Socioeconomic Inequalities in Health in 22 European Countries

Table 1. Table 1. Countries Included in the Analysis and Sources of Data.

We obtained data on mortality according to age, sex, cause of death, and indicators of socioeconomic status from mortality registries (Table 1). The data were based on 3.5 million deaths in 16 countries among more than 54 million persons ranging in age from 30 to 74 years at the beginning of the study. The data were drawn from national populations, except for the United Kingdom, with data from England and Wales only; Italy, with data from Turin only; and Spain, with data from Madrid, Barcelona, and the Basque country only. With regard to the mortality data from England and Wales, this article has received clearance from the Office for National Statistics Longitudinal Study (reference number 20037C). We performed two analyses of the data on death according to cause; one analysis focused on common causes of death (cancer, cardiovascular disease, and injuries), and the other focused on more specific causes of death (smoking-related causes, alcohol-related causes, and causes amenable to medical intervention, such as tuberculosis and hypertension^13,14). Code numbers of the causes of death according to the ninth and tenth revisions of the International Classification of Diseases, Clinical Modification (ICD-9-CM and ICD-10-CM) are given in Table 1 in the Supplementary Appendix, available with the full text of this article at www.nejm.org.

Data on self-assessed health and risk factors for disease (e.g., smoking and obesity) according to age, sex, and indicators of socioeconomic status were obtained from national health or multipurpose surveys that also included self-reported socioeconomic data (Table 1). The data came from 19 countries and almost 350,000 respondents who ranged in age from 30 to 64 years in some surveys and from 30 to 69 years in others. All data are nationally representative. For self-reported illness, our study focused on the single-item question on self-assessed health (“How is your health in general?”), which has five possible answers, ranging from “very good” to “bad.” In order to make use of the full range of levels of self-assessed health, we gave quantitative weights to each level (i.e., a multiplicative factor of 1.85 for each level worse than “very good”) that were derived from the average number of chronic conditions in each level¹⁵ (details of the calculation are given in the legend to Figure 2). The only risk factors for disease for which data were available in a form that enabled them to be compared across countries were current tobacco smoking and obesity, defined as a body-mass index (the weight in kilograms divided by the square of the height in meters) greater than 30.

Socioeconomic status was measured by education, occupation, and income. Education levels were categorized as no education or primary education (up to approximately 6 years of education), lower secondary education (up to approximately 9 years), higher secondary education (up to approximately 11 years), and tertiary education (bachelor's degree or higher). Data on education level were available in a comparable form for most countries from both mortality registries and health interviews or multipurpose surveys. Occupations were classified as “manual” (considered the lower level) or “nonmanual.” Data on occupation were available from mortality registries for middle-aged men in a limited number of countries only. Income was categorized in approximate quintiles of equivalent net household income. The self-reported after-tax incomes of all household members, including benefits, were added, and the total was corrected for household size by dividing it by the total number of persons in the household to the power of 0.36. Income data were available from surveys in a limited number of countries only. Tables 2, 3, and 4 in the Supplementary Appendix show the distribution of study populations according to education level, occupational classification, and income level. The proportion of the population with less education tended to be large in the southern and eastern regions, whereas inequalities in income were large in England and Wales and in Portugal.

All measures were adjusted for age. Because both relative and absolute measures of inequalities in health are important, we have presented both the relative index of inequality and the slope index of inequality^16,17 for each country separately. Both indexes are regression-based measures that take into account the whole socioeconomic distribution and that remove variability in the size of socioeconomic groups as a source of variation in the magnitude of inequalities in health.¹⁷ In the regression analysis, mortality, morbidity, or risk-factor prevalence was related to a measure of the rank of education, occupation, or income, in which the rank was calculated as the mean proportion of the population having a higher level of education, occupation, or income.

The relative index of inequality is the ratio between the estimated mortality, morbidity, or risk-factor prevalence among persons at rank 1 (the lowest education, occupation, or income level) and rank 0 (the highest level). The relative index of inequality was calculated with the use of Poisson regression analysis, which also generated 95% confidence intervals. The slope index of inequality measures absolute differences in rates (e.g., in deaths per 100,000 person-years) between the lowest and the highest ends of the socioeconomic scale. The slope index of inequality is derived from the relative index of inequality and the age-adjusted overall mortality rate according to the following formula: slope index of inequality=2×mortality rate×(relative index of inequality−1)÷(relative index of inequality+1).¹⁶ Because the slope index of inequality depends on the overall mortality rate in the population, we have presented these overall mortality rates together with the slope indexes of inequality.

Figure 1. Figure 1. Relative Inequalities in the Rate of Death from Any Cause.

Panel A shows inequalities between men with the lowest level of education and those with the highest, and Panel B shows education-related inequalities for women. Panel C shows inequalities between men in the lower and higher occupational classes. Economically inactive men whose last occupation was unknown were excluded from the analysis. Because exclusion of these men may lead to underestimation of mortality differences between occupational classes, we applied an adjustment procedure that was developed and tested in a previous European comparative study of inequalities in mortality; the procedure is based on national estimates of the proportion of economically inactive men in each occupational class and of the mortality rate ratio of inactive as compared with active men in each occupational class.¹⁸

Figure 1A and 1B show relative inequalities in the rate of death from any cause according to education level. The relative index of inequality is greater than 1 for both men and women in all countries, indicating that, throughout Europe, mortality is higher among those with less education. The magnitude of these inequalities varies substantially among countries. For example, in Sweden, the relative index of inequality for men is less than 2, indicating that mortality among those with the least education is less than twice that among those with the most education; on the other hand, in Hungary, the Czech Republic, and Poland, the relative index of inequality for men is 4 or higher, indicating that mortality differs by a factor of more than 4 between the lower and upper ends of the education scale. The smallest inequalities for both men and women are found in the Basque country of Spain, whereas the largest inequalities are found in the Czech Republic and Lithuania. Education-related inequalities in mortality are smaller than the average for Europe in all southern European populations included in this analysis and larger than average in most countries in the eastern and Baltic regions. Data on occupation-related inequalities in mortality among middle-aged men (Figure 1C) confirm that relative inequalities in mortality tend to be smaller in southern European populations.

Table 2. Table 2. Absolute Inequalities in Overall and Cause-Specific Mortality Rates between Persons with the Lowest and Those with the Highest Level of Education.

Table 2 shows that the international pattern observed for relative education-related inequalities in mortality also generally applies to absolute education-related inequalities in mortality, as indicated by the slope index of inequality. In Europe as a whole, persons with less education have higher rates of death from all causes except breast cancer, as indicated by a negative slope index of inequality for this cause of death. Inequalities in the rate of death from cardiovascular disease account for 34% of education-related inequalities in the rate of death from any cause among men (451 of 1333 deaths per 100,000 person-years) and 51% of those among women (251 of 492 deaths per 100,000 person-years). Although death from almost any cause is more frequent among those with less education than among those with more education, the range of variation for a single cause of death sometimes includes both “reverse” inequalities (higher mortality in groups with higher education) and “regular” inequalities (higher mortality in groups with lower education).

These data help to explain how smaller education-related inequalities in the rate of death from any cause in southern European populations and larger inequalities in the eastern and Baltic regions arise. Among men and women, smaller inequalities in the rate of death from any cause in the southern region are due mainly to smaller inequalities in the rate of death from cardiovascular disease. For example, among men in the Basque country, where the education-related inequality in the rate of death from any cause is below the European average, death from cardiovascular disease accounts for 46% of this difference (i.e., [451−16 deaths per 100,000 person-years]÷[1333−384 deaths per 100,000 person-years]). Larger inequalities in the rate of death from cardiovascular disease make an important contribution to larger inequalities in the rate of death from any cause in the eastern and Baltic regions as well; however, important contributions are also made by cancer in the eastern region and injuries in the Baltic region.

In Europe as a whole, inequalities in mortality from smoking-related conditions account for 22% of the inequalities in the rate of death from any cause among men and 6% of those among women (Table 2). Inequalities in smoking-related mortality tend to be larger in the eastern and Baltic regions (among men only) and smaller (or even “reverse”) in the southern region. In Europe as a whole, inequalities in alcohol-related mortality account for 11% of inequalities in the rate of death from any cause among men and 6% of those among women. Larger inequalities in alcohol-related mortality contribute to larger inequalities in the rate of death from any cause in Hungary (among men and women) and the Baltic region (among men only). In Europe as a whole, deaths from conditions amenable to medical intervention account for 5% of inequalities in the rate of death from any cause. However, these inequalities are larger than the European average in Lithuania and Estonia, where they contribute to the larger inequalities in the rate of death from any cause (among men only).

Figure 2. Figure 2. Relative Inequalities in the Prevalence of Poorer Self-Assessed Health.

Panels A and B show inequalities between persons with the lowest and those with the highest level of education for men and women, respectively. Panels C and D show inequalities between persons with the lowest and those with the highest level of income for men and women, respectively. In order to make use of the full range of levels of self-assessed health, we calculated the estimated burden of disease associated with each level on the basis of the number of chronic conditions reported by respondents to these surveys. Relative differences in self-reported chronic conditions between answer categories of the self-assessed health question were remarkably similar between countries and varied only marginally around a multiplicative factor of 1.85 (i.e., each step down on the self-assessed health scale was found to be associated with 1.85 times more chronic conditions). On the basis of this analysis, we assigned a weight for burden of disease to each category of answer to the question “How is your health in general?” “Very good” was assigned a weight of 1.85⁰=1, “good” a weight of 1.85¹=1.85, “fair” a weight of 1.85²=3.42, and “poor” or “very poor” a weight of 1.85³=6.33. Sensitivity analyses showed that the ranking of countries according to the magnitude of inequalities in self-assessed health did not change when these weights were varied within the range of observed values.¹⁵

Figure 2 shows the relative inequalities in the prevalence of poorer self-assessed health (weighted on the basis of the burden of chronic disease) according to education and income level. The relative index of inequality is greater than 1 in all countries, indicating worse health in groups of lower socioeconomic status throughout Europe. The variation of this measure among countries is considerably less than that of inequalities in the rate of death from any cause, and the international pattern also tends to be different from that of death from any cause. In Italy and Spain, education-related inequalities in self-assessed health are smaller than average, a finding that mirrors the smaller education-related inequalities in the rate of death from any cause observed in Turin, Barcelona, Madrid, and the Basque country. In the Baltic region, on the other hand, education-related inequalities in self-assessed health are smaller than average, whereas education-related inequalities in death from any cause are larger. Income-related inequalities in self-assessed health are not larger in the eastern and Baltic regions than in other parts of Europe and are remarkably large in the northern and western regions, particularly England and Wales, where income inequalities are also large (see Table 4 in the Supplementary Appendix).

Figure 3. Figure 3. Relative Inequalities in the Prevalence of Current Smoking (Panel A) and Obesity (Panel B) between Persons with the Lowest and Those with the Highest Level of Education, According to Sex.

In Europe as a whole, both smoking and obesity are more common among people of lower education level; education-related inequalities in smoking are larger among men, and education-related inequalities in obesity are larger among women (Figure 3). There are striking differences among countries in the magnitude and even the direction of these inequalities, however. Large education-related inequalities in smoking are seen in the northern, western, and continental regions; small inequalities (and, among women, even reverse inequalities, in which smoking rates are higher in groups with more education) are seen in the southern region. In the eastern and Baltic regions, the pattern is unclear. Large education-related inequalities in obesity are seen in the southern region, particularly among women, for whom the relative indexes of inequality are above 4, indicating that the prevalence of obesity among those with the least education is more than four times higher than that among those with the most education. By contrast, education-related inequalities in obesity tend to be smaller than average in the eastern and Baltic regions.

As compared with our study of inequalities in mortality and morbidity related to socioeconomic status in 10 western European countries during the 1980s,³ the present, more extensive study of the situation during the 1990s and early 2000s found much larger among-country variability in the magnitude of inequalities in health. Inequalities in mortality from selected causes suggest that some variations may be attributable to socioeconomic differences in smoking, excessive alcohol consumption, and access to health care. We also found among-country variations in the magnitude of inequalities in self-assessed health, but in a different pattern, precluding a generalization from inequalities in mortality to inequalities in overall health.

Our study had several limitations. International comparability of data on socioeconomic inequalities in health is still imperfect, and the degree of comparability is likely to decline with increasing geographical coverage. There are differences among countries in various aspects of data collection, and some of these might affect the size of inequalities in health, as we have shown previously.¹⁸ We found smaller inequalities in mortality in some urban, relatively prosperous southern European populations that are not necessarily representative of the whole of Italy or Spain. Some studies have shown, however, that inequalities in health tend to be larger in urban than in rural areas.¹⁹ Our previous study in the 1980s, which used national data for Italy and Spain from methodologically less-refined sources, also showed smaller inequalities in mortality in these countries.^4,5 We found larger inequalities in mortality in the eastern and Baltic regions. All these countries except Slovenia, which has smaller inequalities in mortality, provided data from cross-sectional, non-census–linked studies. Although this may suggest bias,²⁰ it is also possible that Slovenia, which is close to Italy, shares some of the favorable characteristics of the southern region.

Internationally comparable data on inequalities in specific determinants of mortality and morbidity are scarce, and we could study only smoking and obesity. The contribution to inequality of other factors, such as alcohol consumption, use of health care, working and housing conditions, and psychosocial stressors, could not be studied directly.

Both smoking and obesity have been shown to contribute to inequalities in health related to socioeconomic status in studies of individual persons in some countries.^21-23 Obesity, however, is unlikely to be a major contributor to international variations in inequalities in health, because inequalities in obesity related to socioeconomic status are large where inequalities in mortality related to socioeconomic status, particularly mortality from cardiovascular disease, are small (i.e., in the southern region). Smoking, on the other hand, does appear to be a major explanatory factor. It has been well documented that countries in the southern region are in an earlier stage of the smoking epidemic than countries in the northern, western, and continental regions.^24,25 We still found reverse inequalities in smoking among women and small inequalities among men, findings that are consistent with the smaller inequalities in mortality in the southern region, particularly from conditions related to smoking. The history of the smoking epidemic is much less well documented for the eastern and Baltic regions,^26,27 and it is therefore difficult to determine why inequalities in mortality from smoking-related conditions are large, whereas inequalities in smoking are often small.

The role of hazardous drinking (daily consumption of large amounts of alcohol-containing beverages, binge drinking, or consumption of surrogate alcohols) in generating high mortality rates in eastern Europe, particularly among men, has been well documented.^28-30 We have not been able to find comparable survey data on inequalities in alcohol consumption related to socioeconomic status in eastern Europe, but our analysis of cause-specific mortality suggests that rates of hazardous drinking are substantially higher in the lower than in the higher socioeconomic groups, particularly among men. Low levels of social support, lack of control over one's life, and material hardship, combined with a culture that approves of excessive alcohol consumption, are likely to be involved.^8,9

Although the role of deficiencies in health care in the high mortality rates of eastern Europe has been pointed out before,^31,32 our study demonstrates the magnitude of inequalities in mortality related to socioeconomic status from conditions amenable to medical intervention in this part of Europe. Our results suggest that inequalities in access to good-quality health care have a role in generating inequalities in mortality. Inequalities in access to health care leading to inequalities in survival from chronic conditions may also partly explain the discrepancy between our results for mortality and those for self-assessed health. Inequalities in the prevalence of poorer self-assessed health are the result of inequalities in both the incidence and the duration of health problems, which may be shortened by lower survival rates among less-educated persons in eastern Europe.

Smoking, obesity, excessive alcohol consumption, and deficiencies in health care represent only some of the immediate determinants of inequalities in health, and both lifestyle choices and patterns of use of health care are likely to be constrained by inequalities in general living conditions, as structured by political, economic, social, and cultural forces. Within western Europe, there is little evidence that among-country variations in the magnitude of inequalities in health are related to variations in political factors. For example, Italy and Spain have welfare policies that are less generous and less universal than those of northern Europe,^33,34 but they appear to have substantially smaller inequalities in mortality, perhaps partly because of cultural factors, such as the Mediterranean diet and the reluctance of women to take up smoking.^35,36 Cultural factors seem to have prevented differences in access to material and other resources in these populations from translating into inequalities in lifestyle-related risk factors for mortality.

We also found no evidence for systematically smaller inequalities in health in countries in northern Europe. This is surprising, because these countries have long histories of egalitarian policies, reflected by, among other things, welfare policies. These policies provide a high level of social-security protection to all residents of the country, resulting in smaller income inequalities and lower poverty rates.^33,34,37 Our results suggest that although a reasonable level of social security and public services may be a necessary condition for smaller inequalities in health, it is not sufficient. Lifestyle-related risk factors have an important role in premature death in high-income countries³⁸ and also appear to contribute to the persistence of inequalities in mortality in the northern region.³⁹

Our study shows that although inequalities in health associated with socioeconomic status are present everywhere, their magnitude is highly variable, particularly for inequalities in mortality. This result implies that there is opportunity to reduce inequalities in mortality. Developing policies and interventions that effectively target the structural and immediate determinants of inequalities in health is an urgent priority for public health research.⁴⁰

In addition to the authors, the following members of the European Union Working Group on Socioeconomic Inequalities in Health participated in this study: Scientific Institute of Public Health, Brussels — H. van Oyen, S. Demarest; Department of Demography and Geography, Faculty of Science, Charles University in Prague, Prague, Czech Republic — J. Rychtarikova; Department of Social Geography and Regional Development, Faculty of Science, Charles University in Prague, Prague, Czech Republic — D. Dzurova; National Institute of Public Health, Copenhagen — O. Andersen; National Institute of Public Health, University of Southern Denmark, Copenhagen — O. Ekholm; School for Health, University of Bath, Bath, England — K. Judge; National Institute for Health Development, Department of Epidemiology and Biostatistics, Tallinn, Estonia — M. Tekkel; Department of Health Promotion and Chronic Disease Prevention, National Public Health Institute, Helsinki — R. Prättäla; Department of Sociology, University of Helsinki, Helsinki — P. Martikainen; Institut National de la Statistique et des Études Économiques, Paris — G. Desplanques; Research and Information Institute for Health Economics, Paris — F. Jusot; Center for Social Policy Research, University of Bremen, Bremen, Germany — U. Helmert; Demographic Research Institute, Hungarian Central Statistical Office, Budapest, Hungary — K. Kovacs; Hungarian National Center of Epidemiology, Budapest, Hungary — F. Marton; Economic and Social Research Institute, Dublin — R. Layte; Department of Public Health, University of Turin, Turin, Italy — G. Costa; Servizio di Epidemiologia, Grugliasco, Italy — F. Vannoni; Faculty of Public Health, Riga Stradins University, Riga, Latvia — A. Villerusa; Kaunas University of Medicine, Kaunas, Lithuania — R. Kalediene, J. Klumbiene; Centraal Bureau voor de Statistiek, Voorburg, the Netherlands — J.J.M. Geurts; Research Program Care, Health and Welfare, Oslo University College, Oslo — E. Dahl; Division of Epidemiology, Norwegian Institute of Public Health, Oslo — B.H. Strand; Department of Medical Statistics, National Institute of Hygiene, Warsaw, Poland — B. Wojtyniak; Centro de Estudos Geográficos, Universidade de Coimbra, Coimbra, Portugal — P. Santana; Košice Institute for Society and Health, Pavol Josef Safarik University, Košice, Slovakia — A. Madarasova Geckova; Department of Public Health, Faculty of Medicine, Ljubljana, Slovenia — B. Artnik; Agencia de Salut Pública de Barcelona, Barcelona — C. Borrell; Research Unit, Department of Health, Basque Government, Vitoria-Gasteiz, Spain — S. Esnaola; Department of Preventive Medicine and Public Health, Universidad Complutense de Madrid, Madrid — E. Regidor; Department of Public Health Sciences, Karolinska Institute, Stockholm — B. Burström; Center for Health Equity Studies Stockholm, Stockholm University, Stockholm — J. Fritzell, O. Lundberg; Institute of Social and Preventive Medicine, University of Zurich, Zurich, Switzerland — M. Bopp; Office of National Statistics, Newport, United Kingdom — M. Glickman.