Does the pattern of occupational class inequalities in self-reported health depend on the choice of survey? A comparative analysis of four surveys and 35 European countries

Background: Despite increasing overall life expectancy, substantial differences in health between socioeconomic groups persist. Research on inequalities in health often draws on data from different, single surveys. An important question that arises is whether these surveys reflect health and inequalities in the same way. When occupational class is utilized, data are often not analysed for women. The aim of this study therefore is to investigate whether patterns of occupational class inequalities in self-reported health differ across sex and country, between four major European surveys. Methods: Data on self-reported health and occupational class are taken from the European Social Survey (ESS), the EU Statistics on Income and Living Conditions (EU-SILC), the European Working Conditions Survey (EWCS) and the International Social Survey Programme (ISSP). Data from 35 countries for men and women aged 25-65 years are analysed. Occupational class is measured according to manual and non-manual workers. Age-standardized prevalence rates, and prevalence ratios (PR) between nonmanual and manual workers and likelihood ratio (LR) tests are estimated to determine occupational class inequalities in self-rated health in Europe. Results: Results show that prevalence rates of less than good health differ noticeably between countries and surveys. Furthermore, occupational class inequalities in health differ between countries. In some countries inequalities are larger for women than for men. This is especially true in Eastern, Central and Baltic European countries. Besides that no regional patterns, consistent over all surveys, in inequalities could be detected. Inequalities differed significantly between surveys. Conclusion: The magnitude of inequalities in all countries depend on the survey used in the analysis. When undertaking a comparative analysis of inequalities in health, or other determinants, these differences have to be taken into account, as results might differ according to the data source used.