Article Text
Abstract
Background Between the 1990s and 2000s, relative inequalities in all-cause mortality increased, whereas absolute inequalities decreased in many European countries. Whether similar trends can be observed for inequalities in other health outcomes is unknown. This paper aims to provide a comprehensive overview of trends in socioeconomic inequalities in self-assessed health (SAH) in Europe between 1990 and 2010.
Methods Data were obtained from nationally representative surveys from 17 European countries for the various years between 1990 and 2010. The age-standardised prevalence of less-than-good SAH was analysed by education and occupation among men and women aged 30–79 years. Socioeconomic inequalities were measured by means of absolute rate differences and relative rate ratios. Meta-analysis with random-effects models was used to examine the trends of inequalities.
Results We observed declining trends in the prevalence of less-than-good SAH in many countries, particularly in Southern and Eastern Europe and the Baltic states. In all countries, less-than-good SAH was more prevalent in lower educational and manual groups. For all countries together, absolute inequalities in SAH were mostly constant, whereas relative inequalities increased. Almost no country consistently experienced a significant decline in either absolute or relative inequalities.
Conclusions Trends in inequalities in SAH in Europe were generally less favourable than those found for inequalities in mortality, and there was generally no correspondence between the two when we compared the trends within countries. In order to develop policies or interventions that effectively reduce inequalities in SAH, a better understanding of the causes of these inequalities is needed.
- Health inequalities
- SELF-RATED HEALTH
- SOCIAL EPIDEMIOLOGY
Statistics from Altmetric.com
Introduction
Europe offers excellent opportunities for conducting between-country comparative research of socioeconomic inequalities in health, and this can help to identify successful strategies to reduce these inequalities.1 ,2 Between 1990 and 2010, many European countries have made the reduction of health inequalities an explicit aim of national health policies.3–5 A comprehensive overview of trends in health inequalities among a large set of European countries can show whether there are differences between countries in the extent to which a reduction of health inequalities has actually been achieved.
A recent study assessed trends in inequalities in premature mortality between the 1990s and 2000s.6 The study found that relative inequalities increased in most populations in Europe except for Southern Europe, while absolute inequalities decreased in many European countries. Whether these trends can be generalised to other domains of health, however, is unknown. One other health outcome for which trend data on inequalities is available is self-assessed health (SAH), which has been shown to be a reliable indicator of general health and well-being,7–9 and an independent predictor of mortality and survival10–12 that can be used in population health monitoring.
SAH is strongly associated to indicators of socioeconomic position in all countries with available data,1 ,13–16 but studies on trends in these inequalities often focus on one country or at most, on a small number of countries.17–26 One notable exception is a study covering 10 European countries between the 1980s and 1990s that showed a high degree of stability of these inequalities.27 A comprehensive overview of recent trends in SAH inequalities based on a larger set of European countries is still lacking.
This paper aims to provide such a comprehensive overview by analysing trends in socioeconomic inequalities in SAH among adults in 17 European countries between 1990 and 2010.
Methods
Data
We obtained nationally representative health surveys from 17 countries (table 1). All available and comparable surveys between 1990 and 2010 were used, which led to a different number of surveys for each country. The first and last observation years differed between countries, but all years were between 1990 and 2010. Data came from the same survey over time for most countries, except for the Netherlands, Austria and Italy. However, the chosen surveys within the three countries have a high comparability,28–31 and thus can be used for trend analysis. The age range used for most countries was 30–79 years. Younger respondents were excluded because many of them were still receiving full-time education. Older respondents were excluded to avoid the potential bias caused by the exclusion of institutionalised population in many surveys. As shown in table 1, some countries had upper age limits that were different from 79 years (England, Scotland, Ireland, Poland and Lithuania). Although these different age limits could influence the comparability of the level of inequalities in SAH between countries, the risk of bias was considered limited in our analysis of trends in inequalities over time within each country. The number of included respondents per year ranged from 1137 (Czech Republic, 1993) to 87 673 (Italy, 2000). The prevalence of less-than-good SAH ranged from 12% (Switzerland, 2002, 2007) to 70% (Czech Republic, 1993) among men, and ranged from 12% (Ireland, 2002) to 77% (Portugal, 1995–1996) among women.
To measure SAH, we used answers from a question which was framed similarly to “how is your health in general?” In all countries except for England (in which 3 answer categories were used, ie, “good/fairly good/not good”), five answer categories were distinguished, which were normally “very good/good/fair/bad/very bad”. The precise answer categories varied slightly in some countries, but the consistency over time was retained in all countries. We dichotomised the answers by collapsing those that reported a less-than-good SAH into one category.
Socioeconomic position was measured by education and occupation. Educational levels were recorded as the highest level of education completed or currently being attended by the respondent. It was harmonised on the basis of the International Standard Classification of Education (ISCED) and reclassified into three categories: levels 0–2 (no, primary or lower secondary education, considered “low-educated”), levels 3–4 (upper secondary and postsecondary non-tertiary education, considered “middle-educated”), levels 5–6 (tertiary education, considered “high-educated”). Occupational classes were classified as “manual” versus “non-manual”. Respondents who were economically inactive, and who could not be classified on the basis of their last or main occupation were classified as missing. For trends in education-related inequalities, we included all 17 available countries with 62 country-year observations. For trends in occupation-related inequalities, we included 16 countries with 53 country-year observations, where Belgium was excluded due to a large number of missing values for occupation. Some recent years for the Netherlands and Finland were also excluded as information on occupation was not available.
Statistical methods
The prevalence rates of less-than-good SAH were calculated by country, year, sex, and education or occupation, and age standardised to the European Standard Population32 using the direct standardisation method.
Inequalities were measured by means of absolute prevalence rate differences (RD) and relative prevalence rate ratios (RR) of low versus high level of education or manual versus non-manual occupation. A bootstrap procedure with 1000 iterations was used to calculate 95% CIs. We also calculated the slope index of inequality (SII) and relative index of inequality (RII) based on education, which took into account the distribution of the population by education.33 In order to facilitate the comparison between education-related and occupation-related inequalities, however, we mainly used the RD and RR as the inequality measures, and provide the results for the RII and SII in the online supplementary figure S1. Survey weights were available for some countries or years. Unweighted results are reported in the results section. Analyses with available weighting factors are reported in the online supplementary figures S2 and S3 as a sensitivity analysis.
To study the trends over time in each country and in the ensemble of countries as a whole, we employed meta-analysis with random-effects models, using the prevalence of less-than-good SAH, the RD and RR by education or occupation as the outcomes. In the analysis of the trends within each country, the year of data collection was used as the only moderator variable in the models. In the analysis of the trends in the ensemble of countries, we pooled all available observations and additionally added the country dummies into the regressions. In all models, the regression parameters for the year of data collection were taken as indicators of the linear time trends. The estimated country-specific linear time trends and their 95% CIs were plotted in the form of forest plots together with the estimate for all countries as a whole (the “average” in the forest plots). A reference vertical line representing no change over time was plotted. The I2 statistic measuring the proportion of total variability explained by heterogeneity and the p value testing the residual heterogeneity were reported. The meta-analysis was performed with the R (3.0.1.) package metafor.34
Results
Figure 1 shows the results of the meta-analysis for the trends in age-standardised prevalence of less-than-good SAH in the 17 European countries between 1990 and 2010. As shown in the forest plots, a statistically significant decline in the prevalence of less-than-good SAH was observed in a pooled analysis including all countries (represented by “average”), among males and females. These declines were mainly the result of significant declines in some, but not all countries, including Italy, Portugal, Czech Republic, Poland, Lithuania and Estonia.
To give an overview of the levels of and variation in inequalities among the 17 European countries, the absolute and relative inequalities in SAH by educational level (figure 2) are presented, using the most recent year for which data were available in each country. Absolute education-related inequalities in SAH, as measured by RD, were found in all countries and these ranged between 0.08 and 0.35, indicating between 8% and 35% points difference in less-than-good SAH between the low and high educated, with no clear pattern emerging between different regions in Europe. Relative inequalities as measured by RR were also present in all countries, and these ranged between 1.26 and 4.14. These were particularly high in Scotland, Ireland, Switzerland (among males only) and Austria. Inequalities in SAH by occupational class are presented in the online supplementary figure S4. Absolute and relative inequalities by occupation were observed in all countries. Again, no clear pattern was observed among different European regions.
Results from the meta-analysis for the trends in absolute and relative inequalities based on education are reported as forest plots (figure 3A, B). Pooling all countries together, we found no statistically significant trend in absolute educational inequalities among males. Also, no significant trend was observed within most countries, except for an increasing trend in inequalities in Denmark and Switzerland, and a declining trend in Italy. Among females, a significantly increasing trend was found in a pooled analysis of all countries and in England, Scotland, the Netherlands, Poland and Lithuania. The general picture became less favourable when relative inequalities were used. Pooling all countries together, we found significantly increasing trends in relative inequalities among males and females. Among males, significant increases in relative inequalities were detected in Switzerland, Austria and Poland. Among females, significant increases in relative inequalities were observed in England, the Netherlands, Switzerland, Portugal, Poland and Lithuania. No country showed a significantly decreasing trend in relative inequalities.
Results from the meta-analysis for trends in inequalities based on occupation are presented in figure 4A, B. For males, the trends resembled those seen for education, with relatively stable trends for absolute inequalities in a pooled analysis, as well as in most countries. The stable trends were also found among females in a pooled analysis and in most countries. Again, the general picture became less favourable when relative inequalities were studied. When we pooled all countries together, significant increases in relative inequalities were found among males and females, as well as in a number of separate countries. Significant decline in relative inequalities was only observed in England among males.
Discussion
Summary of findings
We observed declining trends in the prevalence of less-than-good SAH in many countries, particularly in Southern and Eastern Europe and the Baltic states. In all countries, less-than-good SAH was more prevalent in lower educational and manual groups. For all countries together, absolute inequalities in SAH were mostly constant, whereas relative inequalities increased. Almost no country consistently experienced a statistically significant decline in either absolute or relative inequalities.
Interpretation
For all countries together, the prevalence of less-than-good SAH declined during the study period, while socioeconomic inequalities (mainly relative inequalities) in SAH generally increased. This can perhaps be partly explained by the commonly observed negative association between the prevalence of health problems and the magnitude of relative inequalities, which results from the fact that relative declines in prevalence tend to be larger in higher socioeconomic groups because of their lower prevalence.35 However, this cannot explain the fact that among females absolute educational inequalities also increased in some countries.
We mainly used the RD and RR of low versus high education in order to facilitate the comparison to the RD and RR by occupation. However, measures like the RD and RR cannot take into account the distribution of the population by socioeconomic group.33 In many European countries, the proportion of individuals with a low level of education is decreasing over time.18 ,27 ,36 Thus, the unfavourable trends in inequalities measured by RD and RR might reflect the fact that the shrinking low-educated group is increasingly composed of people who have been socially marginalised.37 This is partly confirmed by the results based on the SII and RII by education, which adjust for these differences in population composition (see online supplementary figure S1). This analysis finds stable trends in inequalities in all countries among males, and significantly increasing trends in a few countries and in the ensemble of countries as a whole only among females.
Although we included more countries and adopted a new technique to assess the trends in inequalities, it is worthwhile to compare our findings to the previous ones. Discrepancies between the inequalities in SAH and inequalities in mortality were found when we compared the magnitude of the inequalities,6 which is consistent with existing findings.1 As for the trends in inequalities, our overall results about relative inequalities in SAH are consistent with those about relative inequalities in mortality between the 1990s and the 2000s,6 where increasing inequalities were found in many populations. However, on a country-specific basis, the two do not always correspond. Countries for which increasing relative inequalities in mortality and SAH were found include Switzerland and Lithuania, but Finland, Sweden, Belgium and Estonia showed stable relative inequalities in SAH and increasing relative inequalities in mortality. Reductions of absolute inequalities were commonly found for mortality (except for the Baltic states), but were not generally found for SAH. One potential explanation for the different trends in absolute inequalities is that the mortality among higher educated is reaching a level below which it is difficult to decline further without new breakthroughs in prevention or treatment,6 whereas this might be not true for SAH as it is a subjective self-reported measure of health. Our results suggest that the trends in inequalities seen for mortality cannot be generalised to other health outcomes such as SAH.
In agreement with other studies, socioeconomic inequalities in SAH were found to be persistent over time.17 ,18 ,20 ,22 ,24 ,25 Although the study periods were not exactly the same, our findings are generally consistent with those from previous studies using data from one country or a small number of countries, for example, stable inequalities in Nordic countries18 and increased educational inequalities among Dutch women.21 In an earlier study,27 stable trends in SAH inequalities between the 1980s and 1990s was found in many European countries. It is worthwhile to compare this to the recent trends found in our study. England and the Netherlands, where stable education-related inequalities were reported between the 1980s and 1990s, showed increasing trends—particularly among females—in the recent years covered by our study. This is disappointing against the background of the increasing awareness among policymakers of socioeconomic inequalities, and the implementation of policies to tackle inequalities. In contrast, significant widening of education-related inequalities in Spain and Italy was found between the 1980s and 1990s; however, this was not detected in our recent period. Stable trends in education-related inequalities in Finland and Sweden were consistently found in earlier and recent periods.
More generally, our results do not support the idea that countries with national policies to tackle health inequalities have fared better in terms of inequalities in SAH than countries without such policies. England is the first and only European country that pursued a systematic and well-resourced policy to reduce inequalities in health.38 As indicated above, inequalities in SAH have not narrowed down in England during this period except for the relative inequalities by occupation among males; in fact, education-related inequalities among females increased. Scotland also pursued a coordinated action plan to tackle health inequalities,39 but inequalities in SAH were rather stable and even increased among females during the study period. The Netherlands is another country that has had some national activities to reduce health inequalities,40 but while absolute occupation-related inequalities among males declined, no positive changes were seen for education-related inequalities. Among all the countries, Italy shows the most encouraging trend as it had a significantly deceasing trend in prevalence of less-than-good SAH, and declines in absolute inequalities by education. However, efforts to reduce inequalities have not been stronger in Italy than in other countries.4 Other structural developments may have undone the effects of policies on health inequalities such as the economic recessions in some European countries in the early 1990s and the late 2000s,41 increases in income inequalities (eg, England),20 ,42 and changes of working conditions (eg, more work-related stress).21 Nevertheless, the observed trends of inequalities in SAH are not consistent with the amount of effort made to reduce health inequalities in some countries.
Strengths and limitations
This is the largest study ever of inequalities in SAH in terms of the number of countries and years included. It is also for the first time that a meta-analysis has been used to systematically assess the trends in inequalities. Two indicators (education and occupation) were used to capture the multidimensional nature of the concept of socioeconomic position, and we included absolute and relative inequalities.
One issue to consider in all international comparisons is data comparability. Despite great harmonisation efforts, we were not able to remove all differences between countries in data collection such as the framing of survey questions, the population coverage or the response rates. The tendency to report less-than-good SAH may differ between countries due to different cultural backgrounds,9 ,43 for example, persons in Central and Eastern European countries may tend to report their health as less good than persons in other European countries.44 ,45 However, as we focused on the trends in inequalities in SAH, we retained comparability over time within each country, and therefore consider the risk of bias due to between-country variations in these aspects to be limited. Weighting factors were available for some countries or years as the aim was to compensate the survey design and make the sample representative of the population. Essentially similar results were obtained in a sensitivity analysis of trends in RD and RR, where weighting factors were incorporated when available (see online supplementary figures S2 and S3).
Another concern is that the populations used to assess education-related and occupation-related inequalities differed. When we assessed the occupation-related inequalities, respondents who were economically inactive and who could not be classified on the basis of their last or main occupation were coded as missing, whereas they were included when education-related inequalities were assessed. This resulted in a larger percentage of missing values among people older than 65 years in occupation-based analysis (eg, in Denmark and Austria), than in education-based analysis. It may have resulted in smaller occupation-related inequalities, since economically inactive people tend to have worse health as compared with employed people, and tend to originate from the lower occupational groups. Again, the impact on the comparison between trends in occupation-related and education-related inequalities is likely to be limited.
The broad age range used in the analysis might hide the potential heterogeneity in the trends in inequalities in SAH among different age groups. Therefore, we did a supplementary analysis using a smaller age range of 30–64 years, which could also facilitate the comparison between trends in occupation-related and education-related inequalities. We found that limiting the analysis to the age group 30–64 years did not essentially change our results (see online supplementary figures S5 and S6).
Our meta-analysis assumed a linear trend in the outcome measures—an assumption that might not always hold. It cannot be excluded that results based on a non-linear trend assumption would have changed our conclusions. Nevertheless, this technique gives a useful “helicopter view” of the trends in inequalities in SAH in Europe. Future hypothesis-driven research should assess whether non-linear trends better fit the data.
Owing to lack of appropriate data, our analysis was mainly focused on recent trends in inequalities in SAH among adults. Future research should consider exploring the trends in SAH inequalities among adolescents.
Conclusions
For all countries together, relative socioeconomic inequalities in less-than-good SAH widened, whereas absolute inequalities were more stable. Trends in inequalities in SAH in Europe were generally less favourable than those found for inequalities in mortality over the same time period, and there was generally no correspondence between the two when we compared the trends within countries. In order to develop policies or interventions that effectively reduce inequalities in SAH, a better understanding of the causes of these inequalities is needed.
What is already known on this subject
Between the 1990s and 2000s, relative inequalities in all-cause mortality increased, whereas absolute inequalities decreased in many European countries. Whether these trends can be generalised to inequalities in self-assessed health is unknown.
Trends in inequalities in self-assessed health showed a high degree of stability in many European countries in the 1980s and 1990s. However, a comprehensive overview of recent trends in inequalities in self-assessed health in Europe is still lacking.
What this study adds
This paper provides an overview of trends in inequalities in self-assessed health in 17 European countries between 1990 and 2010.
For all countries together, absolute inequalities in self-assessed health were mostly constant, whereas relative inequalities increased. Almost no country consistently experienced a significant decline in either absolute or relative inequalities.
Trends in inequalities in self-assessed health were less favourable than those found for inequalities in mortality; on a country-specific basis, there was no correspondence between the two.
Acknowledgments
The authors thank Giuseppe Costa (University of Turin), Stefaan Demarest (Scientific Institute of Public Health, Belgium), Chris Dibben (University of St Andrews, Scotland), Satu Helakorpi and Anni Helldán (the Finnish National Institute for Health and Welfare THL), Ken Judge (University of Bath, England), Johannes Klotz (Statistics Austria, Austria), Richard Layte (Economic and Social Research Institute, Ireland), and Miroslawa Wolowicz (Central Statistical Office of Poland) for providing the data.
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
- Data supplement 1 - Online supplement
Footnotes
Contributors JPM had the original idea for the study. MB, BOB, DD, OE, JK, EL, ML, ER, PS and RdG made substantial contributions to the data acquisition. YH, CWNL and GJB performed the data analyses. YH, FJvL and JPM made substantial contributions to the drafting of the manuscript. All the authors contributed to the interpretation of the results, and reviewed and approved the final manuscript.
Funding This study was supported by the project DEMETRIQ (Developing Methodologies to Reduce Inequalities in the Determinants of Health, grant agreement number 278 511), which received funding from the European Union under the FP7 Health programme.
Competing interests None declared.
Provenance and peer review Not commissioned; externally peer reviewed.