Socioeconomic status has been operationalised in a variety of ways, most commonly as education, social class, or income. In this study, we also use occupational complexity and a SES-index as alternative measures of socioeconomic status. Studies show that in analyses of health inequalities in the general population, the choice of indicators influence the magnitude of the observed inequalities. Less is known about the influence of indicator choice in studies of older adults. The aim of this study is twofold: i) to analyse the impact of the choice of socioeconomic status indicator on the observed health inequalities among older adults, ii) to explore whether different indicators of socioeconomic status are independently associated with health in old age.
We combined data from two nationally representative Swedish surveys, providing more than 20 years of follow-up. Average marginal effects were estimated to compare the association between the five indicators of SES, and three late-life health outcomes: mobility limitations, limitations in activities of daily living (ADL), and psychological distress.
Our results indicate overlapping properties between socioeconomic status indicators in relation to late-life health. However, income is associated to late-life health independently of all other variables. Moreover, income did not perform substantially worse than the composite SES-index in capturing health variation. Thus, if the primary objective of including an indicator of socioeconomic status is to adjust the model for socioeconomic differences in late-life health rather than to analyse these inequalities per se, income may be the preferable indicator. If, on the other hand, the primary objective of a study is to analyse specific aspects of health inequalities, or the mechanisms that drive health inequalities, then the choice of indicator should be theoretically guided.
The overarching aim of this study was to explore how the three most common indicators of SES (education, social class, and income) are associated with health in old age. We also included occupational complexity as an alternative indicator of SES, as recent research suggests that complexity is a key driver of labour market stratification [8,9,10]. Education, social class, occupational complexity, and income all have overlapping properties, but they may also be independently associated with health in old age. Therefore, we explored the relation between these variables, a composite measure of the variables, and change in mobility limitations, activities of daily living, and psychological distress from working ages to old age.
Some research have suggested that skill requirements, productivity and efficiency may provide a better explanation of social stratification at the labour market than conventional class theories [8,9,10]. Using Swedish data, Tåhlin found that the occupational complexity level is more strongly associated to the earned wages than traditional social class schemas [9, 10]. Tåhlin further argues that the level of occupational complexity associated with a given occupation can be used as a proxy for efficiency and productivity. Against this background, we have included a measure of occupational complexity as an alternative indicator of SES.
Income is often considered to be a straightforward indicator of material resources, and income is, robustly and positively, associated with longevity [22, 23]. Income might affect the health of people by enabling those with high income to lead healthy lifestyles, while those at the lower rungs of the income distribution have fewer of these enabling resources . Income has an important role as an enabling resource for health care access and use in the Andersen health behavioural model [25, 26]. On the other hand, poor health can also result in lower incomes . Hence, the association between income and health is likely shaped by bi-directional causal mechanisms. In the literature, measures indicating financial situation is commonly most strongly associated to health and mortality in old age, when compared to e.g. educational level and social class [14,15,16].
The rationale for including specific indicators of SES in studies of health in old age may vary. It could be to monitor, or to understand, how the social patterning of resources affect health. In this case, it is important to consider the different pathways and mechanism that education, social class, income, and occupational complexity might have on health. Another reason could be to merely adjust a model for as much socioeconomic variance as possible, without any specific interest in the mechanisms driving the health inequalities. Research suggest that a composite measures of individual level indicators could be appropriate for this purpose . The findings from a previous study, based on the same data as the present study, suggested that a composite measure of individual level variables of SES was more suitable for this purpose than any of the individual variables . On the other hand, using a composite measure of individual level variables of SES, may obscure the underlying mechanisms and prevent progression in the understanding of how different aspects of SES contributes to health.
Studies of the working age population have showed that different indicators of SES are differently associated with different health outcomes, which suggest that different underlying mechanisms may generate these associations [5, 33]. These findings suggest that the choice of indicator may be of importance for the results and the interpretations, when studying socioeconomic inequalities in health. Less is known of how different SES indicators relate to health in old age.
A thorough exploration of how the association between SES and health in old age varies by indicator of SES may provide important insights into the mechanisms generating socioeconomic inequalities in late-life health. As most commonly used indictors of SES are rooted in educational and occupational stratification, we chose to assess these variables in late working life when people tend to have reached their peak positions, in terms of educational level, social class, and income.
We also used all of the SES indicators described above to construct a composite measure of SES (the SES-index). We used education, social class, occupational complexity, and income as classified above, and summarized them. The index was then divided into tertiles. This was done to investigate whether a composite measure could, statistically, capture as much, or more, of the variance in late-life health as the individual indicators.
The associations between education, social class, occupational complexity, income, the SES-index, and health in later life were all analysed separately in model 1 (Table 2), adjusted only for age, sex, and linkage. In model 2 (Table 2), all the SES indicators were analysed simultaneously in relation to each health outcome. We also adjusted the associations between each individual SES indicator and each health outcome, by the other indicators of SES one-by-one. For these results, see Additional file 1: Table S1 and Additional file 2: Table S2.
Education was significantly and negatively associated with both mobility limitations and psychological distress, but not with ADL limitations (Table 2; model 1). Including education increased the model fit by 12% in the analyses of mobility limitations, 3% for ADL limitations, and 15% for psychological distress. However, the inclusion of the variable did not significantly improve the model fit for ADL limitations and psychological distress. Only the high-educated group deviated significantly from the other groups, with better health. Adjusted for the other SES indicators (Table 2; model 2), education was no longer significantly associated with the health outcomes, and contributed with only 1% to the total model fit for mobility limitations, 0% for ADL limitations, and 3% for psychological distress (Table 2; model 2).
Income was statistically significantly associated with all outcomes. The high-income group consequently had the lowest probabilities of adverse health of all income groups. Income contributed to model fit by 13% for mobility limitations, 7% for ADL limitations, and 27% for psychological distress (Table 2). Income was the only indicator that remained statistically significantly associated with the health outcomes in the fully adjusted models (model 2). In addition, income contributed the most to the model fits in model 2: 7% to mobility limitations, 3% to ADL limitations, and 18% to psychological distress.
The high SES-index group had better health than the other groups. The SES-index increased model fit more than any of the other indicators in the models of mobility limitations, about the same as income for ADL limitations, and less than income for psychological distress.
Including all the indicators of SES simultaneously increased model fit for mobility limitations by 29%, compared to a model only adjusted for age, sex, and linkage. The individual contribution of each variable sum up to an increased model fit of 46% (education 12% + social class 12% + occupational complexity 9% + income 13%). The difference between 46 and 29% indicate that the properties of the socioeconomic indicators overlap. The composite measure of the variables (SES-index) explained less variance than simultaneously including all SES indicators. A similar pattern emerged for ADL limitations. The contribution, when including all SES indicators simultaneously was 12%, while the sum of the individual contributions amounted to 19%, and the SES-index only explained 6%. The same pattern holds when psychological distress is the outcome, the summed increase in model fit for all SES indicators was 72%, whereas the combined increase was 37%, and the SES-index increased model fit by 20%. These result suggests that properties of the indicators overlap when analysing psychological distress. 2b1af7f3a8