Article Text
PDF
Abstract
Background Unemployment and economic inactivity are associated with poor health. There are social gradients in unemployment and economic inactivity, so it was hypothesised that they may contribute to the social gradient in self-rated health.
Methods Data on employment status, socio-economic position (SEP) and self-rated heath were obtained for people of working age (25–59) who had ever worked from a 3% sample of the 2001 English census. The age-adjusted prevalence differences in poor general health for four separate measures of SEP were compared with the prevalence differences obtained after additional adjustment for employment status.
Results Prevalence differences for poor health were reduced by 50% or over when adjusting for employment status (for men ranging from 57% to 81%, for women 50% to 74%).
Discussion The social gradient in employment status contributes greatly to the social gradient in self-reported health. Understanding why this is the case could be important for tackling social inequalities in health.
- Employment
- social gradient
- health inequalities
- work
- self reported health
- self-rated health
- social inequalities
- unemployment and health
Statistics from Altmetric.com
- Employment
- social gradient
- health inequalities
- work
- self reported health
- self-rated health
- social inequalities
- unemployment and health
Background
Health varies by employment status, with unemployment and economic inactivity being associated with an increased likelihood of morbidity and mortality.1–8 The social gradients in the risk of unemployment and economic inactivity may contribute to the social gradient in health. So, using census data, this study asks ‘to what extent does adjusting for employment status reduce the social gradient in self-reported health among men and women in England’?
Methods
A 3% sample of the 2001 English census was used (the Individual Sample of Anonymised Records).9 The analysis sample was people of working age (25–59) excluding full-time students, those living in communal residences (eg, hospitals) and the 1.3% of men and 3.9% of women who had never worked. We excluded the never worked, as there is some evidence, particularly for men, that this group includes many people unable to work due to pre-existing health conditions.10 The final sample sizes were 349 699 women and 349 181 men. Data on self-reported health, employment status, socio-economic position (SEP) and age were extracted. The census form is available online (http://www.statistics.gov.uk/census2001/pdfs/H1.pdf). Self-rated general health was assessed in the census, for the first time in 2001, by the following question ‘Over the last 12 months, would you say your health has on the whole been … Good? Fairly good? Poor?’ The coding of employment status and four census measures of SEP are given in table 1. Income was not measured in the census.
Differences in prevalence of poor health general health for English men and women aged 25–59 in 2001
All analysis was conducted using Stata version 10.11 As men and women have different overall rates and patterns of employment, it was decided a priori to analyse each gender separately.
As the data are cross-sectional, we focused on the prevalence of poor health by SEP. We studied the prevalence difference in poor health to ascertain the contribution of employment status to the absolute rate of poor health for each SEP group. We fitted generalised linear models with a binomial distribution and an identity link to obtain the prevalence difference. In all analyses, the ‘highest’ SEP group is the reference. To judge the extent of inequalities by each separate measure of SEP (Models A in table 1) and by employment status (Model C) we fitted age-only-adjusted models. To assess the impact of controlling for employment status on the separate age-adjusted SEP models, we then additionally adjusted for employment status (Models B) and calculated the percentage reduction in prevalence difference in each category using the formula (prevalence difference in age adjusted model–prevalence difference in age and employment adjusted model)/(prevalence difference in age adjusted model). To assess the overall health gradient for the SEP measures (education and housing tenure) with more than two groups, we also calculated the slope index of inequality. It can be regarded as a measure of the health gradient from the highest to lowest group.12 For employment status (Model C), we additionally controlled for all measures of SEP (Model D) to assess their attenuating impact.
Results
The majority of men and women were in employment, although women's rate of employment was lower (table 1). For those out of work, economic inactivity was more common than unemployment. Overall, 7.6% of men and 8% of women reported poor general health. As expected in the age-only-adjusted analysis (Models A), the rates of poor health increased with lower SEP. In Models B, which additionally controlled for employment status, socio-economic differences were attenuated by 50% or over (for men ranging from 57% to 81%, for women 50% to 74%).
As an example, 5.6% of men living in owner-occupied housing had poor general health compared with 19.1% of men in social rented housing, an age-adjusted difference of 13% points. After further adjustment for employment status, this difference was reduced to 2.5% points, a reduction of 81%. The slope index of inequality for housing tenure, as an overall measure of SEP, was reduced by 81% after controlling for employment status.
Rates of poor health were higher among non-employed groups (table 1, Model C). Adjusting for all measures of SEP (table 1, Model D) attenuated the differences somewhat for those looking after the family or the home and the unemployed but made little impact for other groups.
Discussion
Our results suggest that for both men and women, employment status attenuates significantly SEP differences in poor health. These results suggest that the social gradient in employment status could be an important contributor to the social gradient in self-rated general health. Arber using survey data also reported an important role for non-employment in the association between social class and self-rated health.13 Our work extends this to a wider range of socio-economic variables and a more recent, larger and highly representative dataset as well as providing a quantification of the impact.
Why employment status may play such an important role in the relationship between SEP and self-rated health cannot be ascertained in this cross-sectional study. We can only suggest some reasons. The relationship may run from poor health to non-employment to low SEP. In other words, health selection may be important. It is notable that the highest rate of poor health was in those describing their economic inactivity as being due to permanent sickness and disability. However, education as an SEP measure may be more robust to health selection, as it is mainly achieved earlier in adult life.14
If employment status affects, directly or indirectly, self-rated health, it is therefore more likely to be through either poverty or differences in income (employment may better reflect income differences than our other measures of SEP) or differences in health behaviours (although the evidence is more equivocal on this) or through psychosocial mechanisms such as stress.5 Finally, employment status may capture very well socio-economic differences across the lifecourse better than any single measure of SEP at one single point of time.6
In conclusion, our results suggest that it will be important for tackling health inequalities to understand why the social gradient in employment status seems to contribute so much to the social gradient in health.
What is known on this subject
Previous studies have shown that unemployment and economic inactivity are associated with worse morbidity and mortality and that unemployment and economic inactivity are socially graded.
Unemployment and economic inactivity are therefore considered to be potentially very important social determinants of health which contribute greatly to the social gradient in health and to socio-economic health inequalities.
However, existing studies have tended to focus specifically on the health of the unemployed or inactive, not on the contribution of employment status to the social gradient in health.
What this study adds
Using data on self-reported health from the 2001 English Census, this study is the first to quantify the effects of employment status on the social gradient and socio-economic inequalities in health.
In keeping with existing research, the study found that regardless of socio-economic position, people experiencing unemployment or economic inactivity have worse self rated than those in employment.
In addition, the results demonstrate that employment status contributes greatly to the social gradient in self-reported health, with unemployment and economic inactivity contributing up to 81% of the excess in self-reported poor health among the lowest socio-economic groups.
Acknowledgments
The 2001 SARs are provided through the Cathy Marsh Centre for Census and Survey Research (University of Manchester), with the support of the ESRC and JISC. All tables containing Census data, and the results of analysis, are reproduced with the permission of the Controller of Her Majesty's Stationery Office and the Queen's Printer for Scotland.
References
Footnotes
Competing interests None
Provenance and peer review Not commissioned; externally peer reviewed.