An assessment of the effect of Health Selection bias into employment status at 1991 in the analysis of mortality from major causes by National Statistics Socioeconomic Classification 90 (NSSEC90) 1991-2001 and 1996-2001

Myer Glickman, Chris White and Tania Corbin, Office for National Statistics

[Project number 20105]

This LS application contributes to a wider project to monitor patterns and trends in mortality by NSSEC, based on all death registrations and using the 2001 Census and subsequent annual population estimates (modified by the LFS) as denominators.

Health selection into employment status at cross-section is an important alternative explanation for inequality in mortality risk by socioeconomic status. Traditional mortality analysis by social class excluded deaths occurring in the first five years of follow-up to allow such health selection effects to wear off. However, we wish to formally test the assumption of health selection effects with the new National Statistics Socioeconomic classification (NSSEC90) assigned from SOC90 and employment status at 1991. We will compare inequalities in major causes of death (i.e. ischaemic heart disease, all malignant neoplasms and all respiratory diseases) in the time periods 1991-2001 and 1996-2001 for men aged 20-64 and women aged 20-59 in 1991 and assess statistical variation.

Social class has been superseded by the NSSEC in the 2001 Census and it is constructive to determine a mortality baseline for this new classification. NSSEC is designed to be more relevant than Social Class to modern employment and social conditions; importantly, it distinguishes the self-employed in a way Social Class does not; and it is intended to be the basis for future analyses by socioeconomic status. Analyses undertaken by Fitzpatrick and Dollamore demonstrate the mortality pattern in a surrogate NSSEC measure (SECLOW), derived from SOC90 and employment status at the 1991 Census for the period 1991-93, deviates from the established Social Class pattern, especially with regard to the lower mortality found in the class Small Employers and Own Account Workers. It is relevant to detect the inequality pattern by NSSEC for the period 2001-04 and how it differs to Social Class inequality estimates in the 1990s for pensions policy development.

With the exception of decennial publications, all past analyses of mortality by socioeconomic status for England and Wales are based on the ONS Longitudinal Study (LS). This study aims to investigate the validity of using death registrations and a combination of census denominators and synthetically derived mid-year NSSEC population estimates for examining mortality differentials in the intercensal years. A key weakness postulated in the use of these sources is health selection into employment status at cross-section as a valid alternative explanation for mortality differentials by socioeconomic status. The LS's unique linkage over time of census and vital events data allows a comparison of the population at risk between different time periods, and thereby the potential scale of bias from this source in cross-sectional analyses of mortality and socioeconomic status.