Trends in socioeconomic inequalities in self ... - Semantic Scholar

Published by Oxford University Press on behalf of the International Epidemiological Association © The Author 2004; all rights reserved. Advance Access publication 24 November 2004

International Journal of Epidemiology 2005;34:295–305 doi:10.1093/ije/dyh342

Trends in socioeconomic inequalities in self-assessed health in 10 European countries Anton E Kunst,1* Vivian Bos,1 Eero Lahelma,2 Mel Bartley,3 Inge Lissau,4 Enrique Regidor,5 Andreas Mielck,6 Mario Cardano,7 Jetty AA Dalstra,1 José JM Geurts,8 Uwe Helmert,9 Carin Lennartsson,10 Jorun Ramm,11 Teresa Spadea,12 Willibald J Stronegger13 and Johan P Mackenbach1

Accepted

24 August 2004

Background Changes over time in inequalities in self-reported health are studied for increasingly more countries, but a comprehensive overview encompassing several countries is still lacking. The general aim of this article is to determine whether inequalities in self-assessed health in 10 European countries showed a general tendency either to increase or to decrease between the 1980s and the 1990s and whether trends varied among countries. Methods

Data were obtained from nationally representative interview surveys held in Finland, Sweden, Norway, Denmark, England, The Netherlands, West Germany, Austria, Italy, and Spain. The proportion of respondents with self-assessed health less than ‘good’ was measured in relation to educational level and income level. Inequalities were measured by means of age-standardized prevalence rates and odds ratios (ORs).

Results

Socioeconomic inequalities in self-assessed health showed a high degree of stability in European countries. For all countries together, the ORs comparing low with high educational levels remained stable for men (2.61 in the 1980s and 2.54 in the 1990s) but increased slightly for women (from 2.48 to 2.70). The ORs comparing extreme income quintiles increased from 3.13 to 3.37 for men and from 2.43 to 2.86 for women. Increases could be demonstrated most clearly for Italian and Spanish men and women, and for Dutch women, whereas inequalities in health in the Nordic countries showed no tendency to increase.

Conclusions The results underscore the persistent nature of socioeconomic inequalities in health in modern societies. The relatively favourable trends in the Nordic countries suggest that these countries’ welfare states were able to buffer many of the adverse effects of economic crises on the health of disadvantaged groups. Keywords

Educational level, poverty, income, socioeconomic status, socioeconomic factors, inequalities, self-assessed health, health surveys, trends, international perspectives, European Union

1 Department of Public Health, Erasmus MC, Rotterdam, The Netherlands. 2 Department of Public Health, University of Helsinki, Helsinki, Finland. 3 Department of Epidemiology and Public Health, University College,

London, United Kingdom.

9 Centre for Social Policy Research, University of Bremen, Bremen, Germany. 10 Swedish Institute for Social Research, Stockholm University, Stockholm,

Sweden. 11 Health Division, Statistics Norway, Oslo, Norway. 12 Department of Public Health and Microbiology, Turin University,

4 National Institute of Public Health, Copenhagen, Denmark. 5 Department of Epidemiology, Ministry of Health, Madrid, Spain. 6 GSF Institute of Health Economics and Health Care Management,

Neuherberg, Germany.

Turin, Italy. 13 Institute of Social Medicine and Epidemiology, University of Graz, Graz,

7 Department of Social Research, University of Piemonte Orientale,

Allessandria, Italy. 8 Division KPE, Statistics Netherlands, Heerlen, The Netherlands.

295

Austria. * Corresponding author. Department of Public Health, Erasmus MC, University Medical Centre Rotterdam, PO Box 1738, 3000 DR Rotterdam, The Netherlands. E-mail: [email protected]

296

INTERNATIONAL JOURNAL OF EPIDEMIOLOGY

Ever since the Black Report,1 a key question in health inequalities research has been whether inequalities in mortality and morbidity are narrowing or widening over time. Many trend studies are motivated by the concern that socioeconomic inequalities in health have increased owing to unfavourable social and economic developments.2 During the 1980s and 1990s, several European countries experienced serious economic crises or widening income inequalities—changes that might affect the health of national populations as a whole, and of disadvantaged groups in particular.3–5 Reducing inequalities in health now figures high on the health agendas of many national governments,6 and targets have been set to reduce these inequalities in the near future.7–9 In order to evaluate progress towards reaching these targets, monitoring trends in health inequalities will remain important. Until recently, trends in inequalities in health have been documented mostly by using data on differences in premature mortality according to occupational class or educational level. A widening of relative inequalities in mortality was observed consistently for all European countries for which data were available.5,10–18 This widening appeared to be a long-term trend that can be traced back to 1950s for both England and Wales and The Netherlands,19 and to the 1960s and 1970s for other northern European countries.11,16–18,20 A recent overview showed that these trends continued until the 1990s in all northern European countries.21 However, the evidence from southern European countries is limited and does not consistently show that the social distribution of premature death has become more unequal over the last decade.12,15,22 These observations prompt the question of what trends would be observed if data on inequalities in morbidity instead of mortality were used. A main source of information is national health interview (and similar) surveys. In the 1980s, these surveys were used to produce detailed descriptions of socioeconomic inequalities in health and risk factors for health in most European countries.23,24 During the late 1990s, a second and even a third round of national health surveys were conducted in most European countries, offering an opportunity to assess whether inequalities in self-reported health have increased or decreased between the 1980s and thus 1990s. Results of national studies from several countries have been published recently.4,25–35 These results are not as consistent as those for mortality. For example, with regards to self-assessed general health, some studies observed a widening of inequalities between the 1980s and the 1990s whereas other studies suggested narrowing inequalities.25,26,28–33,36 It is uncertain whether this variation between studies reflects real differences among the countries studied. International overviews and comparisons are needed to assess whether there was a general tendency for inequalities in self-assessed health either to increase or to decrease and whether trends varied among countries. This article is based on the first European study of trends in socioeconomic inequalities in mortality and morbidity.37 It provides a comprehensive overview of changes in inequalities in self-reported morbidity. This overview includes 10 European countries, considers both men and women, and assesses health differences according to both education and income. The specific objectives were to determine (i) whether there was a general tendency for inequalities in morbidity either to increase or to

decrease between the 1980s and the 1990s and (ii) whether the direction and magnitude of change varied among countries. Self-assessed health is used as the health indicator because nearly identical questions on this indicator are included in interview surveys from almost all European countries. This measure has been shown to be a reliable and valid indicator of general health and well-being38 and is recommended for international overviews and comparisons.39

Materials and methods Table 1 gives an overview of the countries and surveys that are included in this article. Data were obtained from nationally representative health interview surveys or from multipurpose surveys with a significant health component. Data were obtained from surveys that were at least 7 years apart, with the first survey held in the 1980s and the second survey held in the 1990s. The selected surveys differ with respect to the interview method used and the non-response rate. In no country, however, have there been major changes over time in these aspects. In addition to the 10 countries included in this article, several other European countries have two or more national health interview surveys.40 We had to exclude these countries from our overview either because the first survey was held in the 1990s instead of the 1980s (e.g. Switzerland, some Eastern European countries) or because the 1980s survey did not include a question on self-assessed health (e.g. France and Ireland). In most countries, first- and second-generation migrants were included in the surveys. In Germany, however, those with nonGerman nationality were excluded from the surveys. Exclusion of these people might have biased health inequality estimates because most migrants have a low socioeconomic status and may have generally worse health than native people of the same status. On the other hand, selection on good health during immigration may result in relatively good health among recent migrants. Unfortunately, no reliable estimates are available on the general health of migrants as compared with the native German population. The data presented in this paper relate to men and women aged 25–69 years. Younger respondents were excluded because many had not completed full-time education. Older respondents were excluded because of the bias that the exclusion of the institutionalized population might cause at higher ages. The number of respondents aged 25–69 years varies among countries. The Austrian and Italian surveys and also the first Spanish survey include relatively large numbers. For each country, the two surveys had comparable data on educational levels. Unfortunately, comparable data on income were available only for the surveys from Finland, Sweden, The Netherlands, and West Germany. In order to add another country to these four, we included the results of an analysis of data from the British General Household Surveys of 1984 and 1996: for both years this survey had comparable data on both self-assessed health and income.

Measurement of self-assessed health The health indicator is quantified as the proportion of respondents who state that their general health is ‘good’ or

SOCIOECONOMIC TRENDS IN SELF-ASSESSED HEALTH

297

Table 1 Characteristics of surveys

Country Finland

Sweden

Norway

Denmark

England

Netherlands

W. Germany

Austria

Italy

Spain

Survey year

% Nonresponse

Survey name

Total number of respondents aged 25–69 years Men

Women

1986

Finnish Survey on Living Conditions

13

4115

4514

1994

Finnish Survey on Living Conditions

27

3518

3407

1988

Swedish Survey on Living Conditions

21

2112

2083

1997

Swedish Survey on Living Conditions

22

2038

2143

1985

Health Interview Survey

21

2733

2859

1995


25

2625

2719

1987

Health and Morbidity Survey

20

1627

1644

1994

Health and Morbidity Survey

22

1622

1729

1985

Health and Lifestyle Survey

26

2403

3252

1995

Health Survey for England

29

5434

6302

1985–88


37

9936

10 083

1997–99

Permanent Survey on Living Conditions

42

8524

8868

1984–86

First National Health Survey

34

2416

2370

1990–91

Third National Health Survey

31

2590

2664

1983

Mikrozensus 1983, round 4

12

15 629

16 952

1991

Mikrozensus 1991, round 4

18

15 780

15 585

1986–87


8

21 587

22 632

1994


8

18 236

19 118

1987

National Health Interview Survey

10

9893

10 695

1997

National Health Interview Survey

16

2124

2260

‘fair/poor’ when asked a question similar to ‘How do you rate your general state of health: very good, good, fair, poor, or very poor?’. In nearly all countries, the same survey question with five answer categories was used in both surveys. Some of the Swedish and English surveys used a three-fold distinction, but they retained the essential distinction between ‘good’ and ‘fair’. A major exception is that the first Italian survey made a distinction only between ‘felt good’ and ‘not good’. The German survey used answer categories that lacked a clear distinction between ‘good’ and ‘fair’ health. The categories were ‘very good’, ‘good’, ‘satisfying’, ‘not so good’, and ‘poor’. Therefore, we used two indicators for Germany, one that is restricted to ‘not so good’ and ‘poor’ and another that includes ‘satisfying’. Since the two indicators showed nearly identical changes over time, only the latter indicator will be presented in this article. In addition to using the ‘fair/poor’ indicator, we also assessed inequalities in ‘poor’ health only. Generally, larger relative inequalities were observed for ‘poor’ health than those reported in this article for ‘fair/poor’ health. Unfortunately, owing to the small number of respondents reporting ‘poor’ health, confidence intervals for these inequality estimates were wide in many countries, and changes over time could not be determined with sufficient precision.

Measurement of educational level In each survey, men and women were classified into three categories according to their completed educational level. These

categories correspond broadly to levels 1 and 2 (elementary and lower secondary education), level 3 (upper secondary education), and levels 4–6 (post-secondary education) of the International Standard Classification of Education.41 For some countries, however, deviations from this standard classification had to be made in order to ensure as even a distribution of the survey population as possible over the three levels. Table 2 shows the resulting distribution of the respondents according to these levels. As might be expected, in every country the share of the highest educational level is higher among men than among women, and this share increases over time. Relatively large changes are observed for Norway and Spain, probably resulting from slight changes in the educational classification applied in the two surveys. The share of the highest educational level is about the same in most countries, but it is smaller in Spain and Italy. The educational distributions for Germany and Austria are somewhat different because of different educational systems in which, among other factors, more emphasis is given to vocational training at secondary levels.

Measurement of income level Income level was measured as the household equivalent income (HEI). Its calculation involved three steps. First, basic data on income were obtained from the national surveys. Income had to be measured for the entire household (or family). Where necessary, we added the income of all household members with substantial earnings. If data were

298


Table 2 Population distribution by educational level: men and women aged 25–69 years Proportion of the total population (%) Men Country

Educational level

Finland

Sweden

Norway

Denmark

England

1990s

1980s

Post-secondary education

18.2

23.7

14.3

19.8

Upper secondary education

34.2

40.3

36.4

45.2

W. Germany

Austria

Italy

Spain

1990s

Up to lower secondary education

47.6

36.1

49.3

35.0

Upper or post-secondary education

24.7

29.6

23.6

30.3

Lower secondary education

43.9

48.9

43.4

48.5

No or elementary education

31.4

21.5

33.0

21.2


20.9

28.3

14.7

25.0


20.0

33.2

13.0

21.8


59.1

38.5

72.3

53.1


22.3

27.6

19.1

27.8

10 years’ school

54.2

53.3

44.8

47.4

Up to 9 years’ school

23.5

19.2

36.1

24.8

Advanced level exam (age 18 yr) or NVQ level 3 or higher

16.6

32.3

14.2

21.4

29.3

33.4

27.2

36.2

Certificate of Secondary Education, Ordinary level, or NVQ levels 1, 2

Netherlands

Women

1980s

No qualifications

54.1

34.3

58.6

42.4


20.0

25.4

13.1

17.2


37.0

37.8

26.9

32.5


43.0

36.8

60.0

50.3

High level (incl. Hochschule)

18.5

25.3

11.3

14.6

Medium level (incl. Abitur)

14.8

16.1

19.7

23.6

Basic level (incl. Hauptschule, Realschule)

66.7

58.6

69.0

61.8

Upper or post-secondary education

16.3

19.8

10.8

15.2

Lower secondary education

54.1

57.0

34.8

40.2


29.5

23.2

54.4

44.6

7.4

8.3

5.4

6.5

Secondary education


49.8

60.8

42.0

54.5


42.9

30.9

52.6

39.1


14.1

15.5

8.2

14.2

Secondary education

20.3

43.8

14.4

38.8


65.6

40.7

77.4

47.0

available on different income components, all available income components were added. Income was measured as net income, that is, after deductions of taxes and social security premiums. Second, the net household income was adjusted for the household size. The size of the household was measured as the total number of members of that household, including children and the elderly. The HEI was calculated by dividing the net household income by the square root of the household size. This formula is similar to the one used in a previous comparative project.42 Third, respondents were classified on the basis of their HEI into income quintile groups. These five groups had about the same proportion of the total number of male and female respondents aged 25–69 years.

In the British and Dutch surveys, income levels were unknown for about 20% of the respondents in both periods. Income was unknown for about 8% of the respondents to the two German surveys. In the Finnish and Swedish surveys, income levels were known for all respondents as a result of linkage to data from the tax registry.

Statistical methods The prevalence of ‘fair/poor’ health per socioeconomic group was measured by means of standardized prevalence rates. Standardization by 5-year age group was done by means of the direct method, with the European population of 1987 as the standard. This age standardization procedure allowed us to control for differences in age structure among socioeconomic


groups and, in addition, between men and women, among countries, and among periods. In order to facilitate comparisons over time, odds ratios (ORs) were calculated that compared the lowest educational level (or income quintile) with the highest educational level (or income quintile). ORs were estimated by means of logistic regression analysis with the proportion of respondents with ‘fair/poor’ health as the dependent variable. The independent variables were a series of dummy variables representing 5-year age groups and a variable that represented the contrast between the highest and lowest educational levels (or income quintiles). The difference between the ORs for the two periods is used to measure changes over time in the magnitude of health inequalities. Confidence intervals were calculated for this difference by means of a standard formula on the difference between the estimates for two independent samples. For each period separately, we also calculated an unweighted average of the ORs for all countries together. Confidence intervals for this pooled measure were calculated by means of a standard formula on the average of the estimates from independent samples. Italy was excluded from this pooled estimate because of a problem with the comparability of the question on self-assessed health in the two Italian surveys (see ‘Discussion’ for details). In trend analyses that are presented elsewhere,37 we also assessed trends with respect to the absolute (instead of relative) difference in prevalence rates between the highest and lowest groups. Moreover, we assessed changes over time using the Relative Index of Inequality, which has the advantage of being able to take into account all socioeconomic groups separately and to account for changes in the population distribution across these groups.43 Finally, trends in health inequalities were also assessed by applying prevalence rate ratios instead of ORs.34 These different health inequality indices are not presented in this article because they showed basically the same changes over time as those presented in the next section.37

Results Table 3 presents estimates of the prevalence of ‘fair/poor’ general health for the total male and female survey populations.

299

The overall prevalence of ‘fair/poor’ health varies considerably among countries, with the highest rates observed for Finland and Italy. In Germany, the prevalence rates are much higher than elsewhere, but this is due to the inclusion of ‘satisfying’ health. Excluding ‘satisfying’ reduces the German prevalence rate to about 15%. In most cases, the national prevalence of ‘fair/poor’ health declined over time. The major exception is Italy, where prevalence rates increased by more than 6 percentage points for the population at large. This increase is probably due to changes in the Italian survey question on self-assessed health (see ‘Material and Methods’). In Table 4, a distinction is made by educational level. In most countries, the extent to which prevalence rates changed over time varies according to educational level. In several countries, notably The Netherlands, Italy, and Spain, changes were least favourable among the lowest educated men and women. (Trends in the national rates, presented in Table 3, are generally more favourable because these rates also reflect the increase in the percentage of the population belonging to higher educational levels.) Although Table 4 demonstrates that inequalities in selfassessed health persisted over time in each country, it is not clear to what extent these inequalities increased or decreased over time. In Table 5, the magnitude of health inequalities is expressed by means of ORs. A substantial increase in ORs is observed in a few countries, especially Italy and Spain. The opposite trend of decreasing inequalities is observed in a few other cases, for example, Norwegian and Danish men. For all countries together (see last row), the ORs comparing low with high educational levels remained stable for men (2.61 in the 1980s and 2.54 in the 1990s) but increased slightly for women (from 2.48 to 2.70). The confidence intervals for the ORs of Table 5 are large, implying that the observed changes in health inequalities may be strongly influenced by chance fluctuations. Their effect is evaluated in Figure 1, which presents estimates of the absolute change in ORs together with 95% confidence intervals for these change estimates. In only a few cases (Italian men and women, Spanish women) do confidence intervals not overlap the 0 value, implying that a widening of health inequalities can be demonstrated with statistical significance. In a few other cases

Table 3 Prevalence of ‘fair/poor’ self-assessed health in total population: men and women aged 25–69 years Prevalence rate (per 100 respondents) Men Country

Women

1980s

1990s

Change

1980s

1990s

Change

Finland

41.7

37.0

4.7

41.5

35.4

6.0

Sweden

19.6

19.2

0.4

24.8

21.4

3.4

Norway

20.4

16.8

3.6

22.0

19.2

2.9

Denmark

20.2

16.7

3.5

21.2

22.6

1.4

England

26.4

20.7

5.7

26.1

22.0

4.1

Netherlands

19.3

19.0

0.3

21.8

24.6

2.8

W. Germany

54.3

53.9

0.3

58.2

54.5

3.7

Austria

26.8

26.3

0.5

31.6

28.5

3.0

Italy

27.9

34.2

6.3

35.4

42.0

6.6

Spain

27.3

27.2

0.1

36.7

32.5

4.2

300


Table 4 Prevalence of ‘fair/poor’ self-assessed health by education level: men and women aged 25–69 years Prevalence rate (per 100 respondents) Country

Level of education

Finland

Sweden

Norway

Denmark

England

Netherlands

W. Germany

Austria

Italy

Spain

Men

Women

1980s

1990s

Change

1980s

1990s

Change

High

25.9

23.9

2.1

26.9

20.7

6.2

Mid

40.2

36.9

3.2

38.3

35.8

2.5

Low

48.8

45.7

3.1

48.0

43.3

4.6

High

12.0

13.1

1.1

12.0

13.1

1.2

Mid

20.1

19.5

0.6

24.8

22.3

2.5

Low

24.7

26.9

2.2

34.1

31.0

3.0

High

10.4

11.8

1.3

8.0

10.0

2.0

Mid

12.9

14.9

2.0

15.7

17.5

1.8

Low

26.5

22.1

4.4

26.0

24.2

1.9

High

13.7

12.6

1.1

11.7

16.3

4.6

Mid

18.2

16.4

1.8

17.8

19.3

1.5

Low

31.0

23.6

7.4

30.4

35.9

5.5

High

14.2

12.8

1.5

17.4

14.2

3.3

Mid

21.7

12.8

9.0

19.2

17.1

2.1

Low

32.7

30.3

2.4

31.4

30.1

1.3

High

10.6

11.1

0.5

15.2

15.9

0.7

Mid

16.7

17.4

0.7

17.0

20.8

3.8

Low

24.6

25.9

1.3

24.3

29.5

5.2

High

46.4

45.8

0.6

45.5

46.1

0.6

Mid

53.8

50.7

3.1

51.5

45.1

6.4

Low

56.6

58.3

1.8

62.1

60.1

2.1

High

14.8

16.0

1.2

18.9

19.1

0.3

Mid

27.0

26.1

0.9

27.2

24.5

2.7

Low

33.2

35.7

2.4

36.9

35.4

1.5

High

19.5

21.3

1.7

25.0

31.8

6.7

Mid

25.3

32.0

6.7

32.2

38.3

6.2

Low

32.5

42.1

9.6

39.0

48.8

9.8

High

19.6

18.8

0.8

25.1

17.8

7.3

Mid

22.6

23.1

0.6

26.9

27.0

0.0

Low

30.5

34.8

4.3

39.7

41.5

1.8

(English women, Dutch women, Spanish men), the increase in health inequalities reaches borderline significance. In most other situations, especially in the Nordic countries, confidence intervals are so wide that no precise estimates can be made of the extent to which health inequalities changed over time. Results for income are given in Table 6. In each country for which data are available, men and women with lower income more often assess their health as ‘fair’ or ‘poor’. The association between income and self-assessed health is generally linear, with no consistent evidence of a threshold effect separating the poorest from the remainder of the population. In only a few cases, especially British men and women, is a large gap found between the prevalence rates of the lowest quintile and the rates observed for the remainder of the population. Changes over time in prevalence rates are not consistently related to income level. An exception to this rule is The Netherlands, where the prevalence of ‘fair/poor’ health declined in most income groups but increased in the lowest groups.

The ORs given in Table 7 help to assess whether incomerelated health differences widened or narrowed between the 1980s and the 1990s. These differences were more or less stable in most cases. However, substantial increases were observed for Sweden (women), Great Britain (women), and The Netherlands (men and women). For all countries together, income-related inequalities increased among men (ORs increased from 3.06 to 3.40) and to a larger extent among women (from 2.43 to 2.86). Figure 2 shows the extent of the increase in individual countries and adds 95% confidence intervals in order to assess the precision of the change estimates. Increases in inequalities can be demonstrated with statistical significance for Dutch women only. In all other cases, changes cannot be demonstrated with statistical significance. It would, however, be erroneous to interpret this non-significance as a sign of stability. The figure shows that the confidence intervals for the change estimates are wide and that they do not exclude the possibility that substantial increases or decreases did occur.


301

Table 5 Magnitude of educational differences in ‘fair/poor’ self-assessed health: men and women aged 25–69 years Odds ratio (95% confidence interval) Men

Women

Country

1980s

1990s

1980s

1990s

Finland

3.15 (2.55–3.88)

2.99 (2.44–3.66)

2.86 (2.28–3.58)

3.29 (2.60–4.18)

Sweden

2.37 (1.71–3.29)

2.37 (1.70–3.30)

3.32 (2.37–4.66)

3.06 (2.22–4.23)

Norway

2.93 (2.16–3.98)

2.30 (1.73–3.04)

3.10 (2.13–4.50)

2.84 (2.10–3.82)

Denmark

2.96 (1.99–4.40)

2.16 (1.45–3.21)

2.88 (1.91–4.34)

3.00 (2.13–4.21)

England

3.11 (2.27–4.25)

3.08 (2.57–3.68)

2.08 (1.59–2.71)

2.66 (2.21–3.19)

Netherlands

2.95 (2.46–3.52)

2.81 (2.39–3.30)

1.95 (1.63–2.35)

2.12 (1.81–2.49)

W. Germany

1.50 (1.20–1.88)

1.76 (1.44–2.14)

1.89 (1.43–2.50)

1.91 (1.50–2.44)

Austria

3.39 (2.92–3.93)

3.22 (2.79–3.71)

2.75 (2.37–3.19)

2.67 (2.31–3.07)

Italy

2.05 (1.79–2.34)

2.94 (2.54–3.40)

1.86 (1.62–2.15)

2.55 (2.20–2.95)

Spain

1.86 (1.58–2.17)

2.58 (1.81–3.67)

1.97 (1.63–2.37)

3.10 (2.18–4.41)

Total (excl. Italy)

2.61 (2.41–2.83)

2.54 (2.35–2.75)

2.48 (2.28–2.69)

2.70 (2.50–2.92)

Men

3 2 1 0 –1 –2

Germany

Austria

Italy

Spain

Germany

Austria

Italy

Spain

Netherl.

England

Denmark

Norway

Sweden

Finland

–3

Women 3 2 1 0 –1 –2 Netherl.

England

Denmark

Norway

Sweden

Finland

–3

Figure 1 Changes in educational differences in self-assessed health: absolute difference between odds ratios for the 1980s and the 1990s, with 95% confidence interval. Positive figures indicate an increase in odds ratios

Discussion Evaluation of data sources Until recently, studies on trends in health inequalities relied mainly on mortality data. Increasingly, however, national interview surveys are used to assess whether health inequalities widened or narrowed over time. Data from health interview surveys are used not only because of the wish to move from mortality to morbidity, but also because these surveys offer a number of specific advantages. First, these surveys are available

for many of the European countries that still lack national longitudinal mortality studies. In addition, health surveys allow the consideration of a broader set of variables than is usually available for mortality studies, including different health indicators, a set of complementary socioeconomic indicators, and information on health-related behaviours and other health determinants.27,40,44 Nonetheless, the use of interview surveys to monitor inequalities in health poses a number of problems. A comprehensive discussion of all potential data problems is given elsewhere.37 In this article, we discuss only those that are probably most influential. One major problem is the low statistical power of most interview surveys. Estimates of the magnitude of changes over time in health inequalities were found to have generally low levels of precision (large confidence intervals). This may seem surprising given the fact that the national surveys used here are among the largest available in the respective countries and given the experience that their sample sizes are sufficiently large for many purposes. For example, they can be used to demonstrate inequalities in health at a single moment in time with high levels of statistical significance. However, estimates of the magnitude of these inequalities have large confidence intervals, with the consequence that changes in this magnitude cannot be estimated with much precision. This problem is compounded by the fact that during the relatively short time periods covered in this study (between 7 and 10 years in each country), the true changes in health inequalities were probably modest. A second problem, which is inherent in the use of any interview survey, relates to the reliance on reporting by respondents. The indicator used in this study, self-assessed health, has been extensively evaluated for both its validity and reliability. In both aspects, this indicator performs fairly well, implying that it enables a valid and reliable assessment of a person’s general health and well-being.38,45–47 Nonetheless, one cannot exclude the possibility that changes over time in its overall prevalence rates are influenced by shifts in, for example, the expectations that people have with regard to their health and factors that may change the awareness of health problems, such as improved detection of diseases. If these changes differ by socioeconomic group, they would bias estimates of changes

302


Table 6 Prelavence of ‘fair/poor’ self-assessed health by income level: men and women aged 25–69 years Prevalence rate (per 100 respondents) Country

Income quintile

Finland

Sweden

Great Britain

Netherlands

W. Germany

Men

Women

1980s

1990s

Change

1980s

1990s

Change

1 (highest)

34.0

26.7

7.3

32.9

26.9

6.0

2

39.7

38.9

0.8

39.5

34.4

5.0

3

44.6

41.0

3.7

42.3

37.2

5.1

4

50.7

41.6

9.1

49.1

43.1

6.0

5 (lowest)

54.5

49.0

5.6

51.4

42.4

9.1

1 (highest)

12.0

11.8

0.2

19.0

14.7

4.3

2

20.8

18.0

2.7

19.5

18.8

0.7

3

22.2

17.4

4.8

24.4

21.1

3.4

4

31.5

30.8

0.8

34.3

29.9

4.4

5 (lowest)

31.0

30.0

1.0

31.4

30.0

1.4

1 (highest)

14.1

19.2

5.0

21.1

22.7

1.5

2

19.5

25.0

5.5

24.1

34.4

10.3

3

21.0

27.3

6.3

26.5

31.8

5.3

4

25.7

32.7

7.0

32.6

38.1

5.5

5 (lowest)

34.4

44.2

9.7

41.8

49.9

8.1

1 (highest)

11.4

10.8

0.6

15.4

16.8

1.4

2

17.8

13.5

4.3

20.7

19.9

0.8

3

18.6

18.3

0.3

23.6

24.4

0.8

4

23.4

21.4

2.0

24.8

31.1

6.3

5 (lowest)

29.8

33.3

3.5

28.0

36.5

8.5

1 (highest)

46.7

46.9

0.1

51.1

43.2

7.9

2

51.3

53.2

2.0

53.9

47.6

6.4

3

50.9

55.1

4.2

58.1

53.6

4.4

4

59.2

55.9

3.3

58.2

58.9

0.7

5 (lowest)

59.8

62.5

2.7

66.8

63.2

3.5

Table 7 Magnitude of income-related differences in ‘fair/poor’ self-assessed health: men and women aged 25–69 years Odds ratio (95% confidence interval) Men Country

Women 1980s

1990s

1980s

1990s

Finland

2.92 (2.29–3.71)

3.09 (2.42–3.94)

2.65 (2.09–3.35)

2.43 (1.86–3.18)

Sweden

3.93 (3.45–7.15)

4.11 (2.82–6.04)

2.16 (1.48–3.16)

2.80 (1.92–4.09)

Great Britain

3.65 (2.83–4.70)

3.88 (3.09–4.88)

3.12 (2.49–3.92)

3.92 (3.21–4.79)

Netherlands

3.68 (3.00–4.50)

4.50 (3.66–5.52)

2.21 (1.84–2.67)

3.01 (2.49–3.63)

W. Germany

1.79 (1.33–2.39)

2.05 (1.55–2.72)

2.11 (1.53–2.91)

2.40 (1.81–3.18)

Total

3.06 (2.69–3.49)

3.40 (3.00–3.86)

2.43 (2.14–2.74)

2.86 (2.54–3.23)

in health inequalities. The extent of this bias cannot be measured and should perhaps not be exaggerated, especially for studies that cover a relatively short time period. A third problem relates to changes over time in survey designs and questionnaires. Even though we selected the surveys on the basis of a high degree of continuity and comparability over time, some changes may have biased our trend estimates. One particular problem is that the question on self-assessed health used in the Italian surveys changed over time. This change is likely to have caused the large increase

in overall prevalence rates which was observed for Italy only (Table 3). One cannot exclude the possibility that this effect might be larger among lower educational groups than among higher groups. A similar problem is that large increases in overall prevalence rates were also observed for Great Britain when using the General Household Surveys (GHSs) of 1984 and 1996 (Table 6), but not when using the other British surveys (Tables 3 and 4). The causes of this discrepancy are not known, but they may be related to changes in the GHS questionnaire. Therefore, the substantial changes in the size of health


3 2 1 0 –1

Ger - F

Ger - M

Net - F

Net - M

GBr - F

GBr - M

Swe - F

Swe - M

Fin - F

–3

Fin - M

–2

Figure 2 Changes in income-related differences in self-assessed health: absolute difference between odds ratios for the 1980s and the 1990s, with 95% confidence interval. Positive figures indicate an increase in Odds ratios. Fin, Finland; Swe, Sweden; GBr, Great Britain; Net, Netherlands; Ger, West Germany; M, male; F, female

inequalities that we observed for Italy (by education) and for Great Britain (by income) should be viewed with caution. The last problem that needs to be discussed are the high rates of non-response for some surveys. This applies particularly to the survey of The Netherlands, where the total non-response rate is about 40% and where item non-response on income is 20%. Together, these two factors imply that income-related inequalities in health in The Netherlands are estimated for only about one-half of the initial survey sample. The effects of nonresponse are perhaps less dramatic than one might fear, because response rates in The Netherlands do not clearly differ according to socioeconomic status,48 and response rates changed only slightly over time. Based on more detailed evaluations,35,37 we judged that non-response problems are unlikely to explain the increase in income-related health differences observed for The Netherlands. The data problems discussed above warn that results from individual surveys should be treated with some caution. They underscore the fact that it may be highly problematic to determine for individual countries the extent to which health inequalities have widened or narrowed during recent years. In the discussion below, we will therefore concentrate on the few international patterns that are not likely to be due to specific data problems.

Discussion of key findings The main finding from this study is that socioeconomic inequalities in self-assessed health showed a high degree of stability in European countries. This finding adds to the increasing evidence that socioeconomic inequalities are highly persistent in Europe. The same persistence was observed in our previous cross-national comparisons for the 1980s, in which we observed that there was no European country where socioeconomic inequalities in morbidity and mortality were clearly smaller than elsewhere.42,49 The persistence of large health inequalities stresses that these inequalities are deeply rooted in the social stratification systems of modern societies. It also underlines the increasing recognition that health and social disadvantage are intimately related as a result of multiple links that unfold and persist over a person’s life course.50,51 Finally,

303

the observed persistence of health inequalities warns that it would not be realistic to expect a substantial reduction in health inequalities within a short period. Targets for policies that aim to reduce inequalities in morbidity or mortality should therefore avoid overoptimism.52 A second tendency in our international overview—and perhaps a tendency that will give rise to more optimistic interpretations—is that changes in health inequalities seem to be more favourable in the Nordic countries. Even though estimates from individual Nordic countries were not precise because of the small sample size of each Nordic survey, health differences according to educational level (the only indicator available for all Nordic countries) did not show a general tendency to increase. Similar findings were obtained from a series of national studies that analysed the same Nordic surveys with different analytical methods or that analysed data from other surveys.4 The relatively favourable record of the Nordic countries is remarkable given the economic crises that most experienced in the late 1980s or early 1990s.53 Even though Nordic studies reported adverse effects of the economic crises on the psychological well-being of the population, especially the younger generations,54 improvements in somatic health outcomes seem to have continued, also among lower socioeconomic groups. These trends suggest that Nordic welfare states may have been able to moderate, at least in the short term, the adverse effects that economic crises could have had on the general health of disadvantaged sections of the population.4 A country where inequalities in health might have widened over time is Great Britain, where income inequalities increased at an unprecedented rate in the 1980s.53 Our results suggest that inequalities in self-assessed health tended to increase as well. However, this increase was not clearly larger than those observed in many other European countries. Similar patterns were observed for mortality. In England and Wales, substantial increases are observed for inequalities in mortality, measured both at the area level5 and according to occupational class.5,10 However, the pace of increase does not appear to be larger than in other northern European countries.21 Thus, the exceptionally strong rise in income inequalities in England does not seem to be accompanied by a more-than-average increase in inequalities in mortality or morbidity. This finding does not necessarily imply that health inequalities are insensitive to income inequalities, as the full effect of increasing income inequalities may appear later over the life course of those who are affected.55 Firm evidence on the impact of social and economic change should come from future studies of trends in socioeconomic inequalities in health. Economic developments and socioeconomic policies are expected to have large effects on the health of disadvantaged population groups.55–57 Attention needs to be given to different stages of the economic cycle, including the occurrence of major economic crises,3 and to gradual socioeconomic change, such as changes in welfare state provision4,58 and economic reform.59 The impact of these changes on health inequalities can be assessed by studying trends over a longer span of time and by studying social gradients in intermediate factors such as living standards, working conditions, and health-related behaviours.25,36,60,61 Important opportunities for such studies will arise in the coming

304


years, when a new round of health interview surveys will be completed in many European countries.

Acknowledgements This study was supported financially by the Health Monitoring Program of the European Commission under contract number SOC 98 201376 05F03. We very much thank Charlie Owen of

the Institute of Education at London for analysing data from the General Household Survey on income-related health inequalities. During two workshops, in Rotterdam (1999) and Bielefeld (2000), useful comments on the design and preliminary results of this study were given by several colleagues, and especially by Seeromanie Harding, Örjan Hemström, Hubert Isnard, Christophe Koller, Richard Layte, and Tapani Valkonen.

KEY MESSAGES •

In each of the 10 countries included in this study, large differences in self-assessed health were observed in relation to both educational level and household income.

•

The magnitude of inequalities remained more or less stable between the 1980s and the 1990s. However, the extent of changes could not be estimated with much precision.

•

Increasing inequalities were observed for Spain, Italy, and The Netherlands, whereas inequalities in the Nordic countries showed no tendency to increase.

•

These findings suggest that Nordic welfare states were able to protect disadvantaged groups against many of the adverse health effects of economic crises.

14 Lostao L, Regidor E, Aiach P, Dominguez V. Social inequalities in

References 1 Townsend P, Davidson N Whitehead M (eds). Inequalities in Health. The

Black Report & The Health Divide. London: Penguin Books, 1988. 2 Smith GD, Shaw M, Mitchell R, Dorling D, Gordon D. Inequalities

in health continue to grow despite government’s pledges. BMJ 2000;320:582. 3 Martikainen PT, Valkonen T. Excess mortality of unemployed men

and women during a period of rapidly increasing unemployment. Lancet 1996;348:909–12. 4 Lahelma E, Kivela K, Roos E et al. Analysing changes of health

inequalities in the Nordic welfare states. Soc Sci Med 2002;55:609–25. 5 Shaw M, Dorling D, Gordon D, Davey Smith G. The Widening Gap:

Health Inequalities and Policy in Britain. Bristol: The Policy Press, 1999. 6 Mackenbach JP, Bakker M (eds). Reducing Inequalities in Health: A

European Perspective. London: Routledge, 2002. 7 World Health Organisation. Health 21—Health for All in the 21st Century.

Copenhagen: World Health Organisation, 1999. 8 Van Herten LM, Van de Water HPA. Health policies on target? Review

of health target setting in 18 European countries. Eur J Public Health 2000;10(4 Suppl.):11–16. 9 Department of Health. Saving Lives: Our Healthier Nation. London: The

Stationery Office, 1999. mortality by cause, 1986 to 2000. Health Statistics Quaterly 2003;20:25–37. 11 Diderichsen F, Hallqvist J. Trends in occupational mortality among

in

15 Regidor E, Gutierrez-Fisac JL, Rodriguez C. Increased socioeconomic

differences in mortality in eight Spanish provinces. Soc Sci Med 1995;41:801–7. 16 Lang T, Ducimetiere P. Premature cardiovascular mortality in France:

divergent evolution between social categories from 1970 to 1990. Int J Epidemiol 1995;24:331–39. 17 Jozan P, Forster DP. Social inequalities and health: ecological study of

mortality in Budapest, 1980–3 and 1990–3. BMJ 1999;318:914–15. 18 Dahl E, Kjaersgaard P. Trends in socioeconomic mortality differentials

in post-war Norway. Sociol Health Illn 1993;15:587–611. 19 Kunst AE, Looman CW, Mackenbach JP. Socio-economic mortality

differences in The Netherlands in 1950–1984: a regional study of cause-specific mortality. Soc Sci Med 1990;31:141–52. 20 Martikainen P, Valkonen T, Martelin T. Change in male and female life

expectancy by social class: decomposition by age and cause of death in Finland 1971–95. J Epidemiol Community Health 2001;55:494–99. 21 Mackenbach JP, Bos V, Andersen O et al. Widening socioeconomic

inequalities in mortality in six Western European countries. Int J Epidemiol 2003;32:830–37. 22 Cardano M, Costa G, Demaria M, Merler E, Biggeri A. Le

10 White C, Galen FV, Huang Chow Y. Trends in social class differences in

middle-aged men 1997;26:782–87.

ischaemic heart and cerebrovascular disease mortality in men: Spain and France, 1980–1982 and 1988–1990. Soc Sci Med 2001;52:1879–87.

Sweden

1961–1990.

Int J Epidemiol

12 Borrell C, Plasencia A, Pasarin I, Ortun V. Widening social inequalities

in mortality: the case of Barcelona, a southern European city. J Epidemiol Community Health 1997;51:659–67. 13 Valkonen T, Martikainen P, Jalovaara M, Koskinen S, Martelin T,

Makela P. Changes in socioeconomic inequalities in mortality during an economic boom and recession among middle-age men and women in Finland. Eur J Public Health 2000;10:274–80.

diseguaglianze di mortalita negli studi longitudinali italiani [Inequalities in mortality in the Italian longitudinal studies]. Epidemiol Prev 1999;23:141–52. 23 Cavelaars AE, Kunst AE, Geurts JJ et al. Differences in self reported

morbidity by educational level: a comparison of 11 western European countries. J Epidemiol Community Health 1998;52:219–27. 24 Cavelaars AE, Kunst AE, Geurts JJ et al. Morbidity differences by

occupational class among men in seven European countries: an application of the Erikson–Goldthorpe social class scheme. Int J Epidemiol 1998;27:222–30. 25 Borrell C, Rue M, Pasarin MI, Rohlfs I, Ferrando J, Fernandez E.

Trends in social class inequalities in health status, health-related


behaviors, and health services utilization in a Southern European urban area (1983–1994). Prev Med 2000;31:691–701. 26 Anitua C, Esnaola S. Changes in social inequalities in health in the

Basque Country. J Epidemiol Community Health 2000;54:437–43. 27 Bartley M, Fitzpatrick R, Firth D, Marmot M. Social distribution of

illustrated with 1997;44:757–71.

two

examples

from

Europe.

305

Soc Sci Med

44 Mason V, Bridgwood A. Methods of Collecting Morbidity Statistics. Revised

Report to the Eurostat Task Force on ‘Health and Health-Related Survey Data’. London: Office for National Statistics, 1997.

cardiovascular disease risk factors: change among men in England 1984–1993. J Epidemiol Community Health 2000;54:806–14.

45 Lundberg O, Manderbacka K. Assessing reliability of a measure of

28 Lahelma E, Arber S, Rahkonen O, Silventoinen K. Widening or

46 Martikainen P, Aromaa A, Heliovaara M et al. Reliability of perceived

narrowing inequalities in health? Comparing Britain and Finland from the 1980s to 1990s. Social Health Illness 2000;22:110–36.

47 Segovia J, Bartlett RF, Edwards AC. An empirical analysis of the

29 Lissau I, Rasmussen NK, Hesse NM, Hesse U. Social differences in

self-rated health. Scand J Soc Med 1996;24:218–24. health by sex and age. Soc Sci Med 1999;48:1117–22. dimensions of heath status measures. Soc Sci Med 1989;29:761–68.

illness and health-related exclusion from the labour market in Denmark from 1987 to 1994. Scand J Public Health 2001; 29(Suppl. 55): 19–30.

48 Geuzinge L, van Rooijen J, Bakker B. Project SSB-Fase 2. Correctie voor

30 Manderbacka K, Lahelma E, Rahkonen O. Structural changes and

49 Kunst AE, Groenhof F, Mackenbach JP. Mortality by occupational

social inequalities in health in Finland, 1986–1994. Scand J Public Health 2001; 29(Suppl. 55):41–54. 31 Lundberg O, Diderichsen F, Yngwe MA. Changing health inequalities

selectieve non-respons in persoonsenquêtes door gebruik van informatie uit registers. Heerlen: CBS, 1999. class among men 30–64 years in 11 European countries. EU Working Group on Socioeconomic Inequalities in Health. Soc Sci Med 1998;46:1459–76.

in a changing society? Sweden in the mid-1980s and mid-1990s. Scand J Public Health 2001; 29(Suppl. 55):31–39.

50 Elstad J. Social Inequalities in Health and Their Explanations. Oslo:

32 Dahl E, Elstad JI. Recent changes in social structure and health

51 Holland P, Berney L, Blane D, Smith GD, Gunnell DJ, Montgomery SM.

inequalities in Norway. Scand J Public Health 2001; 29(Suppl. 55):7–17. 33 Mackenbach JP, Kunst AE. Measuring the magnitude of socio-

Norwegian Social Research (NOVA), 2000. Life course accumulation of disadvantage: childhood health and hazard exposure during adulthood. Soc Sci Med 2000;50:1285–95.

economic inequalities in health: an overview of available measures illustrated with two examples from Europe. Soc Sci Med 1997;44:757–71.

52 Kunst AE, Mackenbach JP. Setting Realistic Targets on Inequalities in

34 Regidor E, Gutierrez-Fisac JL, Dominguez V, Calle ME, Navarro P.

53 Kautto M, Fritzell J, Hvinden B, Kvist J, Uusitalo H (eds). Nordic

Comparing social inequalities in health in Spain: 1987 and 1995/97. Soc Sci Med 2002;54:1323–32. 35 Dalstra JA, Kunst AE, Geurts JJ, Frenken FJ, Mackenbach JP. Trends

in socioeconomic health inequalities in the Netherlands, 1981–1999. J Epidemiol Community Health 2002;56:927–34. 36 Bartley M, Owen C. Relation between socioeconomic status,

employment, and health during economic change, 1973–93. BMJ 1996;313:445–49. 37 Kunst AE, Bos V, Mackenbach JP, EU Working Group Inequalities in

Health. Monitoring Socioeconomic Inequalities in Health in the European Union: Guidelines and Illustrations. Rotterdam: Erasmus University, 2001. 38 Manderbacka K. Examining what self-rated health question is

understood to mean by respondents. Scand J Soc Med 1998;26:145–53. 39 Bruin AD, Pichavet HSJ, Nossikov A. Health Interview Surveys: Towards

International Harmonization of Methods and Instruments. Copenhagen: WHO Regional Publications, 1996. 40 Hupkens C. Coverage of Health Topics by Surveys in the European Union.

Luxembourg: Eurostat, 1998. 41 OECD. International Standard Classification of Education. Paris: OECD,

1997.

Health: A New Method Applied to Smoking and Smoking-Related Mortality in the Netherlands. Rotterdam: Erasmus University, 2001. Welfare States in the European Context. London and New York: Routledge, 2001. 54 Hallsten L, Grossi G, Westerlund H. Unemployment, labour market

policy and health in Sweden during years of crisis in the 1990’s. Int Arch Occup Environ Health 1999;72(Suppl.):S28–S30. 55 Graham

H. Building an inter-disciplinary science of health inequalities: the example of life course research. Soc Sci Med 2002;55:2005–16.

56 Whitehead M, Burstrom B, Diderichsen F. Social policies and the

pathways to inequalities in health: a comparative analysis of lone mothers in Britain and Sweden. Soc Sci Med 2000;50:255–70. 57 Navarro V, Shi L. The political context of social inequalities in health.

Int J Health Serv 2001;31:1–21. 58 Lundberg O, Lahelma E. Nordic health inequalities in the European

context. In: Kautto M, Fritzell J, Hvinden B, Kvist J, Uusitalo H (eds). Nordic Welfare States in the European Context. London: Routledge, 2001, pp. 42–65. 59 Leinsalu M, Vagerö D, Kunst AE. Estonia 1989–2000: enormous

increase in mortality differences by education. Int J Epidemiol 1993;32:1081–87.

42 Mackenbach JP, Kunst AE, Cavelaars AE, Groenhof F, Geurts JJ.

60 Helmert U, Shea S, Maschewsky-Schneider U. Social class and

Socioeconomic inequalities in morbidity and mortality in western Europe. The EU Working Group on Socioeconomic Inequalities in Health. Lancet 1997;349:1655–59.

cardiovascular disease risk factor change in West Germany 1984–1991. Eur J Public Health 1995;5:103–8.

43 Mackenbach JP, Kunst AE. Measuring the magnitude of socio-

economic inequalities in health: an overview of available measures

61 Lahelma E, Rahkonen O, Berg MA et al. Changes in health status and

health behavior among Finnish adults 1978–1993. Scand J Work Environ Health 1997;23(Suppl. 3):85–90.