Social epidemiology at ONS

Social epidemiology at the Office for National Statistics

Myer Glickman 23 March 2006

Overview

• Who/what/why is ONS?• What do we do?• Social epidemiology – examples• Current developments in ONS

1. Who/what/why is ONS?

What is ONS?

• The Office for National Statistics (ONS) is the government department that provides UK statistical and registration services.

• ONS is responsible for producing a wide range of key economic and social statistics which are used by policy makers across government to create evidence-based policies and monitor performance against them.

• The Office also builds and maintains data sources both for itself and for its business and research customers. It makes statistics available so that everyone can easily assess the state of the nation, the performance of government and their own position.

History

• 1996 Office for National Statistics– including Central Statistical Office

• 1970 Office for Population Censuses and Surveys– merger of GRO and Govt Social Survey

• 1941 Wartime Social Survey• 1837 General Register Office

ONS offices & functions

• London: head office, methodology, health and care, social reporting, labour market, economics

• London: Family Records Centre• Newport: business surveys, economic

methodology, finance and IT support functions• Titchfield: vital events processing, census

management, population estimates and projections, migration statistics, Neighbourhood Statistics, geography support

• Southport: registration management, certificate services, NHS Central Register

2. What do we do?

Census

And it came to pass in those days, that there went out a decree from Caesar Augustus, that all the world should be taxed.

• 1801 First census in Great Britain• 1841 First ‘modern’ census• 1911 Partial mechanised processing• 1961 Computer processing

Vital Statistics

Social Surveys - THEN

“Prying around and asking a lot of silly questions about morale and upsetting the public.”

• Corset stocks and needs, with special reference to the allocation of steel

• Shortage of domestic brooms and brushes• Diets of young people aged 14-18• Evaluation of publicity campaign to eat more

potatoes• Prevention of venereal disease• Prevalence of illness in the general population

Social Surveys - NOW

• We carry out 8 continuous surveys throughout the year, and over 30 ad-hoc surveys.

• We collect information from more than half a million people each year through personal interviews, telephone interviews, and postal surveys.

• General Household Survey• Labour Force Survey• Family Resources Survey• Expenditure and Food Survey (FES & NFS)• International Passenger Survey• 2008 - Integrated Household Survey

The ONS Longitudinal Study (LS)

• 1% representative sample of England and Wales• Four birth dates plus household members• Started from 1971 census• All census data 1971-2001, births, deaths, cancer reg.,

migration (some years)• Forthcoming: benefits claims, estimated income

Mortality of Males 15-64 by Economic Position (Standardised Mortality Ratios)

0

100

200

300

400 Active Inactive

Employed Seeking work Off work Retired Permanently

disabled Student Other

Source: Fox A.J. and Goldblatt P.O. (1982)

SMR

3. Social epidemiology in ONS

• Miners die in undue proportions … • tailors die in considerable numbers at the younger

ages (25-45) … • labourers’ mortality is at nearly the same rate as

that of the whole population, except in the very advanced ages.

– William Farr, observations based on comparison of the 1851 census with death registrations

Our research focusses mainly on:

• Data to which we have privileged access– Census, birth & death registrations, major surveys

• Methodology and epidemiological ‘basic science’– Data quality, coding & classification, statistical methods

• Issues prioritised by government– Other government departments’ targets & indicators– Background on important policy issues– Making information available to public and academics

We do not:– Comment on government policy– Do ‘academic’ research (but… )

Examples of ‘unpublished’ work

• Implementation of ICD10• Issues in occupational coding• Development of NS SEC social classification• Implementation of SOC 2000 and NS SEC in vital

statistics• Maintenance of Longitudinal Study (decennial,

annual, coding & classification, access)• Development of small area life expectancy

methods• Building GHS time series database

Examples of publications:

• Neighbourhood Statistics website• Annual report on life expectancy for local authority

and health organisations• Decennial volume on geographical variations in

health (includes social factors)• Decennial volume on occupational health

(previously mainly mortality & cancer)• Life expectancy by social class (methods & results)• Mortality by social class and cause of death• Articles from EU collaborations

Trends in social class differences in mortality by cause, 1986 to 2000

Chris WhiteFolkert van GalenYuan Huang Chow

Office for National Statistics

Background

• Death rates for all social classes fell over the 1990s• Inequality increased, however, because deaths in the

higher social classes fell more than deaths in the lower social classes

• Previous analysis of social class differences by cause of death showed social gradients for all major causes in men

• The picture for women is more complex and shows lesser inequalities

Methods

• Figures shown here are directly standardised mortality rates for men and women in 1986-92, 93-96 and 97-99, using the ONS Longitudinal Study

• Paper also shows proportional mortality ratios for men and women in England and Wales and men in Scotland, based on death registrations

• Age group is 35-64 for consistency with previous research and to maximise proportion of deaths coded to a social class

Social class differences in deaths from coronary heart disease

Directly age standardised mortality rates per 1000 men aged 35-64 in England and Wales

0

50

100

150

200

250

300

1986-92 1993-96 1997-99

I&II IIIN IIIM IV&V

Source: ONS Longitudinal Study

Social class differences in deaths from lung cancer

Directly age standardised mortality rates per 1000 men aged 35-64 in England and Wales

0

25

50

75

100

1986-92 1993-96 1997-99

I&II IIIN IIIM IV&V

Source: ONS Longitudinal Study

Social class differences in deaths from respiratory diseases

Directly age standardised mortality rates per 1000 men aged 35-64 in England and Wales

0

25

50

75

100

1986-92 1993-96 1997-99

I&II IIIN IIIM IV&V

Source: ONS Longitudinal Study

Social class differences in deaths from breast cancer

Directly age standardised mortality rates per 1000 women aged 35-64 in England and Wales

0

25

50

75

100

1986-92 1993-96 1997-99

I&II IIIN IIIM IV&V

Source: ONS Longitudinal Study

Key Findings

• Between 1986-92 and 1997-99 inequalities in major causes of death in men increased

• The relative likelihood of a man aged 35 to 64 in social class IV/V dying of lung cancer, compared to a man on social class I/II, rose from 2.2 times to 3.1 times

• For respiratory diseases, the rise was from 3.8 times to 5.6 times

• Inequalities for women were less marked• A woman in social class IV/V in 1997-99 was less likely

to die of breast cancer than a woman in social class I/II

The effect of individual circumstances, area and changes over time on mortality in men, 1995-2001

• Chris White Health & Care (now SEMARD), ONS• Prof Dick Wiggins City University, London• Prof David Blane Imperial College, London• Alison Whitworth Methodology Group, ONS• Myer Glickman Health & Care (now SEMARD), ONS

With acknowledgements to: Nargis Rahman, Methodology Group, ONS Adele Russell, Health & Care, ONS

Person, Place or Time?

Research issues

• Significance of life-course perspective on inequalities: accumulation of risks over time

• Long-running debates on composition v. context in geographical inequalities in health: the place or the people who live there?

• Not enough evidence on how geographical, household and individual socioeconomic factors compare in their effect on mortality

Methods - inclusion criteria and outcome

• Sample of 49,951 men from the ONS Longitudinal Study

– Aged between 26 and 71 years in 1971;

– Resident in private households in E&W;

– Relevant data recorded at 1971,1981 and 1991 censuses;

– Traced at NHS Central Register;

– Not known to have emigrated between 1991-2001

• Outcome measure was death in 1995-2001

– 6,906 deaths (13.8%)

Methods - analytical approach

• Hierarchical two-level logistic random intercepts model using MLWin and Stata

• Additional ‘virtual levels’ within individual level variables

• Calculation of transition variables reflecting change

• Level 2 – Local authority district

• Level 1 –» 1a South East Region or not

» 1b LA ward (Carstairs index quintile)

» 1c Household

» 1d Individual

Effect of (a) Ward deprivation and (b) South East residence

1.13

1.31

1.15

1.05

0.90

1.00

1.10

1.20

1.30

1.40

1.50

Quintile 1 Quintile 2 Quintile 3 Quintile 4&5 Yes No

CARSTAIRS WARD DEPRIVATION SOUTH EAST RESIDENCE

ODDS OF DEATH 1995-2001

Effect of (c) Social Class in 1971

1.54

1.371.31

1.26

0.90

1.00

1.10

1.20

1.30

1.40

1.50

1.60

1.70

1.80

I II IIINM IIIM IV&V

ODDS OF DEATH 1995-2001

Effect of (d) Housing tenure 1991 and (e) Unemployment

1.22

1.41

1.27

0.90

1.00

1.10

1.20

1.30

1.40

1.50

1.60

1.70

Owner Occupied Private Rented Social Housing No Yes

HOUSING TENURE 1991 UNEMPLOYED AT CENSUS

ODDS OF DEATH 1995-2001

Effect of (f) Social mobility and (g) Housing tenure mobility

1.17

1.121.15

1.05

0.90

0.95

1.00

1.05

1.10

1.15

1.20

1.25

1.30

Upward Stable Downward Owner Occ. 3 Rented 3 Tenurechange

SOCIAL MOBILITY TENURE TRAJECTORY

ODDS OF DEATH 1995-2001

Interaction of Social class mobility 1971-91 and Ward deprivation score 1991

Effect of illustrative profiles on riskRisk of death in males 1995-2001 by illustrative group and age: probability and 95% CI

0

0.1

0.2

0.3

0.4

0.5

0.6

Advantaged Disadvantaged

50

60

70

Probability

Conclusions – individual factors

• Occupational social class 25-30 years ago important for life chances

• Extremes of social position have clearly distinct life chances

• Social mobility has weak effect on original social class risk

• Social housing population in 1991 highly disadvantaged

• Accumulated social exclusion or multiple deprivation is accompanied by a greatly increased risk to health

• LA level classification inferior predictor to ward deprivation and SE residence

• Socioeconomic nature of ward is significant

• South East effect reflects career opportunities and economic prosperity

• Social mobility modifies mortality risk profoundly in deprived wards

Conclusions – area factors

4. Current developments in ONS

• Relocation (Lyons review)• ‘Efficiency savings’• Generic analytical divisions• Process re-engineering• Web-first dissemination

Working with academia

• Commissioned work• Research advisory groups• Collaborative research/access to ONS data

sources• CASE studentships• Public health training posts• Recruitment