasets in future

ONS Longitudinal Study ONS Longitudinal Study www.celsius.lshtm.ac.ukwww.celsius.lshtm.ac.uk

Northern Ireland Longitudinal StudyNorthern Ireland Longitudinal Studyhttp://www.nisra.gov.uk/nils/default.asp.htmhttp://www.nisra.gov.uk/nils/default.asp.htm

Scottish Longitudinal StudyScottish Longitudinal Studywww.lscs.ac.ukwww.lscs.ac.uk

For more information:For more information: