Changing health inequalities in east and west Germany since unification

Changing health inequalities in east and west Germany since unification

ARTICLE IN PRESS Social Science & Medicine 58 (2004) 119–136 Changing health inequalities in east and west Germany since unification Ellen Nolte*, Ma...

219KB Sizes 1 Downloads 112 Views

ARTICLE IN PRESS

Social Science & Medicine 58 (2004) 119–136

Changing health inequalities in east and west Germany since unification Ellen Nolte*, Martin McKee London School of Hygiene and Tropical Medicine, European Centre on Health of Societies in Transition, Keppel Street, London WC1E 7HT, UK

Abstract The unification of Germany in 1990 brought about substantial social and economic changes in its eastern part, with new uncertainties and, despite increasing overall income, rising inequality. This paper explores the potential impact on health of these changes during the 1990s, looking specifically at income-related health inequalities in east and west Germany and its modulation by psychosocial factors. We used data from the German Socio-Economic Panel (GSOEP) for the years 1992 and 1997, including individuals aged 25+. We investigated changes in self-perceived health in the two parts of Germany and its socio-economic and psychosocial determinants. Analyses estimated odds ratios of less than good health using logistic regression. In 1992, 47% of east Germans rated their health worse than good compared with 54% in the west. By 1997, the east– west gap in self-rated health had disappeared, with the prevalence of poor health increasing to 56% in both parts. Income and education were important determinants of health in east and west, with, in the age-sex-adjusted model, those having available less than 60% of median equivalent income being at increased risk of poor health in 1992 (OReast 2.39, 1.45–3.94; ORwest 2.04, 1.65–2.52). Addition of education reduced the strength of this relationship only slightly. In the west, income-related health inequalities widened between 1992 and 1997 yet the initially stronger gradient declined in the east, despite an overall increase in income inequality (OReast 1.63, 1.04–2.56; ORwest 2.65, 2.19–3.21). The impact of education remained stable. Psychosocial variables were important determinants, mediating the effects of income, with leisure-cultural social involvement exerting the strongest effect in both east and west. The results show that, unlike in the west, the overall increase in income inequality in east Germany between 1992 and 1997 was not accompanied by a simultaneous increase in income-related health inequalities. This suggests that mechanisms involved in the association of socio-economic factors and health possibly behave differently in east and west. r 2003 Elsevier Science Ltd. All rights reserved. Keywords: East and west Germany; Self-perceived health; Health inequalities; Psychosocial factors

Introduction Early research on the determinants of patterns of health in populations, which typically focused on standard measures of socio-economic position, such as education, income or occupation, has given way to much more complex models that embrace psychosocial *Corresponding author. Tel.: +44-20-7612-7809; fax: +4420-7612-7812. E-mail address: [email protected] (E. Nolte).

and related factors such as optimism (Kunzansky & Kawachi, 2000) or participation in social networks (Berkman & Glass, 2000), with a growing body of work showing how these modulate the health effects of other factors. Large-scale social change offers a natural experiment from which it may be possible to derive new insights into the determinants of health. Thus, the decline in the civilian death rate in World War II in the United Kingdom has been attributed to a much more equal distribution of resources as well as a marked increase in

0277-9536/03/$ - see front matter r 2003 Elsevier Science Ltd. All rights reserved. doi:10.1016/S0277-9536(03)00153-9

ARTICLE IN PRESS 120

E. Nolte, M. McKee / Social Science & Medicine 58 (2004) 119–136

social cohesion in the face of a common adversary (Wilkinson, 1996). Research on the health effects of the collapse of the Soviet Union has provided new insights into the complex determinants of health in a fractured society (McKee, 2001). Yet, in most cases the circumstances in which such large-scale change arise mean that the researcher is dependent almost entirely on data on mortality, and in many cases even they are unavailable. The unification of Germany is one such natural experiment. Over the course of a few months a failing communist state merged with one of the world’s leading economies. The introduction of democratic structures, coupled with a massive financial injection, created both winners and losers. The freedom symbolised by the tearing down of the Berlin Wall was accompanied by a new uncertainty, as established social hierarchies broke down. The creation of a free market brought both opportunities for advancement and risks of failure. While overall income increased, some fared better than others and, while the increase in inequality was relatively moderate compared with other former socialist countries (Krause & Habich, 2000; UNICEF, 1997), it was still considerable. Importantly, unlike many other largescale transitions, data exist that make it possible to track changes in health and its determinants in both east and west Germany. The German Socio-Economic Panel (GSOEP), which includes questions on self-perceived health, has been conducted annually in the Federal Republic of German since 1984, with inclusion of the east since 1990 (Wagner, Burkhauser, & Behringer, 1993). This paper explores the changing pattern of health in the two parts of Germany during the 1990s. It concentrates on the changing scale and nature of inequalities in relation to income in east and west Germany but also how this relationship is modulated by a range of psychosocial factors.

GSOEP sample members are invited to participate in the study as long as they continue to share a dwelling with a member of the original household. Foreign residents were oversampled as were east German residents who were included as an additional sample in 1990. Samples were selected randomly by multistage (west German residents) or probability sampling (foreign and east German residents). Data collection used structured person and household questionnaires, usually completed during face-to-face interviews with trained interviewers. Sample response rates in the initial round ranged from 61% in the west German sample to 70% among east German residents and selected foreign residents. Overall representativeness of the samples is considered satisfactory, especially of the foreign resident sample. However, in both east and west German samples the elderly (age 70+) are underrepresented. The present analysis used a 95% sample of the original GSOEP, for 1992 and 1997; the restriction of the sample in the public use file is a consequence of German data protection legislation (Department of Policy Analysis and Management, 1999). The 1992 wave included a total of 12,676 individuals (east German sample: 3891) aged 16+ in 6326 households and in 1997 the survey covered 12,560 individuals (east German sample: 3644) in 6442 households. To adjust for disproportionate sampling of subgroups and nonresponse, data were weighted using pre-calculated, cross-sectional weights developed by GSOEP, normalised to the distribution of the non-institutionalised resident population according to official statistics (Pischner, 2000). The analysis presented in this paper was based on respondents aged 25+; respondents younger than 25 were excluded to minimise possible distortion of associations between health and socioeconomic factors by those who have not yet completed their education and/or vocational training.

Methods

Measurements

Data

Self-perceived health was used as the dependent variable, which, albeit being a subjective indicator of health has been shown to be a good predictor of mortality (Kaplan & Camacho, 1983; Idler & Benyamini, 1997; Schwarze, Andersen, & Anger, 2000) and later disability (Ferraro & Su, 2000). It was assessed by a standard question ‘‘How would you describe your health at present’’, using a five-point scale: ‘very good’, ‘good’, ‘satisfactory’, ‘less good’, and ‘poor’. For the analysis, the dependent variable was dichotomised into ‘good/very good health’ and ‘less than good health’.

The GSOEP has been described in detail elsewhere (Wagner et al., 1993; Burkhauser, Butrica, Dalz, & Lillard, 2000). In brief, it is a representative longitudinal study of private households and their adult members in Germany that was started in 1984 and has been repeated annually since. The original sample consists of about 6000 households containing over 12,000 individuals aged 16 and older who were individually interviewed. All adult members of sample households are considered members of the original sample, as are children who became adults during the course of the panel. Individuals remain members of the panel even if they leave their original household to live alone or with other people. Persons joining households with original

Socio-demographic variables Among socio-demographic variables, age was classified into five age groups 25–34, 35–44, 45–54, 55–64 and 65+. Marital status was categorised into ‘single’,

ARTICLE IN PRESS E. Nolte, M. McKee / Social Science & Medicine 58 (2004) 119–136

‘married/cohabiting’ and ‘separated/divorced/widowed’. Education was re-classified into categories based on the school leaving certificate, vocational qualifications and higher education, and further collapsed into four categories ranging from ‘very high’ (e.g. university degree) to ‘low’ (e.g. no or secondary school 1st stage certificate and no vocational qualifications) (Hoffmeister, Huttner, . Stolzenberg, Lopez, & Winkler, 1992; . Jockel et al., 1998). Employment status was dichotomised into ‘currently employed’ and ‘not employed’. Income was defined as total monthly household income (in German Marks, DM) of all household members after deduction of tax and social security contributions but including transfer income such as child benefits or housing subsidy. Income was then recalculated as needs-weighted per capita household income per month, i.e. equivalent income, by adjusting total household income as defined above for household size1 (Buhmann, Rainwater, Schmaus, & Smeeding, 1988). As the income distribution is usually highly skewed and thus not well represented by the mean, analyses further subdivided equivalent income into four categories to reflect distance from the median, using the following scale: ‘less than 60% of median income’, ‘60– o100%’, ‘100–o150%’ and ‘150+% of median income’. The lower level of less than 60% of median income was chosen to represent the poverty line as defined by the statistical office of the European Union (Eurostat) (Krause & Habich, 2000). Psychosocial variables Social involvement and/or social network was assessed asking the respondent ‘‘Which of the following activities do you practise in your spare time?’’ using a four-point response scale ‘weekly’, ‘monthly’, ‘less frequently’ and ‘never’: (1) Attend cultural events, e.g. concerts, theatre, lectures; (2) Go to cinema, pop/jazz concerts, dance halls, clubs, sports events; (3) Participate in sports; (4) Social gatherings with friends, relatives or neighbours; (5) Helping out with friends, relatives or neighbours; (6) Unpaid activities in clubs (Verein), associations or social services; (7) Participation in citizen organisations, political parties, community politics; (8) Attending church and/or other religious events. The eight variables were further investigated in a factor analysis to identify common underlying factors (Tabachnick & Fidell, 1996). The analysis was based on the principal component method, only factors with eigenvalues of greater than 1.0 were retained. An initial analysis was undertaken separately for the east and the west German samples to assess the possibility of different underlying factors. The factor loadings obtained from varimax rotation were subsequently used to 1 Equivalent family income=disposable income/household size0.73.

121

calculate individual (additive) scores2 and, based upon the distribution of the scores in the east and west German samples, dichotomised into ‘less than median’ and ‘higher than median’. Personal resources were assessed using the following statements ‘‘When I am thinking about the future I’m filled with confidence’’ and ‘‘Everything has become so complicated that I’m barely able to cope with it all’’. Both were rated on a four-point scale ranging from ‘agree completely’ to ‘completely disagree’, which for the purpose of this analysis was further collapsed into two categories ‘probable’ and ‘improbable’. Statistical analysis Our analysis uses a cross-sectional approach that considers the two waves as two samples taken at different points in time, using weights that were developed for cross-sectional analyses of the data (see above). About 75% of the subjects in the two periods were the same individuals; however, the age distribution of the two samples was very similar (see Appendix A). The associations between less than good health and socio-economic factors in east and west Germany were estimated using logistic regression and are expressed in terms of odds ratios (OR). Initial analyses were performed for men and women separately to identify potential interactions between gender and other variables. Formal tests for interaction confirmed a differential effect for marital status, thus sex-specific analyses were performed for this variable. Results are presented from several models, with the first model controlling for age and gender and the second also adjusting for education. To assess whether indicators of social and personal resources mediated the impact of income and health, age–sex–education adjusted OR were calculated before and after the inclusion of corresponding variables in the model.

Results Income inequalities The distribution of equivalent income by decile, excluding the 10th decile, in the two parts of Germany in 1992 and 1997 is shown in Fig. 1. In 1992, absolute income as well as income inequality was considerable higher in the west than in the east. East–west differences in income were, however, more pronounced in the higher income groups while the gap was considerably lower in the lower income deciles. By 1997, absolute income and income inequality had increased in the east, 2 Score=((Var1 * loadVar1)+(Var2 * loadVar2)+?+(VarX * LoadVarX))/(loadVar1+loadVar2+?+loadVarX).

ARTICLE IN PRESS E. Nolte, M. McKee / Social Science & Medicine 58 (2004) 119–136

122

1992

1997 4000 equivalent income (DM)

equivalent income (DM)

4000 3000 2000 1000

west east

3000 2000 west

1000

east 0

0 10

20

30

40

50

60

70

80

10

90

20

30

40

50

60

70

80

90

percentile

percentile

Fig. 1. Income distribution in east and west Germany by deciles, 1992 and 1997.

Income distribution -1997

Income distribution - 1992 25000

monthly equivalent income (DM)

monthly equivalent income (DM)

16000

12000

8000

4000

0 N=

7722

1878

West Germany

East Germany

20000

15000

10000

5000

0 N=

7516

1854

West Germany

East Germany

Fig. 2. Box plot of income distribution in east and west Germany in 1992 and 1997.

thus narrowing the gap with the west where changes in both indicators had been less obvious. In 1992, median income in the east was only 65% of median income in the west, by 1997 this had increased to 87%. This is illustrated in Fig. 2, which also shows the spread of the income distribution, with the upper tail remaining more dispersed in west Germany. Social involvement The factor analysis of the eight variables relating to social involvement extracted three factors that explained 57.4% of the common variance. Table 1 lists the estimates of the loadings obtained from principal component analysis after rotation along with the communalities for each variable, showing that at least 46% of the variability of all variables under investigation are explained by the three factors.

The first factor seemed to measure, largely, elements of ‘leisure-cultural’ social involvement as the variables ‘‘attend cinema, pop/jazz concerts’’ and ‘‘participate in sports’’ loaded highest; consequently, this factor was labelled ‘leisure-cultural social involvement’. The second factor included mainly aspects of ‘purpose-oriented’ social involvement such as performing volunteer work, participating in local politics and attending church or other religious events (‘purposeful social involvement’). The third factor appeared to measure activities associated with a social network of friends, relatives and/or neighbours such as social gatherings and mutual help (‘network-oriented social involvement’). The underlying pattern of social participation was similar in the two parts of Germany and remained stable over time. For the 1997 sample, an almost identical factor structure could be identified (not shown) with the three factors now explaining 58.8% of the variance.

ARTICLE IN PRESS E. Nolte, M. McKee / Social Science & Medicine 58 (2004) 119–136

123

Table 1 Factor loadings (after rotation) of indicators of social involvement (loadings >0.35 marked)—Germany 1992 Factor

1 2 3 4 5 6 7 8

Attend cultural events Attend cinema, pop/jazz concerts Participate in sports Social gatherings with friends, relatives, neighbours Help out friends, relatives, neighbours Volunteer work Participate in local politics Attend church/other religious events

Although the three factors show a fairly distinctive pattern, they do overlap, indicating that they are not independent but rather reflect varieties of a more general ‘sociability’ that clearly distinguishes those who do ‘something’ from those who do not participate in any social occasion at all. This is reflected by very small, although statistically significant, correlation between any of these factors and ‘loneliness’, which was also assessed in the GSOEP (Spearman’s rho, 1997: 0.064 to 0.112; po0:01). Self-perceived health Levels and distribution of self-perceived health, along with descriptive characteristics of the east and west German study samples in 1992 and 1997 are shown in Table 2. In 1992, the proportion of men with very high educational attainment was about twice as high than among women, at 12–16% compared with 7%. Also, low educational attainment was more frequent among women although proportions were generally higher in the west with one-third of west German women having less than medium level education. By 1997, proportions of those with high or very high educational attainment had slightly increased in both sexes and both parts, whereas the proportion of those with low educational attainment had declined somewhat, especially among women. Also, being employed was generally distributed similarly in east and west Germany in 1992, at 65% in east German men and 69% in west German men compared with 47% and 44% in women, respectively. Proportions remained fairly stable over time, although declining slightly in men in both regions and east German women, by up to 2%. These figures have, however, to be distinguished from those registered as unemployed, which, in accordance with official statistics, remained higher at 14.2%, in the east compared with 5.5% in the west (1997). Fairly similar patterns were also seen for marital status with at least two thirds being married or living

I

II

III

Communality

0.545 0.809 0.728 0.225 0.133 0.174 0.140 0.333

0.362 0.027 0.110 0.054 0.022 0.757 0.713 0.491

0.166 0.099 0.125 0.776 0.716 0.080 0.134 0.466

0.456 0.666 0.557 0.656 0.531 0.610 0.547 0.568

with a partner. Initially, this figure was slightly higher in the east but by 1997 had become almost identical to the west. In both east and west, substantially more women were separated, divorced or widowed, with the majority being widowed (61% in the east and 66% in the west in 1997), presumably due to higher mortality in men. Turning to self-perceived health, figures in Table 2 indicate that, in 1992, levels of less than good health in the adult population aged 25+ were significantly lower in the east than in the west, at 47.2% compared with 53.5% (both sexes; w2 (1 d.f.): 24.709, po0:001). These east–west differences were largely attributable to differences at younger age groups (under 45) among whom those in the east rated their health substantially less often as less than good than those in the west. Among those aged 45+, east–west differences were less coherent. Women were generally more likely to report less than good health than men in either east or west with those in the west reporting worse health at almost all ages. Also, health declined with increasing age in both regions and both sexes. By 1997, the proportion of those reporting less than good health had increased in both sexes and both regions, although more substantially so in the east, by almost 12% in men and 8% in women compared with, respectively, 3% and 2% in the west. As a consequence, the east–west differences of combined figures for less than good health became almost identical at 56.6% in the east and 55.9% in the west (w2 (1 d.f.): 0.383, p ¼ 0:536). Proportions of those reporting less than good health increased at all ages, but, in the east, particularly so amongst younger subjects where prevalence rates almost doubled. Inequalities in self-perceived health: 1992 As indicated above, increasing age was significantly associated with risk of poor health although the adverse impact of increasing age was stronger in the east than in the west (Table 3). Female sex was also associated with poor health in both regions although failing to achieve

ARTICLE IN PRESS 124

E. Nolte, M. McKee / Social Science & Medicine 58 (2004) 119–136

Table 2 Descriptive characteristics of the (a) east and (b) west German study population aged 25+, 1992 and 1997 Men

Women

1992

1997

1992

1997

%

(number)

%

(number)

%

(number)

%

(number)

(a) East Germany Self-perceived health Very good Good Satisfactory Less good Poor

13.3 44.5 29.5 8.9 3.8

(117) (392) (260) (78) (33)

6.3 40.0 39.3 11.1 3.3

(55) (350) (344) (97) (29)

10.2 38.1 32.6 14.1 4.6

(107) (398) (341) (148) (48)

4.5 36.3 39.3 13.9 6.0

(47) (375) (405) (144) (62)

Less than good health by age 25–34 35–44 45–49 50–64 65+

17.2 25.9 43.7 62.1 83.5

(39) (48) (78) (103) (104)

30.7 42.9 56.6 73.8 78.4

(66) (89) (84) (123) (107)

19.8 27.5 51.8 69.6 83.2

(44) (53) (99) (115) (226)

30.3 42.2 61.9 75.7 84.2

(72) (81) (91) (148) (219)

All ages

42.2

(372)

53.7

(470)

51.5

(537)

59.2

(611)

Marital status Single Married/cohabiting Separated/divorced/widowed

9.6 84.7 5.8

(84) (748) (51)

12.4 79.5 8.1

(108) (696) (71)

6.9 69.5 23.6

(72) (726) (247)

6.6 68.1 25.2

(69) (705) (261)

Income (% of median income) >150% 100–o150% 60–o100% o60%

12.6 41.6 38.7 7.1

(109) (358) (334) (61)

10.8 41.3 38.9 9.0

(92) (353) (333) (77)

9.1 38.7 43.9 8.3

(92) (394) (446) (84)

8.1 41.5 39.7 10.7

(81) (414) (397) (107)

Educational attainment Very high High Medium Low

11.8 13.0 72.3 2.9

(104) (114) (635) (25)

14.0 12.6 71.4 2.0

(122) (110) (621) (17)

6.5 16.9 60.0 16.6

(67) (174) (619) (171)

7.6 16.9 63.4 12.2

(77) (172) (647) (124)

Employed at present Yes No

64.5 35.5

(569) (314)

62.3 37.7

(545) (330)

46.6 53.4

(486) (559)

46.3 53.7

(479) (555)

Social resources Purposeful social involvement Less than median

65.3

(559)

63.3

(543)

70.6

(720)

71.3

(715)

Leisure-cultural social involvement Less than median 48.7

(416)

52.8

(453)

62.9

(641)

62.8

(629)

Network-oriented social involvement Less than median 65.3

(559)

60.7

(520)

69.5

(709)

64.5

(646)

31.4

(276)

21.5

(188)

36.5

(379)

30.5

(315)

41.5

(365)

60.4

(527)

45.0

(468)

62.9

(648)

Personal resources Barely able to cope Probable Confidence about future Improbable

ARTICLE IN PRESS E. Nolte, M. McKee / Social Science & Medicine 58 (2004) 119–136

125

Table 2 (continued) Men

Women

1992

1997

1992

1997

%

(number)

%

(number)

%

(number)

%

(number)

(b) West Germany Self-perceived health Very good Good Satisfactory Less good Poor

12.5 39.1 31.7 11.4 5.2

(417) (1471) (1191) (430) (196)

7.5 41.5 33.5 13.3 4.3

(278) (1538) (1243) (492) (158)

10.0 32.0 35.2 16.1 6.7

(442) (1419) (1560) (714) (296)

6.9 33.2 36.3 18.5 5.1

(301) (1454) (1594) (812) (223)

Less than good health by age 25–34 35–44 45–49 50–64 65+

26.3 37.1 49.0 69.0 71.8

(247) (290) (361) (426) (491)

26.5 39.9 55.4 67.5 82.2

(249) (345) (374) (400) (527)

27.7 41.4 52.6 75.3 83.2

(247) (341) (377) (531) (1074)

33.5 45.3 61.3 72.8 84.3

(335) (387) (400) (490) (1018)

All ages

48.3

(1815)

51.0

(1894)

58.0

(2570)

60.0

(2630)

Marital status Single Married/cohabiting Separated/divorced/widowed

15.6 75.5 8.4

(587) (2840) (316)

14.7 77.8 7.4

(548) (2894) (276)

9.3 64.7 26.0

(413) (2862) (115)

9.2 65.3 25.5

(405) (2872) (1125)

Income (% of median income) >150% 100–o150% 60–o100% o60%

19.0 36.7 37.6 6.7

(657) (1272) (1304) (231)

22.3 31.6 35.9 10.2

(779) (1106) (1254) (357)

15.2 34.4 40.6 9.8

(609) (1379) (1632) (395)

17.8 32.0 38.2 12.0

(717) (1284) (1533) (484)

Educational attainment Very high High Medium Low

16.0 9.9 61.5 12.6

(593) (365) (2278) (466)

17.0 10.1 61.0 11.9

(623) (368) (2231) (434)

7.1 7.2 55.0 30.7

(310) (316) (2412) (1344)

8.8 9.6 54.9 26.7

(383) (416) (2392) (1164)

Employed at present Yes No

69.1 30.9

(2598) (1161)

68.6 31.4

(2551) (1166)

43.5 56.5

(1931) (2506)

45.8 54.2

(2014) (2388)

Social resources Purposeful social involvement Less than median

45.4

(1646)

45.5

(1640)

45.7

(1919)

46.0

(1905)

Leisure-cultural social involvement Less than median 41.3

(1497)

44.0

(1585)

53.5

(2248)

51.9

(2150)

Network-oriented social involvement Less than median 46.1

(1671)

46.2

(1665)

46.0

(1932)

47.6

(1970)

14.2

(497)

19.0

(695)

19.4

(811)

20.5

(885)

38.3

(1345)

50.0

(1848)

44.8

(2311)

49.0

(2133)

Personal resources Barely able to cope Probable Confidence about future Improbable

ARTICLE IN PRESS 126

E. Nolte, M. McKee / Social Science & Medicine 58 (2004) 119–136

Table 3 Determinants of health in east and west Germany: odds ratios (95% confidence limits) for less than good self-perceived health in 1992 East Germany

West Germany

Less than Adjusted good for age health and gender

Adjusted for age, gender and education

Less than Adjusted for good age and health gender

Adjusted for age, gender and education

%

OR

95% CI

OR

95% CI

%

OR

95% CI

OR 95% CI

Age group 25–34 35–44 45–54 55–64 65+

18.5 26.7 47.9 65.8 83.3

1 1.61 4.06 8.54 21.35

(1.16, 2.24) (2.96, 5.55) (6.15, 11.88) (14.95, 30.51)

1 1.64 3.92 7.74 18.31

(1.18, 2.29) (2.96, 5.39) (5.54, 10.82) (12.63, 26.56)

26.9 39.3 50.8 72.3 79.2

1 1.75 2.80 7.07 10.00

(1.51, (2.42, (6.03, (8.60,

1 1.79 2.72 6.57 9.30

Gender Male Female

42.2 51.5

1 1.22 (0.99, 1.51)

1 1.13 (0.91, 1.40)

48.3 58.0

1 1 1.29 (1.18, 1.42) 1.15 (1.05, 1.27)

Income >150% 100-o150% 60–o100% o60%

34.3 47.0 50.8 47.5

1 1.44 (1.00, 2.05) 1.79 (1.24, 2.59) 2.39 (1.45, 3.94)

1 1.36 (0.94, 1.95) 1.61 (1.10, 2.35) 2.01 (1.19, 3.38)

44.8 49.5 59.0 60.3

1 1 1.14 (0.99, 1.33) 1.03 (0.88, 1.19) 1.84 (1.59, 2.13) 1.58 (1.35, 1.83) 2.04 (1.65, 2.52) 1.63 (1.30, 2.03)

Marital status Male Single 21.4 Married/cohabiting 44.0 Separated/divorced/widowed 50.3

1 1.12 (0.62, 2.04) 1.01 (0.42, 2.42)

1 1.16 (0.63, 2.12) 1.04 (0.43, 2.52)

34.2 50.8 52.5

1 1 0.92 (0.74, 1.14) 0.94 (0.76, 1.17) 0.75 (0.55, 1.03) 0.76 (0.55, 1.05)

Female Single 51.7 Married/cohabiting 45.5 Separated/divorced/widowed 69.2

1 0.90 (0.50, 1.63) 0.79 (0.41, 1.52)

1 0.79 (0.43, 1.43) 0.67 (0.34, 1.30)

48.0 51.4 78.1

1 1 0.92 (0.72, 1.16) 0.84 (0.66, 1.07) 1.43 (1.09, 1.89) 1.31 (0.99, 1.73)

Educationn Very high High Medium Low

31.7 34.1 46.8 82.5

1 1.13 (0.73, 1.76) 1.44 (0.99, 2.09) 3.18 (1.84, 5.49)

1 1.00 (0.64, 1.57) 1.26 (0.86, 1.86) 2.47 (1.41, 4.35)

37.3 41.1 52.2 69.5

1 1 1.45 (1.16, 1.81) 1.34 (1.06, 1.69) 1.57 (1.34, 1.84) 1.45 (1.23, 1.72) 2.46 (2.05, 2.96) 2.03 (1.66, 2.49)

Currently employed Yes No

30.9 67.1

1 1.56 (1.20, 2.04)

1 1.41 (1.07, 1.85)

39.7 70.6

1 1 1.61 (1.42, 1.83) 1.48 (1.31, 1.69)

Social resources Purposeful social involvement Higher than median 45.1 Lower than median 48.8

1 1.11 (0.88, 1.39)

1 1.06 (0.84, 1.33)

50.5 56.0

1 1 1.26 (1.14, 1.39) 1.19 (1.07, 1.31)

Leisure-cultural social involvement Higher than median 31.1 Lower than median 60.5

1 1.92 (1.54, 2.40)

1 1.77 (1.41, 2.23)

38.1 69.2

1 1 2.11 (1.90, 2.34) 1.94 (1.74, 2.16)

Network-oriented social involvement Higher than median 40.5 Lower than median 51.0

1 1.25 (0.99, 1.57)

1 1.22 (0.97, 1.54)

45.7 61.5

1 1 1.55 (1.40, 1.71) 1.50 (1.36, 1.66)

2.02) 3.25) 8.28) 11.62)

(1.55, (2.34, (5.58, (7.98,

2.08) 3.16) 7.74) 10.85)

ARTICLE IN PRESS E. Nolte, M. McKee / Social Science & Medicine 58 (2004) 119–136

127

Table 3 (continued) East Germany

West Germany

Less than Adjusted good for age health and gender

Adjusted for age, gender and education

Less than Adjusted for good age and health gender

Adjusted for age, gender and education

%

OR

OR

%

OR

OR 95% CI

Personal resources Barely able to cope Improbable Probable

38.6 63.7

1 1.95 (1.56, 2.43)

1 1.86 (1.48, 2.33)

48.9 74.9

1 1 2.43 (2.10, 2.81) 2.23 (1.92, 2.58)

Confident about future Probable Improbable

40.0 56.5

1 1.83 (1.49, 2.26)

1 1.85 (1.50, 2.29)

43.9 66.4

1 1 2.21 (2.00, 2.45) 2.14 (1.93, 2.37)

n

95% CI

95% CI

95% CI

In the second model adjusted for age, sex and income.

statistical significance in the east, probably reflecting the smaller sample size (Table 2). Marital status was associated with poorer health in west German women only. Education and income were both important determinants of health in east and west although associations were stronger in the east, with those in the lowest income bracket showing, after adjustment for age and sex, a 2.4-fold increase in risk for less than good health compared with a 2.0-fold increase in risk in the west. Addition of education reduced the strength of the relationship but the risk of less than good health among the poorest was still twice as high in the east. Education had an independent effect on health and seemed to be more important than income in the west with, after controlling for age, gender and income those with high educational attainment showing a significantly elevated risk of poor health, at 34%, compared with those with very high educational attainment whereas, in the east, there was no difference between these two groups. Employment status was also significantly associated with health with those not employed at present showing a 41% (east) to 48% (west) elevated risk of poor health. Measures of social and personal resources were important determinants of health in both regions, with, in the age-sex-adjusted model, lack of leisure-cultural social involvement, coping ability or confidence about the future doubling the risk of poor health. However, ‘purposeful social involvement’ was significantly associated with poor health in the west only, although an inverse relationship with health was also seen in the east. These associations were generally only slightly attenuated when additionally controlling for education.

To assess whether indicators of social and personal resources mediate the effect of income on health, separate models were developed that controlled for these variables (Table 4). In both east and west, the negative impact of low income on health remained significant once coping ability had been controlled for, with the odds ratio for those having available less than 60% of median equivalent income declining from 2.01 to 1.77 in the east and from 1.63 to 1.51 in the west. Comparable effects were seen when adjusting for confidence, although the effect was smaller in the east. Including both indicators simultaneously in the model attenuates the risk in east Germany only slightly (ORo60% median income 1.73 [1.02, 2.95]95% CI). In the west the risk was reduced but remained statistically significant (ORo60% 1.34 [1.06, 1.71]95% CI). Controlling for leisure-cultural social involvement had an effect similar to the combined impact of coping and confidence. However, when simultaneous adjusting for coping and leisure-cultural social involvement the statistical significance of the association between income and health disappeared in the east although the indication of a dose–response relationship remained (Table 4). In the west, simultaneous adjustment for the two indicators further attenuated the income-health association. Inequalities in self-perceived health: 1997 By 1997, the age-related gradient in self-perceived health had become almost identical in the two parts of the country with those aged 65+ showing an about 10fold higher risk of less than good health compared with those aged 25–34 (Table 5). Compared with 1992, the

ARTICLE IN PRESS 128

E. Nolte, M. McKee / Social Science & Medicine 58 (2004) 119–136

Table 4 Odds ratios (95% confidence limits) of less than good health by income, controlled for indices of social and personal resources, east and west Germany, 1992 East Germany OR

West Germany 95% CI

OR

95% CI

(0.89, 1.86) (1.01, 2.17) (1.04, 2.99)

1 1.02 1.50 1.51

(0.88, 1.19) (1.28, 1.75) (1.19, 1.92)

Adjusted for age, gender, education and leisure-cultural social involvement Income (% of median income) >150% 1 100–o150% 1.32 (0.91, 1.92) 60–o100% 1.47 (0.99, 2.17) o60% 1.68 (0.98, 2.87)

1 0.98 1.41 1.32

(0.84, 1.14) (1.20, 1.65) (1.05, 1.67)

Adjusted for age, gender, education, coping and leisure-cultural social involvement Income (% of median income) >150% 1 100–o150% 1.26 (0.87, 1.84) 60–o100% 1.36 (0.92, 2.02) o60% 1.51 (0.88, 2.59)

1 0.97 1.34 1.24

(0.83, 1.14) (1.14, 1.57) (0.97, 1.58)

Adjusted for age, gender, education and coping ability Income (% of median income) >150% 1 100–o150% 1.29 60–o100% 1.48 o60% 1.77

association with gender had become even less important in the east while changing only little in the west. Unlike in1992, marital status had become a significant determinant of health in east German men. Being separated, divorced or widowed was associated with a more than two-fold increase in risk of less than good health compared to being single. Among west German women, not only being separated, divorced or widowed but also being married or cohabiting was associated with a significantly elevated risk of reporting less than good health compared to being single. 1997 was also different in that income had become a less important predictor of health in the east whereas its importance appeared to have increased in the west. After controlling for age, gender and education, only those in the lowest income group, i.e. below the poverty line, showed an elevated risk of reporting average or poor health in the east. In the west, in contrast, a strong dose–response relationship had emerged showing a successively increasing risk with those in the lowest income group being more than twice as likely to report less than good health. Education, on the other hand, remained an independent determinant of health in both east and west, showing increasing risks of poor health with lower educational attainment similar to those seen in 1992. Moreover, the importance of being employed at present had increased in both regions.

The impact of indicators of social involvement remained strikingly similar to that seen in 1992 with, again, purposeful social involvement being statistically significant in the west only, although risks were small. The association of leisure-cultural social involvement with health had remained stable in the east whereas network-oriented social involvement had become a statistically significant predictor of health, increasing the risk for poor health for those being less socially integrated by about 50%. Similarly, personal resources remained significant determinants of health in both regions with odds ratios, having slightly increased in the east but somewhat declining in the west, especially those associated with coping. Looking further at the potential mediating effects of social and personal resources shows that income-related health inequalities disappear in the east while remaining statistically significant at all levels of income in the west (Table 6). In the east, after adjustment for coping, confidence or leisure-cultural social involvement in addition to age, gender and education, there was some indication of increased risk of less than good health at the lowest income level only. This was further reduced after the simultaneous inclusion of coping and confidence (ORo60% median income 1.07 [0.66, 1.75]95% CI). In the west, controlling for coping or confidence attenuated income-related health risks only slightly while adjusting

ARTICLE IN PRESS E. Nolte, M. McKee / Social Science & Medicine 58 (2004) 119–136

129

Table 5 Determinants of health in east and west Germany: odds ratios (95% confidence limits) for less than good self-perceived health in 1997 East Germany

West Germany

Less than Adjusted for good age and health gender

Adjusted for age, gender and education

Less than Adjusted for good age and health gender

Adjusted for age, gender and education

%

OR

95% CI

OR

95% CI

%

OR

95% CI

OR

95% CI

Age group 25–34 35–44 45–54 55–64 65+

30.5 42.6 59.2 74.8 82.2

1 1.70 3.32 6.78 10.41

(1.28, (2.45, (4.94, (7.51,

1 1.80 3.49 6.76 10.05

(1.35, (2.55, (4.93, (7.15,

30.1 42.6 58.3 70.3 83.6

1 1.73 3.28 5.51 11.49

(1.51, (2.83, (4.72, (9.82,

1 1.74 3.17 5.02 10.22

(1.52, (2.73, (4.28, (8.70,

Gender Male Female

53.7 59.2

1 1.11

(0.91, 1.35)

1 1.05

(0.86, 1.29)

51.0 60.0

1 1.28

(1.16, 1.41)

1 1.17

(1.06, 1.29)

Income >150% 100–o150% 60–o100% o60%

51.7 59.8 54.5 54.5

1 1.28 1.27 1.63

(0.89, 1.83) (0.88, 1.82) (1.04, 2.56)

1 1.11 1.08 1.35

(0.76, 1.62) (0.73, 1.59) (0.84, 2.17)

44.6 56.0 57.9 62.3

1 1.52 1.89 2.65

(1.32, 1.75) (1.64, 2.17) (2.19, 3.21)

1 1.43 1.69 2.23

(1.24, 1.65) (1.46, 1.95) (1.83, 2.71)

Marital status Male Single 37.1 Married/cohabiting 54.0 Separated/divorced/widowed 75.7

1 0.92 2.56

(0.58, 1.45) (1.26, 5.22)

1 0.88 2.34

(0.55, 1.41) (1.14, 4.82)

37.2 52.4 64.5

1 0.85 1.08

(0.68, 1.06) (0.77, 1.52)

1 0.84 1.02

(0.67, 1.05) (0.72, 1.43)

Female Single 53.0 Married/cohabiting 54.4 Separated/divorced/widowed 73.7

1 1.04 1.08

(0.59, 1.84) (0.57, 2.05)

1 0.87 0.85

(0.48, 1.55) (0.44, 1.64)

46.3 55.3 76.7

1 1.39 1.74

(1.10, 1.76) (1.33, 2.28)

1 1.30 1.63

(1.03, 1.65) (1.24, 2.14)

Educationn Very high High Medium Low

43.1 49.9 57.2 81.6

1 1.35 1.68 2.31

(0.91, 2.01) (1.21, 2.33) (1.32, 4.02)

1 1.28 1.59 2.19

(0.85, 1.92) (1.12, 2.25) (1.22, 3.93)

40.6 41.6 56.1 72.3

1 1.41 1.72 2.54

(1.15, 1.73) (1.48, 1.99) (2.11, 3.04)

1 1.27 1.48 2.01

(1.03, 1.58) (1.26, 1.73) (1.65, 2.45)

Currently employed Yes No

42.3 73.2

1 1.72

(1.34, 2.21)

1 1.58

(1.22, 2.05)

42.0 73.8

1 1.80

(1.59, 2.04)

1 1.68

(1.48, 1.90)

Social resources Purposeful social involvement Higher than median 55.1 Lower than median 56.9

1 1.02

(0.83, 1.27)

1 0.92

(0.74, 1.15)

53.5 57.0

1 1.21

(1.10, 1.34)

1 1.15

(1.04, 1.27)

Leisure-cultural social involvement Higher than median 41.5 Lower than median 67.0

1 1.80

(1.46, 2.23)

1 1.74

(1.40, 2.16)

41.2 70.1

1 2.03

(1.83, 2.25)

1 1.85

(1.66, 2.06)

Network-oriented social involvement Higher than median 45.8 Lower than median 62.6

1 1.55

(1.25, 1.91)

1 1.53

(1.24, 1.89)

48.0 63.1

1 1.49

(1.35, 1.64)

1 1.50

(1.35, 1.65)

2.25) 4.52) 9.24) 14.43)

2.40) 4.76) 9.28) 14.12)

1.98) 3.80) 6.43) 13.44)

2.00) 3.68) 5.88) 12.00)

ARTICLE IN PRESS E. Nolte, M. McKee / Social Science & Medicine 58 (2004) 119–136

130 Table 5 (continued)

East Germany

West Germany

Less than Adjusted for good age and health gender

Adjusted for age, gender and education

Less than Adjusted for good age and health gender

Adjusted for age, gender and education

%

OR

95% CI

OR

95% CI

%

OR

95% CI

OR

95% CI

Personal resources Barely able to cope Improbable Probable

50.3 74.2

1 2.09

(1.63, 2.67)

1 1.94

(1.52, 2.49)

51.8 71.7

1 1.96

(1.72, 2.23)

1 1.85

(1.62, 2.12)

Confident about future Probable Improbable

42.8 65.2

1 2.15

(1.75, 2.64)

1 2.11

(1.71, 2.60)

44.6 67.3

1 2.14

(1.94, 2.36)

1 2.02

(1.83, 2.24)

n

In the second model adjusted for age, sex and income.

Table 6 Odds ratios (95% confidence limits) of less than good health by income, controlled for indices of social and personal resources, east and west Germany, 1997 East Germany OR

West Germany 95% CI

OR

95% CI

(0.74, 1.59) (0.70, 1.53) (0.77, 2.01)

1 1.42 1.66 2.11

(1.23, 1.64) (1.43, 1.92) (1.73, 2.58)

Adjusted for age, gender, education and leisure-cultural social involvement Income (% of median income) >150% 1 100–o150% 1.06 (0.71, 1.56) 60–o100% 0.95 (0.64, 1.41) o60% 1.21 (0.74, 1.97)

1 1.34 1.55 1.83

(1.16, 1.56) (1.33, 1.80) (1.49, 2.24)

Adjusted for age, gender, education, coping and leisure-cultural social involvement Income (% of median income) >150% 1 100–o150% 1.04 (0.71, 1.54) 60–o100% 0.93 (0.62, 1.38) o60% 1.14 (0.70, 1.87)

1 1.34 1.50 1.73

(1.15, 1.55) (1.29, 1.75) (1.41, 2.12)

Adjusted for age, gender, education and coping ability Income (% of median income) >150% 1 100–o150% 1.08 60–o100% 1.04 o60% 1.24

for leisure-cultural social involvement reduced the odds ratio for the lowest income group from 2.23 to 1.83. Simultaneous inclusion of coping and leisure-cultural social involvement further reduced income-related health inequalities in the west although odds ratios remained statistically significant at all levels of income (Table 6). In the east, this had a smaller effect than simultaneous inclusion of coping and confidence. Finally, separate analyses showed that in both east and west network-oriented social involvement had an

effect similar to coping ability whereas purposeful social involvement had no effect in both 1992 and 1997.

Discussion This study provides important new information on several levels. First, it describes the changing pattern and determinants of self-perceived health in both parts of Germany since unification. In 1992, health was poorer in

ARTICLE IN PRESS E. Nolte, M. McKee / Social Science & Medicine 58 (2004) 119–136

west Germany than in the east but the gap closed by 1997. While income and education are important determinants of health in both east and west, in the west, income-related health inequalities widened between 1992 and 1997 yet the initially stronger gradient declined in the east. In contrast, the impact of education remained stable. Thirdly, variables related to personal and social resources were important determinants, mediating the effects of income, with leisure-cultural social involvement exerting the strongest effect in both east and west. The finding of poorer health in west Germany in 1992 appears somewhat surprising given that mortality rates were (and still are) higher in the east than in the west (Nolte, Shkolnikov, & McKee, 2000). Thus, before discussing the results further it is necessary to reflect on how these findings compare with other published data. Self-perceived health in Germany: is there an east–west difference? In an analysis of determinants of health satisfaction in the former GDR and the old FRG, Dehlinger and Ortmann (1992) showed, that, in 1989/1990, those in the west reported slightly more often being satisfied with their health (higher than average on a ten-point scale) than those in the east, at 66.1% versus 64.7%. No such east–west difference with regard to average level of health was reported in a subsequent study of health satisfaction in 1992 (Fuchs, 1995). As both studies analysed different years of the GSOEP, their findings may imply converging trends of health satisfaction between the two regions within a very short period after unification, although overall differences were not large. However, health satisfaction is not widely used in the international literature (Idler & Benyamini, 1997) and while it is significantly correlated with self-perceived health, the match is inexact (r2 ¼ 0:56). This view receives some support from two independent analyses of the 1992 GSOEP data set, the present analysis and a study by Winkelhage, Mielck, and John (1997), both reporting significant east–west differences in self-perceived health with poorer health reported in the west. In contrast, in a recent study of the adult population aged 18+ in the city of Berlin in 1991, east Berlin residents were significantly more likely to assess their health as poor or bad than their west Berlin counterparts (ORage-adjusted 1.29) (Hillen, Schaub, Hiestermann, Kirschner, & Robra, 2000). This difference was largely attributable to differences among those aged over 60. Levels of more narrowly defined ‘less than good health’ which also includes average levels of self-perceived health were, however, quite similar in east and west Berlin men, at 49.5% and 49.9%, but substantially higher in west Berlin women, at 60.6%, than in the east, at 55.9%, similar to the findings in the present analysis.

131

However, Mielck et al. (2000) did not find significant differences in health or in the prevalence of chronic conditions in the adult population aged 25–69 in the two parts of Germany in 1990–1992. Still, health among men aged 35–59 and women aged 25–39 tended to be poorer in the west while among women aged 45–69 the reverse was true. Luschen, . Geling, JanXen, Kunz, and von dem Knesebeck (1997) reported men in the east in 1992 rating their health best followed by women and men in the west while east German women rated their health worst. Differences were, however, small. They found gender generally to be significantly and negatively associated with self-perceived health in the east but not in the west. Although different in the relative ‘rank’ of different subgroups, their measure of mean levels of perceived health as assessed on a five-point scale compared fairly well with our findings. Analyses of more recent data again provide conflicting results. One study of a representative sample of the German adult population (age 18–92) in 1994 reported significant east–west differences in self-perceived health although the direction of these differences is not quite clear from the published material (Laubach, Schumacher, Mundt, & Br.ahler, 2000). A further study of the adult population aged 16+ in 1996 also found significant east–west differences in self-perceived health, with those in the east being more likely to rate their health worse than those in the west (Hessel, Geyer, . Plottner, Schmidt, & Br.ahler, 1999). Also, east Germans were more likely to report being limited in their daily activities by their health status than west Germans. In contrast, no east–west differences in self-rated health among those aged 18–79 could be inferred from the most recent data from the Federal Health Survey 1998, although no actual figures were reported (Knopf, Ellert, & Melchert, 1999). This situation is further complicated by data from an annual survey of living conditions in east Germany, one of the few studies that have been able to look at changes in self-perceived health over time (Schmidtke, 1999). Using a slightly different concept of self-rated health in which respondents were asked to assess their health as ‘healthy and fit’ (gesund und leistungsfahig), ‘slightly . hindered’ (gelegentlich leicht beeintrachtigt), ‘sustained . somatic complaints’ (dauernd gesundheitliche Beschwerden) or ‘limiting illness’ (stark behinderndes Leiden) it showed that the proportion of those perceiving themselves as healthy and fit increased from 26% in 1990 to over 50% in 1999 while the proportion of those suffering from sustained somatic complaints or limiting illness declined from 27% to 18%. However, these changes mainly occurred between 1990 and 1992. Since 1992, the situation has remained fairly stable. This trend is not consistent with the present analysis and findings from the GSOEP reported elsewhere (Bruckner, . 1998) that

ARTICLE IN PRESS 132

E. Nolte, M. McKee / Social Science & Medicine 58 (2004) 119–136

suggest an increase in levels of less than good health in the 1990s in both east and west. However, the data are not entirely comparable as they use different concepts of self-perceived health. In summary, available evidence on self-perceived health in east and west Germany is still not conclusive as to whether east–west differences actually do exist and if so, which direction these differences take. Published data seem to suggest that immediately after unification, health was better in east German men while for women the evidence is less clear. While inconsistencies are likely to reflect differences in study design and period, they do not provide a basis for rejecting what is otherwise a somewhat counter-intuitive finding. Potential explanations for east–west differences in levels of self-perceived health and changes over time The substantial increase in levels of less than good perceived health in east Germany between 1992 and 1997 may appear somewhat surprising, specifically as mortality levels have declined substantially during the same period (Nolte et al., 2000). While a certain time lag would be expected before changes in self-perceived poor health exert a (measurable) impact on mortality, a recent study in Germany showed that levels of poor selfperceived health may increase probability of death among those aged 50 and over within a short period of two to three years (Schwarze et al., 2000). However, in the present analysis the steepest increase in levels of less than good health was recorded at younger ages, which, as probability of death is much lower at younger ages, may not have exerted noticeable effects on mortality. There are several possible explanations for the changes in levels of less than good self-perceived health in east Germany. Firstly, differences in attitudes or beliefs may have changed over time, which in turn could have impacted levels of self-perceived health. Thus, observed changes would have to be considered artefactual rather than real. As noted above, the increase in levels of self-perceived poor health in east Germany was specifically seen at ages under 45 who may represent a ‘new’ generation with beliefs different from ‘older’ generations, perhaps more closely reflecting attitudes or beliefs seen in the west (Trommsdorf, 1999). This would be consistent with the view of an ‘assimilation’ process towards west German attitudes or beliefs as has been discussed with regard to the sharp fertility decline in east Germany after unification (Conrad, Lechner, & Werner, 1996). It does, however, also raise the more general question on the comparability of measures of self-perceived health between populations. Self-perceived health has repeatedly been shown to be a valid and reliable measure of health (Lundberg & Manderbacka, 1996; Miilunpalo,

Vuoria, Pasanen, & Urponen, 1997; Martikainen et al., 1999). However, the concept has been justified within a broader definition of health not only as a physiological state, but also as a social construct (Blaxter, 1990) that represents a summary of information of different aspects of health considered important by the respondent, depending on contextual psychological and social factors (Miilunpalo et al., 1997; Manderbacka, 1998). The perception of ‘good’ (or poor) health may thus vary between populations in different socio-cultural settings or according to age and experience (Blaxter, 1990; Elkeles & Seifert, 1996). A comparative study of middleaged men in Rotterdam/The Netherlands and Kaunas/ Lithuania did, for example, find self-perceived health a significant predictor for mortality in the Dutch sample only (Appels, Bsoma, Grabauskas, Gostautas, & Sturmans, 1996). This was attributed to possible differences in the understanding of the actual meaning of this construct in the two samples. Whether such differences also apply to the east and west German samples in the present analysis is difficult to assess and has as yet not been studied systematically. One study showed that east Germans tend to report more often somatic complaints with defined morphological correlates such as angina, liver disease or hypertension, whereas west Germans reported more frequently functional limitations such as indigestion, migraine or constipation (Hessel et al., 1999). This might indicate differences in the perception of aspects of physical health although much more research is needed. A second explanation for the observed changes in selfperceived health in east Germany between 1992 and 1997 refers to possible changes in response behaviour, which, in turn, would suggest that the observed east– west difference in self-perceived health in 1992 is artefactual. This may be inferred from findings of the annual survey on living conditions in east Germany mentioned earlier that found substantial increases in self-reported good health between 1990 and 1992 whereas virtually no change was observed in subsequent years. If there were a ‘true’ change in self-perceived good (or poor) health one would expect the change to be more gradual rather than sudden. It has been suggested that, as population surveys were not common in the former GDR, east Germans were not used to being interviewed for research purposes and thus may initially have been reluctant to provide comprehensive information in surveys (Trommsdorf, 1999). Whether this also applies to more ‘simple’ measures such as self-perceived health certainly needs to be investigated further. Health inequalities in east and west Germany Finally, changes in contextual socio-economic and/or psychosocial factors may have impacted self-perceived health and thus contributed to the actual deterioration

ARTICLE IN PRESS E. Nolte, M. McKee / Social Science & Medicine 58 (2004) 119–136

in levels of health in both east and west but particularly so in the east. This explanation may in fact hold for west Germany where income-inequalities and income-related health inequalities appear to have widened between 1992 and 1997. It is, however, contradicted by the findings for east Germany where, despite an observed increase in income inequality, income-related health inequalities seemed to have narrowed. Education, on the other hand, remained a rather strong determinant of health in both regions. This observation lends some support to the view that the association of socio-economic factors and health may differ between east and west, with education a stronger predictor in the east and income a more important determinant in the west (Luschen . et al., 1997; Mielck et al., 2000; van Dooslaaer et al., 1997). This was also suggested in other former communist countries of central and eastern Europe where education was found to be a strong determinant of various health outcomes (Bobak, Skodova, Pisa, Poledne, & Marmot, 1997; Koupilov!a, Bob!ak, HolW!ık, Pikhart, & Leon, 1998; Shkolnikov, Leon, Adamets, Andreev, & Deev, 1998; Jozan & Forster, 1999; Marmot & Bobak, 2000) although this has been challenged by others (Carlson, 1998; Hraba, Lorenz, PechaWov!a, & Liu, 1998; Palosuo, Uutela, Zhuravleva, & Lakomova, 1998). Income was usually not significantly related to health, whereas perceived measures of economic hardship such as material deprivation (Bobak, Pikhart, Rose, Hertzman, & Marmot, 2000; Gilmore, McKee, & Rose, 2002) and self-assessed economic situation were (Carlson, 1998). This has been explained, largely, by education determining position in society in adult life that, in turn, is related to determinants of differential health outcomes (Marmot & Bobak, 2000). On the other hand, in former communist societies, income had been distributed more equally and was also less important in getting access to scarce goods, which was more frequently based on connections with structures of power (Wnuk-Lipinski & Illsley, 1990). This has, however, changed after the political transition when, as in the west, access to luxury goods was increasingly based on individual wealth and earned income. Against this background it seems the more surprising that, in east Germany, an initial gradient in income-related health inequalities, present shortly after unification, had declined over time. Indicators of social involvement and personal resources were shown to be important determinants of health in east and west Germany with specifically lack of coping ability, confidence about the future and leisurecultural social involvement doubling the odds of poor health. The present analysis also suggests that, in 1992, these factors mediate the adverse impact of low income on health in both regions, whereas in 1997, in the east, income-related health inequalities disappear once cop-

133

ing, confidence or leisure-cultural social involvement had been controlled for. However, the lowest income group still showed an elevated, though statistically not significant, risk of poor health possibly indicating that being below a certain threshold, here relative poverty, poses a direct risk to self-perceived health. Furthermore, the relative impact of confidence had increased in the east, while in the west leisure-cultural social involvement still exerted the strongest mediating effect. Overall, these findings are in line with the view that psychosocial factors such as social involvement or personal resources mediate socio-economic differences in health (Wilkinson, 1996; Marmot & Bobak, 2000). The most interesting finding of the present analysis is the considerable role of leisure-cultural social involvement in mediating or ‘buffering’ the impact of income on selfperceived health whereas network-oriented social involvement was less important. Leisure-cultural social involvement includes active engagement in sports, which may be understood as an indicator of a healthier lifestyle. This could be related to better health in two ways. First, participating in sport may improve health. Second, selection bias may be acting, with those who are healthier more able to participate in sport. However, controlling for this last variable only attenuates the odds of less than good health by income only slightly. Purposeful social involvement, i.e. volunteer work or participation in local politics, showed some association with health in the west only. This is consistent with the study by Carlson (1998) who found a weak positive association between membership of an organisation (non-political) and self-rated health. It also showed that west European countries had a substantially higher degree of organisational membership. This is also seen in Germany where, in the 1990s, organisational membership was substantially higher in the west than in the east . 2000). In 1993, 44% of west Germans were (Schob, not member of any kind of organisation compared to 53% in the east, by 1998, in the west, this proportion had declined, to 42%, while increasing in the east, to 62%. However, the findings of the present study also indicate that while indicators of social involvement and personal resources might mediate the impact of socio-economic factors on health, there is still an independent effect of income, at least in the west, lending some support to the neo-materialist interpretation of the association of socio-economic factors and health (Ettner, 1996; Lynch, Davey Smith, Kaplan, & House, 2000; Wagstaff & van Doorslaar, 2000). Finally, one further finding is worth mentioning. In 1992, marital status was associated with health in west German women only, with those being separated, divorced or widowed being more likely to report less than good health than single women whereas marriage/

ARTICLE IN PRESS E. Nolte, M. McKee / Social Science & Medicine 58 (2004) 119–136

134

cohabiting tended to be protective, although this was not significant. This had, however, changed by 1997 when not only being divorced or widowed but also being married/cohabiting posed an increased risk to women in the west. No such an effect was seen among east German women. Among men, marriage had no impact on health although, in 1997, east German men who were divorced or widowed were at higher risk of poor health. The east– west difference in women may be attributed to the fact that the former GDR promoted greater gender equality than the west. Overall, this mixture of findings regarding the impact of socio-economic indicators and the possible mediating role of psychosocial factors on self-perceived health does lend further support to the view that mechanisms involved in these relationships may act differently in east and west. The weak association between income and health in east Germany in 1997 also suggests that income per se is possibly not an appropriate measure of socio-economic status in this part of the country. Further research is thus needed to investigate these associations in more detail, perhaps employing more subjective measures such as self-rated social status. This conclusion is partly supported by an analysis of the perception of social inequalities in Germany (Noll, 1998). It showed that, in 1996, about 55% west Germans considered themselves as ‘middle class’ and another 33% as ‘lower class’ or ‘workers’ class’ whereas in the east these figures were 39% and 59%, respectively, indicating substantial differences between east and west that may only partly be captured by conventional measures of socio-economic status.

Conclusions In conclusion, while there was an increase in income inequality in east Germany between 1992 and 1997, this was not accompanied by a simultaneous increase in income-related health inequalities. The reasons for this somewhat unexpected finding must remain speculative but it would appear that the change in income inequality was relatively low and thus had no measurable impact on indicators such as self-perceived health. In other former communist countries, income inequality as measured by the Gini coefficient rose quite substantially after the political transition, for example, between 1991 and 1996, by over 20% in the Czech Republic and Poland up to almost 50% in Russia and over 100% in Romania (UNICEF, 1999). In east Germany, the corresponding figure was only 8% (Krause & Habich, 2000). However, in west Germany, changes in income inequality were even less visible, with the Gini coefficient remaining fairly stable between 1991 and 1996, yet, income-related health inequality had become stronger. This again suggests that mechanisms involved in the association of socio-economic factors and health possibly behave differently in east and west. A longitudinal study is needed to confirm these preliminary findings and to ensure, for example, that the finding of an worsening in self-perceived health in east Germany is in fact real and not an artefact. Such a design would also allow investigating whether the apparent alleviating impact of leisure-cultural social involvement on the relationship between income and health is indeed due to a healthier lifestyle rather than vice versa.

Table 7 Men

Women

1992 Age East Germany 25–34 35–44 45–54 55–64 65+ All West Germany 25–34 35–44 45–54 55–64 65+ All

% 25.9 20.9 20.2 18.8 14.2 100

25.0 20.8 19.6 16.4 18.2 100

1997 Number 229 184 179 166 126 883

941 782 735 617 685 3759

% 24.6 23.7 17.0 19.1 15.6 100

25.2 23.3 18.2 16.1 17.3 100

1992 Number 215 208 149 167 137 875

938 865 675 598 642 3717

% 21.2 18.4 18.3 15.9 26.2 100

20.2 18.6 16.2 16.0 29.1 100

1997 Number 222 193 191 166 274 1045

895 824 719 708 1291 4437

% 23.1 18.6 14.2 18.9 25.2 100

22.7 19.4 14.8 15.3 27.8 100

Number 239 192 147 196 261 1034

999 854 652 674 1223 4402

ARTICLE IN PRESS E. Nolte, M. McKee / Social Science & Medicine 58 (2004) 119–136

Acknowledgements Ellen Nolte was supported by a European Commission TMR Fellowship, grant no. FMBICT983062. The authors are grateful to Dr. Ilkka M.akinen for his support and advice on methodological issues.

Appendix A Age distribution of the study populations in east and west Germany, 1992 and 1997, is shown in Table 7.

References Appels, A., Bsoma, H., Grabauskas, V., Gostautas, A., & Sturmans, F. (1996). Self-rated health and mortality in a Lithuanian and a Dutch population. Social Science & Medicine, 42, 681–689. Berkman, L. F., & Glass, T. (2000). Social integration, social networks, social support, and health. In L. F. Berkman, & I. Kawachi (Eds.), Social epidemiology (pp. 137–173). New York: Oxford University Press. Blaxter, M. (1990). Health and lifestyles. London: Routledge. Bobak, M., Skodova, Z., Pisa, Z., Poledne, R., & Marmot, M. (1997). Political changes and trends in cardiovascular risk factors in the Czech Republic, 1985–92. Journal of Epidemiology and Community Health, 51, 272–277. Bobak, M., Pikhart, H., Hertzman, C., Rose, R., & Marmot, M. (2000). Socioeconomic factors, perceived control and self-reported health in Russia. A cross-sectional survey. Social Science & Medicine, 47, 269–279. Bruckner, . G. (1998). Health expectancy in Germany: What do we learn from the reunification process?. Tokyo: Nihon University Population Research Institute. Buhmann, B., Rainwater, L., Schmaus, G., & Smeeding, T. M. (1988). Equivalence scales, well-being, inequality, and poverty: Sensitivity estimates across ten countries using the Luxembourg income study (LIS) database. Review of Income & Wealth, 34, 115–142. Burkhauser, R. V., Butrica, B. A., Dalz, M. C., & Lillard, D. (2000). The PSID-GSOEP equivalent file: A product of crosssectional research. Cornell Working Paper. Ithaka, New York: Cornell University. Carlson, P. (1998). Self-perceived health in east and west Europe: Another European health divide. Social Science & Medicine, 46, 1355–1366. Conrad, C., Lechner, M., & Werner, W. (1996). East German fertility after unification: Crisis or adaptation? Population Development Review, 22, 331–358. Dehlinger, E., & Ortmann, K. (1992). Gesundheitszufriedenheit in der Bundesrepublik Deutschland und der ehemaligen DDR—ein Vergleich. Gesundheitswesen, 54, 88–94. Department of Policy Analysis and Management, Cornell University, & German Institute for Economic Research (DIW), Berlin. (1999). GSOEP 1984–1997 and the crossnational equivalent file 1980–1997 GSOEP-PSID-SLID. Ithaca, NY: Cornell University.

135

Elkeles, T., & Seifert, W. (1996). Immigrants and health: Unemployment and health-risks of labour migrants in the federal republic of Germany, 1984–1992. Social Science & Medicine, 43, 1035–1047. Ettner, S. L. (1996). New evidence on the relationship between income, health. Journal of Health Economics, 15, 67–85. Ferraro, K. F., & Su, Y. P. (2000). Physician-evaluated and self-reported morbidity for predicting disability. American Journal of Public Health, 90, 103–108. Fuchs, J. (1995). BeeinfluXt Einkommen die Gesundheit? . Analysen mit Daten des Sozio-Okonomischen Panels. Gesundheitswesen, 57, 746–752. Gilmore, A. B. C., McKee, M., & Rose, R. (2002). Determinants of and inequalities in self-perceived health in Ukraine. Social Science & Medicine, 55, 2177–2188. . Hessel, A., Geyer, M., Plottner, G., Schmidt, B., & Br.ahler, E. (1999). Subjektive Einsch.atzung der eigenen Gesundheit und subjektive Morbidit.at in Deutschland. Psychotherapie Psychosomatische Medizin und Psychologie, 49, 264–274. Hillen, T., Schaub, R., Hiestermann, A., Kirschner, W., & Robra, B.-P. (2000). Self-rating of health is associated with stressful life events, social support and residency in east and west Berlin shortly after the fall of the Wall. Journal of Epidemiology and Community Health, 54, 575–580. Hoffmeister, H., Huttner, . H., Stolzenberg, H., Lopez, H., & Winkler, J. (1992). Sozialer Status und Gesundheit. bgaschriften 2/92. Munchen: . MMV Medizin Verlag. Hraba, J., Lorenz, F. O., PechaWov!a, Z., & Liu, Q. (1998). Education and health in the Czech Republic. Journal of Health and Social Behaviour, 39, 295–316. Idler, E. L., & Benyamini, Y. (1997). Self-rated health and mortality: A review of twenty-seven community studies. Journal of Health and Social Behaviour, 38, 21–37. . Jockel, K. H., Babitsch, B., Bellach, B. M., Bloomfield, K., Hoffmeyer-Zlotnik, J., Winkler, J., & Wolf, C. (1998). Empfehlungen der Arbeitsgruppe Epidemiologische Methoden in der Deutschen Arbeitsgemeinschaft Epidemiologie der Gesellschaft fur . Medizinische Informatik, Biometrie und Epidemiologie (GMDS) und der Deutschen Ge. Sozialmedizin und Pr.avention (DGSMP) zur sellschaft fur Messung und Quantifizierung soziodemographischer Merkmale in epidemiologischen Studien. In: W. Ahrens, B. M. . Bellach & K. H. Jockel (Eds.), Messung soziodemographischer Merkmale in der Epidemiologie. RKI-Schriften 1/ 98. Munchen: . MMV Medizin Verlag. Jozan, P., & Forster, D. P. (1999). Social inequalities and health: Ecological study of mortality in Budapest, 1980–3 and 1990–3. British Medical Journal, 318, 914–915. Kaplan, G. A., & Camacho, T. (1983). Perceived health and mortality: A nine year follow-up of the human population laboratory cohort. American Journal of Epidemiology, 117, 292–298. Knopf, H., Ellert, U., & Melchert, H.-U. (1999). Sozialschicht und Gesundheit. Gesundheitswesen, 61(Suppl. 2), S169–177. Koupilova, I., Bobak, M., Holcik, J., Pikhart, H., & Leon, D. A. (1998). Increasing social variation in birth outcomes in the Czech Republic after 1989. American Journal of Public Health, 88, 1343–1347. Krause, P., & Habich, R. (2000). Einkommen und Lebensqualit.at im vereinigten Deutschland. Vierteljahreshefte zur Wirtschaftsforschung, 69, 317–340.

ARTICLE IN PRESS 136

E. Nolte, M. McKee / Social Science & Medicine 58 (2004) 119–136

Kunzansky, L., & Kawachi, I. (2000). Affective states and health. In L. F. Berkman, & I. Kawachi (Eds.), Social epidemiology (pp. 213–241). New York: Oxford University Press. Laubach, W., Schumacher, J., Mundt, A., & Br.ahler, E. (2000). Sozialschicht, Lebenszufriedenheit und Gesundheitseinsch.atzung Ergebnisse einer repr.asentativen Unter. suchung der Bevolkerung. Sozial- und Praventivmedizin, 45, . 2–12. Lundberg, O., & Manderbacka, K. (1996). Assessing reliability of a measure of self-rated health. Scandinavian Journal of Social Medicine, 24, 318–324. Luschen, . G., Geling, O., JanXen, C., Kunz, G., & von dem Knesebeck, O. (1997). After unification: Gender and subjective health status in east and west Germany. Social Science & Medicine, 44, 1313–1323. Lynch, J. W., Davey Smith, G., Kaplan, G. A., & House, J. S. (2000). Income inequality and mortality: Importance to health of individual income, psychosocial environment, or material conditions. British Medical Journal, 320, 1200–1204. Manderbacka, K. (1998). Examining what self-rated health question is understood to mean by respondents. Scandinavian Journal of Social Medicine, 26, 145–153. Marmot, M., & Bobak, M. (2000). International comparators and poverty and health in Europe. British Medical Journal, 321, 1124–1128. . Martikainen, P., Aromaa, A., Heliovaara, M., Klaukka, T., Knekt, P., Maatela, J., & Lahelma, E. (1999). Reliability of perceived health by sex and age. Social Science & Medicine, 48, 1117–1122. McKee, M. (2001). The health effects of the collapse of the Soviet Union. In D. Leon, & G. Walt (Eds.), Poverty, inequality and health (pp. 17–36). Oxford: Oxford University Press. Mielck, A., Cavelaars, A., Helmert, U., Martin, K., Winkelhage, O., & Kunst, A. E. (2000). Comparison of health inequalities between east and west Germany. European Journal of Public Health, 10, 262–267. Miilunpalo, S., Vuoria, I., Pasanen, M., & Urponen, H. (1997). Self-rated health status as a health measure: The predictive value of self-reported health status on the use of physician services and on mortality in the working-age population. Journal of Clinical Epidemiology, 50, 517–528. Noll, H-H. (1998). Wahrnehmung und Rechtfertigung sozialer Ungleichheit 1991–1996. In H. Meulemann (Ed.), Werte und Nationalitat . im vereinten Deutschland (pp. 61–84). Opladen: Leske+Budrich. Nolte, E., Shkolnikov, V., & McKee, M. (2000). Changing mortality patterns in east and west Germany and Poland: II. Short-term trends during transition and in the 1990s. Journal of Epidemiology and Community Health, 54, 899–906. Palosuo, H., Uutela, A., Zhuravleva, I., & Lakomova, N. (1998). Social patterning of ill health in Helsinki and Moscow. Social Science & Medicine, 46, 1121–1136.

Pischner, R. (2000). Uberarbeitete Querschnittshochrechnung der Wellen G–N (1990–1997) des Sozio-oekonomischen Panels SOEP unter Einbeziehung der Erganzungsstichprobe E . (Welle O). Berlin: Deutsches Wirtschaftsinstitut. Schmidtke, H. (1999). Gesundheit. In G. Winkler (Ed.), Sozialreport 1999. Daten und Fakten zur sozialen Lage in den neuen Bundeslandern (pp. 269–279). Berlin: Sozialwis. senschaftiches Zentrum Berlin-Brandenburg. . A. (2000). Gesellschaftliche Beteiligung und Freizeit. In Schob, Statistisches Bundesamt (Ed.), Datenreport 1999. Zahlen und Fakten uber die Bundesrepublik Deutschland (pp. 530– . 540). Bonn: Bundeszentrale fur . politische Bildung. Schwarze, J., Andersen, H. H., & Anger, S. (2000). Self-rated health and changes in self-rated health as predictors of mortality—first evidence from German panel data. DIW Discussion Paper No. 203. Berlin: Deutsches Wirtschaftsinstitut. Shkolnikov, V. M., Leon, D. A., Adamets, S., Andreev, E., & Deev, A. (1998). Educational level and adult mortality in Russia: An analysis of routine data 1979 to 1994. Social Science & Medicine, 47, 357–369. Tabachnick, B. G., & Fidell, L. S. (1996). Using multivariate statistics. New York: Harper Collins College Publishers. Trommsdorf, G. (1999). Social change and individual development in east Germany: A methodological critique. In R. K. Silbereisen, & A. von Eye A (Eds.), Growing up in times of social change (pp. 171–199). Berlin/New York: Walter de Gruyter. UNICEF United Nations Children’s Fund. (1997). Children at risk in central and eastern Europe: Perils and promises. International Child Development Centre Regional Monitoring Report No. 4. Florence: UNICEF. UNICEF United Nations Children’s Fund. (1999). Women in transition. Regional Monitoring Report No. 6. Florence: United Nations Children’s Fund. van Dooslaaer, E., Wagstaff, A., Bleichrodt, H., Calonge, S., Gerdtham, U. G., Gerfin, M., Geurts, J., Gross, L., H.akkinen, U., Leu, R. E., O’Donnel, O., Propper, C., Puffer, F., Rodriguez, M., Sundberg, G., & Winkelhage, O. (1997). Income-related inequalities in health: Some international comparisons. Journal of Health Economics, 16, 93–112. Wagner, G., Burkhauser, R. V., & Behringer, F. (1993). The English language public use file of the German socioeconomic panel study. Journal of Human Resources, 28, 429–433. Wagstaff, A., & van Doorslaar, E. (2000). Income inequality and health: What does the literature tell us? Annual Review of Public Health, 21, 543–567. Wilkinson, R. G. (1996). Unhealthy societies: The afflictions of inequality. London: Routledge. Winkelhage, O., Mielck, A., & John, J. (1997). Einkommen, Gesundheit und Inanspruchnahme des Gesundheitswesens in Deutschland 1992. Sozial- und Praventivmedizin, 42, 3–10. . Wnuk-Lipinski, E., & Illsley, R. (1990). Introduction. Social Science & Medicine, 31, 833–836.