Buber_Engelhardt_englisch.indd The Association between Age and Depressive Symptoms among Older Men and Women in Europe. Findings from SHARE Isabella Buber, Henriette Engelhardt Abstract: Empirical evidence of the effects of age on depressive symptoms is mixed, ranging from positive to zero to negative effects, depending on the modelling of the age-depression profi le. This paper uses internationally comparative data to analyse the association between age and the prevalence of symptoms of depression, con- trolling for well-known determinants of mental health. Based on the fi rst wave of the Survey of Health, Ageing and Retirement in Europe (SHARE), depressive symptoms of 28,538 persons aged 50 to 89 from eleven European countries and Israel are analysed using a negative binomial regression model. The results indicate that the number of depressive symptoms measured by EURO-D scores increase with age and are higher among women than among men. When including socio-demograph- ic characteristics, health conditions and economic strains, the association between depressive symptoms and age vanishes for men, and even reverses for women. Thus, the association between age and mental health is mediated by the health and living conditions of older persons; age by itself has no explanatory power. Keywords: Depressive symptoms · Older persons · SHARE · EURO-D · Negative binomial model 1 Introduction “There is no health without mental health” (EC 2005: 4). Mental health is an indivis- ible part of health and mental problems, and can drastically reduce the quality of life of those effected and of their families. Good mental health is important for indi- viduals as well as for society. At individual level, it enables people to achieve their intellectual and emotional potential and to fi nd their roles in social and working life. At societal level, good mental health is important for social and economic welfare. Mental disorders are common. Estimates for the adult population in the EU af- fected by some form of mental ill-health and mental disorder within the past 12 months range between 20 % and 27 % (EC 2004b; Wittchen/Jacobi 2005). In the Comparative Population Studies – Zeitschrift für Bevölkerungswissenschaft Vol. 36, 1 (2011): 103-126 (Date of release: 26.05.2011) © Federal Institute for Population Research 2011 URL: www.comparativepopulationstudies.de DOI: 10.4232/10.CPoS-2011-02en URN: urn:nbn:de:bib-cpos-2011-02en9 • Isabella Buber, Henriette Engelhardt104 near future, depression will become the disease group with the second heaviest toll globally (EC 2004a). Accordingly, there is an increasing interest in the mental health of the EU population, and a strong political commitment to action in this fi eld. The most important forms of mental disorder are depression, specifi c phobias, somatoform disorders and alcohol dependence (Wittchen/Jacobi 2005). Depression and dementia constitute the two most important mental problems in later life (Cope- land et al. 1999b). Increasing life expectancy may contribute to a rise in prevalence among the aged. However, empirical evidence of the association between age and depressive symptoms is varied. Analyses of depression in later life (i.e. above age 65) reveal a strong positive association between the prevalence of symptoms of de- pression and age (Stordal et al. 2001; Castro-Costa et al. 2007), a modest association with age (Prince et al. 1999b), or a lack of an overall tendency towards depression rising with age (Trollor et al. 2007; Korten/Henderson 2000; Litwin 2002; Verropou- lou/Tsimbos 2007), except among the oldest old (Copeland et al. 1999b). Controlling for mediating effects, Blazer et al. (1991) and Berkman et al. (1986) found an inverse association between age and depression. Mediating variables generally transmit the effect of an independent variable on a dependent variable (Baron/Kenny 1986; MacKinnon et al. 2007). The different results presented in the literature, ranging from negative to zero to positive effects, may be due to selective samples or caused by different modelling methods of the age-depression profi le, in particular due to adjustments for different sets of covariates that have different infl uences on the age effect of depressive symptoms (Yang 2007). In this paper, we analyse the association between age and depression, control- ling for well-known determinants of mental health, using a representative sample of older persons in several European countries. In contradistinction to previous studies, we pay particular attention to the underlying statistical model and analyse systematically the change in the effect of age on depressive symptoms when con- trolling for these mediating variables. We also want to stress that the present study is not a psychiatric or gerontological contribution targeting physicians or psychia- trists, but addresses sociologists, demographers and the broad group of those who are interested in ageing. 2 Literature review Mental health has two dimensions: positive mental health and negative mental health. The positive dimension refers to the concepts of well-being and ability to cope in the face of adversity. The negative dimension includes psychological dis- tress and psychiatric disorders, and relates to the presence of symptoms. Positive and negative mental health cover different aspects. (EC 2004a). There are several measures for analysing mental health. Some instruments measure more generic factors, such as psychological distress, by recording the presence or absence of certain symptoms such as anxiety or depression. This type of instrument produces a mental health score, and for some of them, cut-off points can be used to categorise people into groups such as “probable cases” with mental The Association between Age and Depressive Symptoms among Older Men and Women • 105 health disorders.1 Other instruments are designed to produce answers which corre- spond to diagnoses of mental disorders (e.g., mood disorder, anxiety disorders and drug and alcohol disorders) and generate estimates of the prevalence of particular disorders2 (EC 2004a). There is a great wealth of literature on mental health, focussing inter alia on clinical aspects and treatments (e.g., Beck 1987; Beck 1991; Adam 1998; Drake et al. 2001; Amber et al. 2006), social and economic costs of mental health (e.g. Hamilton et al. 1997; Stephens/Joubert 2001; Whooley et al. 2002), health care services and their use (e.g., Alonso et al. 2004c, Harris et al. 2006) and the relation between men- tal and physical health (e.g., Braam et al. 2005; Opolski/Wilson 2005). The determinants of mental condition are multiple, including biological (e.g., ge- netics, sex), individual (e.g. health, personal experiences), familial and social (e.g., family status, social support), economic and environmental conditions (e.g., social status and living arrangements). The following brief overview of the literature fo- cuses on central variables associated with depression. In general, a higher number of depressive symptoms is typically found among women (Prince et al. 1999b; Alonso et al. 2004a; Lehtinen et al. 2005; Carta et al. 2005; Barry et al. 2008; Hopcroft/Bradley 2007; Zunzunegui et al. 2007). Copeland et al. (1999a) assessed the prevalence of depression among individuals aged 65 and over in nine European centres, and found that also women outnumber men among older persons with signs of depression. Their meta analysis shows an overall preva- lence of diagnostic depression of 12.3 % (14.1 % for women, and 8.6 % for men). Marital status is an important determinant of depressive symptoms: Widowed and divorced persons have poorer mental health (Dean et al. 1992; Lehtinen et al. 2003; Carta et al. 2005). Mental disorders are more common among persons who were either never married or who were previously married and currently have no partner (Alonso et al. 2004b; Buber/Engelhardt 2008; Schaan 2009). Having a trust- based relationship seems to have a protective effect. Several studies identify links between the prevalence of mental disorders and socio-economic disadvantages. In general, relatively high frequencies of mental disorders are associated with a poor level of education, material disadvantage, low family income, unemployment and living on a pension (e.g. Beekman et al. 1999; Alonso et al. 2004b; Fryers et al. 2005; Lehtinen et al. 2005; Carta et al. 2005; La- din 2008; Litwin/Sapir 2008). Consistent with analyses of European data, Kessler et al. (1994) fi nds elevated rates of affective and anxiety disorders among women and individuals with lower socio-economic status for the US. Other studies show a statistically signifi cant relationship between place of residence and mental health, with the lowest values being registered in large cities (Ayuso-Mateos et al. 2001; Lehtinen et al. 2003; Lehtinen et al. 2005). 1 Instruments in this category include Mental Health Index MHI-5, GHQ (General Health Question- naire) or EURO-D. 2 As an example we mention the CIDI (Composite International Diagnostic Interview) instru- ment. • Isabella Buber, Henriette Engelhardt106 Poor physical health is one of the most important risk factors for depression in older adults. Physical health problems are demonstrated to be a predictor of both the onset and the persistence of depression (e.g. Berkman et al. 1986; Katz 1996; Geerlings et al. 2000; Lenze et al. 2001; Fiske et al. 2003; Braam et al. 2005; Jang et al. 2007). Moreover, cognitive health also turns out to be associated with mental health (e.g. Jorm 2000; Reischies/Neu 2000; Scogin/Rohling 1989). Berkman et al. (1986) show that the addition of functional disability to a multivariate model sub- stantially weakens the association between age and depressive symptoms. Based on US data, Blazer et al. (1991) and Mirowsky and Ross (1992) show that the upward trend of depression in later life reverses after including as covariates mediating vari- ables such as marital, employment, economic and sociodemographic status. More recently, Cairney and Krause (2005) show that depressive symptoms in later life are associated with age, gender, living arrangements and education. They suggest that key social factors are related to depressive symptoms in later life. Thus, the ques- tion arises as to whether there is a true causal link between age and depressive symptoms. We hypothesize that this relationship is mediated by special circum- stances associated with the ageing process. The different and even contradictory fi ndings on the age-depression profi le in the literature may be due to different the consideration of different socio-demographic characteristics, health conditions and economic strains. To answer this question, we take advantage of the multifaceted structure of the Survey of Health, Ageing and Retirement in Europe (SHARE), a representative Euro- pean dataset that allows the comparison of health status in a variety of countries as well as the analysis of the determinants of health in a very broad context. SHARE in- cludes representative samples of the total population of eleven European countries and Israel. It makes it possible to study symptoms of mental health of Europeans aged 50 and older. In this paper, we analysed the association between age and depressive symp- toms – measured by EURO-D score – of persons aged 50 to 89, adjusting for living arrangements, education, economic constraints, limitations in activities of daily liv- ing, cognitive orientation, functional impairments and chronic diseases in a repre- sentative sample in eleven European countries and Israel. The SHARE data permit us to take into consideration these various dimensions which might be responsible for an age-specifi c increase in symptoms of depression. 3 Data, Variables And Method 3.1 Data The study is based on the fi rst wave of the Survey of Health, Ageing and Retirement in Europe (SHARE), which includes detailed cross-national information on health, well-being, economic circumstances and social networks for twelve countries, namely Austria, Belgium, Denmark, France, Germany, Greece, Israel, Italy, the Neth- erlands, Sweden, Switzerland and Spain. The data of the fi rst wave which we utilize The Association between Age and Depressive Symptoms among Older Men and Women • 107 were collected between 2004 and 2005. SHARE covers the non-institutionalised population aged 50 and older. “Release 2.0.1” of wave 1 comprises data on 31,115 individuals in 21,176 households, the weighted average response rate being 61.6 % (Börsch-Supan/Jürges 2005; see also http://www.share-project.org). The focus of the present study is on depressive symptoms of persons aged 50 to 89. Respondents aged 90 or older are excluded due to low numbers (285 per- sons aged 90 to 104). The current sample includes 28,538 persons (13,068 men and 15,470 women) with complete information on depressive symptoms, education and living arrangements. The mean age is 64 for men and 66 for women. 3.2 Variables Our central variable is depressive mood measured by the number of depressive symptoms. Our study measures mental health on the EURO-D scale. This is an in- strument which is symptom oriented, based on the presence or absence of depres- sive symptoms, but does not generate diagnoses of different mental disorders (e.g., mood disorder, anxiety disorders and drug and alcohol disorders). EURO-D was de- veloped in an 11-country European collaboration to compare symptoms of depres- sion in 14 European centres. Five depression measures3 are harmonised to form a 12-item scale (Prince et al. 1999a). The reliability of EURO-D has been reported to be good. In terms of its validity, it has been shown to correlate well with other well- known health measures (Prince et al. 1999a). The EURO-D is an internally consist- ent scale, captures the essence of its parent instruments, has been validated in a cross-European study of depression prevalence, and facilitates a valid comparison of risk factor associations between centres (Prince et al. 1999a). The 12 contribut- ing items for the EURO-D scale are: depression, pessimism, suicidality (wishing death), guilt, sleep, interest, irritability, appetite, fatigue, concentration, enjoyment, tearfulness. The time frame of the symptoms refers mostly to the month preced- ing the interview. The EURO-D is a discrete measure of depressive symptoms; the core ranges from 0 to 12, with higher scores indicating higher levels of depression. Dewey and Prince (2005) suggest to set a threshold at a score of 3 and defi ne clini- cally signifi cant depression as a EURO-D score greater than 3. In the current sample, EURO-D was internally consistent for all countries, with Cronbach alpha being 0.74 for the current pooled sample, ranging from 0.62 (in Switzerland) to 0.78 (in Spain). Thus, EURO-D is a feasible instrument for evaluating different dimensions of mental health. However, we are not able to distinguish between mild and severe mental disorders. To allow for a fl exible, non-parametric association between age and the number of depressive symptoms, 5-year age groups are included in the regressions. Based on the literature, we include in the analyses socio-demographic characteristics and 3 Geriatric Mental State-AGECAT (GMS-AGECAT), SHORT-CARE, Centre for Epidemiological Studies Depression scale (CES-D), Zung Self-Rating Depression Scale (ZSDS), Comprehensive Psychopathological Rating Scale (CPRS). • Isabella Buber, Henriette Engelhardt108 health conditions that were found to have an effect on mental health. The current study includes the following covariates: (a) living arrangements: alone; as a couple; alone with other persons; as a cou- ple with other persons, (b) highest level of education: primary school (ISCED 0-1); lower secondary (ISCED 2); higher secondary (ISCED 3-4); tertiary education (ISCED 5-6), (c) cognitive orientation based on the orientation to date, month, year and day of week; ranging from 0 (bad orientation) to 4 (good orientation), (d) limitations in the following activities of daily living (ADL): dressing; walking across a room; bathing or showering; eating; getting in and out of bed; using the toilet, (e) chronic diseases: no chronic diseases; mild chronic diseases (i.e. high blood pressure, high blood cholesterol, diabetes, asthma, osteoporosis, stomach, duodenal or peptic ulcer, cataracts or hip fracture), and severe chronic diseas- es (i.e. heart attack or chronic lung disease) (Kalwij/Vermeulen 2008), and (f) economic strain: subjective indicator of fi nancial distress, based on the ques- tion of how respondents make ends meet: with great diffi culty; with some diffi culty; fairly easily; easily. The distribution of these variables for the pooled sample is listed in Tables 2a and 2b below. With increasing age, the number of respondents goes down both for men and women. The vast majority of men and women live with a spouse or partner; while about one-third of women live alone, only 17 % of men do so. Differentiating by the highest educational level attained, it turns out that 56 % of all men and 42 % of women have completed higher secondary or tertiary education. Women report economic constraints more frequently than men (34 % of women and 28 % of men). When it comes to health, about 10 % of respondents report one or more limitations in their ADL; 15 % have a less than good cognitive orientation, and about 25 % of men and 21 % of women suffer from severe chronic diseases. 3.3 Statistical procedure In a fi rst step, the association between age and depressive symptoms is analysed by comparing means of EURO-D by age as well as by confi dence intervals. To com- plement the fi rst descriptive results, t-tests are used to estimate mean differences in EURO-D scores between men and women, with 95 % confi dence intervals. Next, multivariate regression models are applied to analyse the association between age and depressive symptoms, controlling for socio-demographic indicators (living ar- rangements, education and economic constraints), as well as diverse dimensions of health (cognitive orientation, limitations in activities of daily living, chronic diseas- es). To account for country-specifi c heterogeneity, we include country dummies in all models. Covariates are included stepwise so as to detect possible changes in the magnitude as well as in the direction of associations between age and depressive symptoms. As a general rule, the reference category is the largest group in each covariate. For the variable highest level of education, the largest group contains The Association between Age and Depressive Symptoms among Older Men and Women • 109 persons with a basic level of education. Since all respondents in Denmark have at least lower secondary education, “basic education” cannot be the reference group in this country. In order to have the same reference group in all countries, “higher secondary or tertiary education”, i.e. the highest educational level, is chosen as the reference group. Analyses are carried out separately for men and women to allow for a different shape of the association between age and the number of depressive symptoms, for a different constant in the estimated model, and for a different asso- ciation between the explanatory variable and the covariates included in the model. The explanatory variable of the current study is the number of depressive symp- toms, a discrete variable ranging from 0 to 12. It can be classifi ed as a count variable indicating how many depressive symptoms a respondent reported. In principle, we were able to analyse these data using standard multiple linear regression. However, the preponderance of zeros (n = 6,760; i.e. 24 %) and the small values indicate that the dependent variable is clearly discrete. The Poisson regression model accounts for these characteristics and is used widely to study such data. A common problem with the standard Poisson model is that the equidispersion assumption (E(Y|X) = V(Y|X) = λ) is violated, i.e. the conditional mean does not equal the conditional variance. To solve this problem, different approaches are proposed, including the generalised event count model, the generalised Poisson model and the negative binomial model to account for overdispersion (E(Y|X) < V(Y|X)) and underdisper- sion (E(Y|X) > V(Y|X)), with X and Y being random variables (Winkelmann 2003). A statistical test for dispersion reveals strong and signifi cant evidence of overdisper- sion, i.e. the conditional variance exceeds the conditional mean in the full sample. Therefore, we estimate a negative binomial model which accounts for overdisper- sion and for the prevalence of zero counts in the data (Winkelmann 2003). In this model, the probability Pr(y|x) of observing any observed count y is given by where X and Y are random variables, x is the vector of observed characteristics, is the gamma function and and μ are parameters to be estimated empirically. The parameter refl ects unobserved heterogeneity and determines the degree of dis- persion in the predictions. Systematic variation can be introduced in the parameter μ as in a log-linear model: μ=exp(x’β). The coeffi cients in this model cannot be in- terpreted directly; only the sign of a coeffi cient indicates the direction of an effect. We refer to Long and Freese (2006) and Winkelmann (2003) for further mathematical details. To visualise the fi t of the current count model with a negative binomial regression model, the observed relative frequencies, both the variable (number of depressive symptoms) and the predicted probabilities are plotted for each value of the count. Figure 1 clearly shows that the negative binomial distribution fi ts the data well and is an appropriate tool for the current study. y 11 1 1 1 1 )(!y )y( x|yPr)xX|yYPr( , (.) • Isabella Buber, Henriette Engelhardt110 4 Results At fi rst sight, the prevalence of depressive symptoms appears to increase with age among men and women, with women reporting more depressive symptoms compared to men. Figure 2 depicts the mean EURO-D scores for men and women. Roughly speaking, the mean EURO-D score increases between ages 50 to 89 from 1.6 to 3.0 among men and from 2.5 to 3.8 among women. Confi dence intervals become broader at older ages, indicating a considerable variation in the number of depressive symptoms at older ages. Table 1 includes the mean number of depressive symptoms among men and women within 5-year age groups and t tests estimating mean differences in EURO- D scores between men and women, as well as confi dence intervals indicating the difference. The results show that women reported signifi cantly more depressive symptoms than men, and, moreover, the gender gap widens with rising age (50-54: difference of 0.85 symptoms; 80-84: difference of 1.05 symptoms). It is therefore appropriate to run the analyses separately for men and women to allow for a differ- ent shape of the association between age and the number of depressive symptoms, Fig. 1: Observed relative frequencies and predicted probabilities for the number of depressive symptoms 0,00 0,05 0,10 0,15 0,20 0,25 0 1 2 3 4 5 6 7 8 9 10 11 12 Observed relative frequency Negative binomial prediction Relative frequency and probalbility Number of depressive symptoms Source: SHARE 2004-05, Release 2.0.1 The Association between Age and Depressive Symptoms among Older Men and Women • 111 for a different constant in the estimated model, and for a different association be- tween the explanatory variable and the covariates included in the model. For the multivariate analysis, we apply negative binomial regression models as described in the previous section. Tables 2a and 2b give the estimated non-expo- nentiated coeffi cients of the various variables separately for males and females, with the respective reference group having a value of zero. Positive coeffi cients imply an increase in the number of depressive symptoms; negative coeffi cients stand for a Fig. 2: Mean number of depressive symptoms by age and gender Tab. 1: Mean number of depressive symptoms by age groups and gender Men Women Mean difference (95%-Confidence Interval) 50-54 1.72 2.57 -0.85 (-0.97; -0.74) 55-59 1.70 2.47 -0.77 (-0.88; -0.66) 60-64 1.63 2.56 -0.92 (-1.04; -0.81) 65-69 1.77 2.66 -0.89 (-1.02; -0.77) 70-74 1.94 2.84 -0.90 (-1.05; -0.75) 75-79 2.29 3.25 -0.97 (-1.16; -0.76) 80-84 2.46 3.50 -1.05 (-1.29; -0.81) 85-89 2.63 3.46 -0.82 (-1.23; -0.42) Source: SHARE 2004-05, Release 2.0.1 Source: SHARE 2004-05, Release 2.0.1 • Isabella Buber, Henriette Engelhardt112 decrease. All models include additionally country dummies to account for country heterogeneity in the number of depressive symptoms (results not reported here). The association between age and number of depressive symptoms is approxi- mated with a piecewise constant function for 5-year age groups. This modeling pro- cedure allows a fl exible form and does not imply a specifi c association between the Tab. 2a: Coeffi cients of a negative binomial regression model for the associations between EURO-D scores and age, living arrangements, education, ADL limitations, cognitive orientation, health and economic strain, male respondents Model 1 Model A Model B Model C Model D Model E Model F Distrib. in % Age 50-54a 0 0 0 0 0 0 0 20 55-59 0.01 0.03 -0.00 0.01 0.00 0.01 -0.04 19 60-64 -0.00 0.02 -0.03 0.01 -0.02 -0.01 -0.11* 17 65-69 0.07 0.11* 0.03 0.08+ 0.05 0.05 -0.06 16 70-74 0.19*** 0.22*** 0.13* 0.20*** 0.14** 0.16** 0.03 12 75-79 0.30*** 0.32*** 0.23*** 0.34*** 0.20*** 0.25*** 0.10+ 9 80-84 0.41*** 0.43*** 0.34*** 0.43*** 0.21** 0.35*** 0.19** 5 85-89 0.53*** 0.50*** 0.44*** 0.55*** 0.30** 0.43*** 0.34** 2 Living arrangements Couple a 0 52 Ego alone 0.22*** 17 Couple with others 0.08* 28 Ego with others 0.34*** 3 Highest level of education Primary school 0.27*** 29 Lower secondary 0.11* 15 Higher secondary or tertiary a 0 56 Make ends meet With great difficulty 0.77*** 8 With some difficulty 0.40*** 20 Fairly easily 0.14** 27 Easily a 0 19 Missing answer 0.21*** 26 ADL limitations None a 0 91 1 and more limitations 0.77*** 9 Cognitive orientation 0 (bad orientation) to 3 0.33*** 15 4 (good orientation) a 0 85 Chronic diseases No -0.31*** 34 Mild a 0 41 Severe 0.30*** 25 Constant 0.49*** 0.40*** 0.43*** 0.24*** 0.44*** 0.78*** 0.60*** MacFadden’s R² 0.012 0.014 0.014 0.022 0.027 0.015 0.024 N 13,017 13,017 13,017 13,017 13,015 13,010 13,017 Notes: All models additionally include country dummies. The number of N differs in the various models due to missing data. a Reference category. + p<0.10; * p<0.05; ** p<0.01; *** p<0.001. Source: SHARE 2004-05, Release 2.0.1; own calculations The Association between Age and Depressive Symptoms among Older Men and Women • 113 number of depressive symptoms and age. This would be the case if, for example, age was entered for a linear relationship or age and age squared for a quadratic one. The sole association between age and depressive symptoms is confi rmed by the estimated signifi cant coeffi cients for ages 70 and above (Tab. 2a and 2b, Model 1). Tab. 2b: Coeffi cients of a negative binomial regression model for the associations between EURO-D scores and age, living arrangements, education, ADL limitations, cognitive orientation, health and economic strain, female respondents Model 1 Model A Model B Model C Model D Model E Model F Distrib. in % Age 50-54a 0 0 0 0 0 0 0 18 55-59 -0.04 -0.03 -0.06+ -0.04 -0.05 -0.04 -0.11** 16 60-64 0.00 0.01 -0.04 -0.00 -0.03 -0.01 -0.09** 15 65-69 0.06+ 0.05 0.00 0.05 0.03 0.05 -0.07+ 15 70-74 0.08* 0.07+ -0.00 0.06+ 0.01 0.06 -0.06+ 13 75-79 0.24*** 0.20*** 0.14*** 0.24*** 0.15*** 0.19*** 0.06 11 80-84 0.34*** 0.28*** 0.23*** 0.33*** 0.18*** 0.25*** 0.15** 9 85-89 0.30*** 0.22*** 0.19*** 0.30*** 0.04 0.19*** 0.14* 3 Living arrangements Couple a 0 40 Ego alone 0.14*** 33 Couple with others 0.04 17 Ego with others 0.18*** 10 Highest level of education Primary school 0.28*** 37 Lower secondary 0.20*** 21 Higher secondary or tertiary a 0 42 Make ends meet With great difficulty 0.55*** 11 With some difficulty 0.35*** 23 Fairly easily 0.12*** 25 Easily a 0 14 Missing answer 0.17*** 27 ADL limitations None a 0 89 1 and more limitations 0.54*** 11 Cognitive orientation 0 (bad orientation) to 3 0.31*** 15 4 (good orientation) a 0 85 Chronic diseases No -0.33*** 29 Mild a 0 50 Severe 0.30*** 21 Constant 0.92*** 0.86*** 0.83*** 0.71*** 0.89*** 1.20*** 1.05*** MacFadden’s R² 0.016 0.017 0.019 0.024 0.027 0.020 0.027 N 15,399 15,399 15,399 15,399 15,396 15,389 15,399 Notes: All models additionally include country dummies. The number of N differs in the various models due to missing data. a Reference category. + p<0.10; * p<0.05; ** p<0.01; *** p<0.001. Source: SHARE 2004-05, Release 2.0.1, own calculations • Isabella Buber, Henriette Engelhardt114 Thus, whilst controlling for country-specifi c heterogeneity, our results reveal in a fi rst step slightly higher levels for men and women aged 65-69, and signifi cantly higher levels of depressive symptoms in the age groups 70-74, 75-79, 80-84 and 85-89, compared to the reference group of respondents aged 50-54. The increase with age is more pronounced among men compared to women. In a next step, variables that are found to be correlated with depressive symp- toms are included separately in the models to analyze the change in size of the age coeffi cients, whilst controlling for these characteristics. Models A, B and C in Tables 2a and 2b include living arrangements, education, and economic constraints as co- variates separately, while Models D, E and F take health conditions into considera- tion. First of all, all covariates are signifi cantly associated with the number of depres- sive symptoms with the expected signs both for men and women. Living alone as well as living with people other than a partner, low levels of educational attainment, increased economic strains, ADL limitations, less than good cognitive orientation and severe chronic diseases are associated with an increased number of depressive symptoms. The best model fi t is measured by McFadden’s R2 results for fi nancial con- straints, chronic diseases and ADL limitations for both genders. For women, whilst controlling for each single covariate, the age effects diminish, with the exception of economic constraints. A signifi cantly increased number of depressive symptoms can only be observed for age 75 and older (Models A to E) and age 80 and older in Model F. For males, we observe only reduced age effects, whilst controlling for health conditions (ADL limitations, cognitive orientation, and chronic diseases). Liv- ing arrangements, level of education and economic strain do not alter the age ef- fects in the way we observe for females. Most interestingly, though, the estimated constant effect increases remarkably both for males and females in Models E and F, where cognitive impairment and chronic diseases are controlled for. Thus, the average number of depressive symp- toms increases, while the effects for the different age groups decline compared to Model 1. Moreover, controlling for chronic diseases, only men aged 75 and older and women aged 80 and older have signifi cantly increased numbers of depressive symptoms. The coeffi cients for age change signifi cantly with the stepwise inclusion of dif- ferent covariates. First, the initial strong positive association between age and the number of depressive symptoms weakens if one includes socio-demographic char- acteristics (Tab. 3, Model 2). Second, including as covariates diverse dimensions of health virtually eliminates the age-depressive symptoms association among men, and even reverses the association among women. While men and women aged 70 and above initially have signifi cantly higher levels of depressive symptoms com- pared to those in their early fi fties, in the model including health indicators (Mod- el 3) the estimated coeffi cients in the male sample are small and no longer signifi - cant (for age 70 and above). For females, all estimated coeffi cients turn negative and are even signifi cant for ages 55 to 74 and 85 to 89, the corresponding levels being between -0.09 and -0.13. The fi nal model includes socio-economic determinants as The Association between Age and Depressive Symptoms among Older Men and Women • 115 Tab. 3: Coeffi cients of a negative binomial regression model for the associations between EURO-D scores and age, living arrangements, education, ADL limitations, cognitive orientation, health and economic strain Men Women Model 2 Model 3 Model 4 Model 2 Model 3 Model 4 Age 50-54 a 0 0 0 0 0 0 55-59 0.02 -0.04 -0.03 -0.05+ -0.11*** -0.12*** 60-64 0.00 -0.10* -0.09+ -0.03 -0.11*** -0.13*** 65-69 0.08 -0.07 -0.05 -0.00 -0.09** -0.14*** 70-74 0.18*** 0.00 0.02 -0.02 -0.12*** -0.19*** 75-79 0.31*** 0.03 0.06 0.14*** -0.03 -0.09* 80-84 0.40*** 0.03 0.06 0.21*** -0.03 -0.11* 85-89 0.47*** 0.12 0.11 0.16** -0.13** -0.22*** Living arrangements Couple a 0 0 0 0 Ego alone 0.16*** 0.16*** 0.11*** 0.11*** Couple with others 0.04 0.04 0.03 0.02 Ego with others 0.28*** 0.28*** 0.10** 0.07* Highest level of education Primary school 0.15*** 0.09** 0.22*** 0.17*** Lower secondary 0.05 0.02 0.16*** 0.13*** Higher secondary or tertiary a 0 0 0 0 Make ends meet With great difficulty 0.71*** 0.60*** 0.48*** 0.40*** With some difficulty 0.37*** 0.33*** 0.30*** 0.26*** Fairly easily 0.13** 0.13** 0.10** 0.10** Easily a 0 0 0 0 Missing answer 0.22*** 0.17*** 0.17*** 0.15*** ADL limitations None a 0 0 0 0 1 and more limitations 0.63*** 0.57*** 0.45*** 0.41*** Cognitive impairment 0 (bad orientation) to 3 0.24*** 0.21*** 0.23*** 0.20*** 4 (good orientation) a 0 0 0 0 Chronic diseases No -0.28*** -0.27*** -0.30*** -0.28*** Mild a 0 0 0 0 Severe 0.20*** 0.25*** 0.19*** 0.18*** Constant 0.16*** 0.53*** 0.25*** 0.62*** 1.00*** 0.75*** MacFadden’s R² 0.025 0.037 0.047 0.027 0.039 0.047 N 13,017 13,008 13,008 15,399 15,386 15,386 Notes: All models additionally include country dummies. The number of N differs in the various models due to missing data. a Reference category. + p<0.10; * p<0.05; ** p<0.01; *** p<0.001. Source: SHARE 2004-05, Release 2.0.1; own calculations • Isabella Buber, Henriette Engelhardt116 well as health indicators (Model 4). Whereas no age effect is observed for men, the estimated coeffi cients for women are all negative and signifi cantly different from zero, indicating a lower number of depressive symptoms for ages 55 to 89 com- pared to women in their early fi fties. Figures 3a and 3b depict the association between age and depressive symp- toms for the stepwise setup of models. Figures 3a and 3b provide unambiguous evidence of the importance of socio-demographic characteristics and indicators of physical and cognitive health for depressive symptoms for both men and women, especially in advanced age. The piecewise constant functions representing the fi nal multivariate model reveal that the increase in depression with age initially observed almost disappears for men, and even reverses for women. Our results indicate that socio-demographic characteristics and physical as well as cognitive health absorbe, and even go so far as to reverse, the association between age and EURO-D scores among older persons. Moreover, the separate analysis of men and women reveals gender-specifi c differences in the association between age and living arrangements on the one hand and EURO-D scores on the other. The association between educa- Fig. 3a: Estimated association between age and the level of depression with and without covariates for socio-demographic and health characteristics, men -0,4 -0,2 0,0 0,2 0,4 0,6 50-54 55-59 60-64 65-69 70-74 75-79 80-84 85-89 Model 1 Model 2 Model 3 Model 4 Coefficients of regression models Age in years Note: The models refer to Tables 2a and 3. Source: SHARE 2004-05, Release 2.0.1 The Association between Age and Depressive Symptoms among Older Men and Women • 117 tion and depressive symptoms is stronger among women, whereas the estimated coeffi cients for economic constraints, ADL limitations and chronic diseases are higher among men. To test whether health, demographic characteristics and economic circumstanc- es are mediating variables as they carry the infl uence of our independent variable (age) on our dependent variable (mental health), we apply a Sobel-Goodman me- diation test. We fi nd that the mediation effect of ADL limitations is highly signifi - cant, with approximately 49 % of the total effect (of age on depressive mood) being mediated. In other words, age infl uences physical health, which in turn infl uences mental health. Gender-specifi c analysis reveals a more pronounced mediating ef- fect among women than among men (52 % versus 42 %). Also, educational status can be regarded as a signifi cant mediator, with 40 % of the total effect being medi- ated. Cognitive impairment and chronic diseases, but also living arrangements and economic constraints, turn out to carry to a lesser degree the infl uence of age on mental health. -0,4 -0,2 0,0 0,2 0,4 0,6 50-54 55-59 60-64 65-69 70-74 75-79 80-84 85-89 Model 1 Model 2 Model 3 Model 4 Coefficients of regression models Age in years Note: The models refer to Tables 2b and 3. Source: SHARE 2004-05, Release 2.0.1 Fig. 3b: Estimated association between age and the level of depression with and without covariates for socio-demographic and health characteristics, women • Isabella Buber, Henriette Engelhardt118 5 Discussion In this study we analyze the association between the prevalence of depressive symp- toms and age, including as covariates socio-economic characteristics and health. In contrast to existing studies which provide mixed empirical evidence, SHARE al- lows the analysis of depressive symptoms of persons living in private households in various European countries based on a representative sample and a standardised questionnaire. In order to assess depressive symptoms, we use the EURO-D scale and estimate the effects of age, whilst controlling for well-known determinants of mental health using a negative binomial regression model as an innovative tool in analysing the number of depressive symptoms. The present study reveals at fi rst sight a signifi cant increase in the number of depressive symptoms with age among both sexes. After including socio-economic characteristics and health conditions, the association almost disappears for men and even reverses for women. Living arrangements, educational level, fi nancial strains, limitations in activities of daily living, chronic diseases and cognitive ori- entation are major determinants of depressive symptoms among older persons. Depressive symptoms are highly correlated with circumstances that are associated with old age, such as health problems, fi nancial strains or the loss of one’s partner. Once we had included these socio-demographic characteristics and health as cov- ariates, age itself no longer had any explanatory power for men, and turns out to be positively correlated with depressive symptoms among women. Our fi ndings for European data are in line with fi ndings for the US by Berkman et al. (1986), Blazer et al. (1991), Mirowsky and Ross (1992) and Cairney and Krause (2005). In addition, gender-specifi c analyses reveal that the association between the number of depressive symptoms and age reverses for women, whereas it “only” vanishes for men. In a recent study based on the Greek SHARE data, Verropoulou and Tsimbos (2007:178) state that the positive association between age and depressive symp- toms may be “spurious and mainly due to the older suffering from disabilities and stressful life events such as bereavement”. Our evidence supports this hypothesis and additionally even reveals for women an inverse association with age, socio- demographic, economic and health indicators being included as covariates. The outcomes of the covariates are in line with previous research (e.g. Dean et al. 1992), but they also reveal remarkable gender differences related to education and living arrangements and ADL limitation. For example, the impact of education is more marked among women than among men. Additional analyses for men and women not shown here reveal that divorce and widowhood have a different impact on the mental health of persons aged 50 and older. For men, it is widowhood that had a stronger negative impact on mental health, while for women it is divorce (see also Buber/Engelhardt 2008). The stepwise introduction of covariates reveals that physical health is indeed an important aspect of mental health, in particular in old age. Accounting for socio- economic status attenuated the positive association between the number of de- pressive symptoms and age; only part of the increase observed with age can be The Association between Age and Depressive Symptoms among Older Men and Women • 119 explained by living arrangements, education and economic situation. The associa- tion between age and mental health is mediated by health determinants, especially among women. With the inclusion of health determinants the initially positive as- sociation vanishes for men and even turns negative for women, indicating a falling number of depressive symptoms with increasing age. Moreover, the measures for model fi t indicate that the differences observed with age are mainly due to physi- cal health (Verropoulou/Tsimbos 2007). Nevertheless, the direction of causality be- tween physical and mental health is not clear. In the current analysis we fi rst include as covariates socio-demographic charac- teristics, followed by health conditions. Alternatively, the covariates are introduced in a different order. It turns out that when including fi rst limitations in activities of daily living as well as cognitive orientation and later living arrangements, the re- sults remain stable. Therefore, both aspects – socio-demographic characteristics and health conditions – are independent determinants of mental health, association with the one group not being absorbed by the other. The unrefl ected use of gender, or sex, as a technical category is criticised by researchers on gender studies who argue that one must not neglect the substan- tial differences between being a man and being a woman (Jylhä 2007). The cur- rent study does not include gender as a covariate, but men and women are ana- lysed separately, allowing different associations between the number of depressive symptoms on the one hand, and age as well as socio-economic and health factors on the other. Indeed, the results for age, living arrangements, educational level, eco- nomic strain, ADL limitations and chronic diseases differ among men and women, either in size or in direction. Two limitations have to be mentioned. First, our analysis is based on self-rated health, not on diagnoses by psychiatrists or general practitioners. Regarding the structure of our data, we use a cross-sectional sample, and are therefore not able to disentangle period and cohort effects. As SHARE is designed as a longitudinal study, it will allow us to address this issue in the near future, when several waves are available to the scientifi c community. What is more, we are not able to investigate dynamic aspects and causalities with our cross-sectional data. Although the current analysis covers a range of different aspects, other dimen- sions such as social support, working conditions or transition to retirement are left out. Another potential confounder in the study of the association of age and depres- sive symptoms in a cross-sectional analysis is the cohort effect (Yang 2007). Moreo- ver, the analysis of the fi rst wave of SHARE does not allow us to detect causalities but only associations. A compression of morbidity is being observed as Europe ages (Vaupel 2010). Due to the increasing relative time spent in good health, the rise in the number of individuals suffering from depression should be less pronounced compared to a situation with prolonged morbidity. This holds true as long as economic strains and socio-demographic characteristics remain constant. • Isabella Buber, Henriette Engelhardt120 The authors would like two thank two anonymous reviewers from Comparative Population Studies for important suggestions. The collection of the data was pri- marily funded by the European Commission through the 5th framework programme (project QLK6-CT-2001-00360 in the thematic programme area “Quality of Life”). Ad- ditional funding came from the US National Institute on Aging (U01 AG09740-13S2, P01 AG005842, P01 AG08291, P30 AG12815, Y1-AG-4553-01 and OGHA 04-064). Data collection in Austria (through the Austrian Science Foundation, FWF), Belgium (through the Belgian Science Policy Offi ce) and Switzerland (through BBW/OFES/ UFES) was funded nationally. The SHARE data collection in Israel was funded by the US National Institute on Aging (R21 AG025169), by the German-Israeli Founda- tion for Scientifi c Research and Development (G.I.F.), and by the National Insurance Institute of Israel. Further support by the European Commission through the 6th framework program (projects SHARE-I3, RII-CT-2006-062193, and COMPARE, CIT5- CT-2005-028857) is gratefully acknowledged. References Adam, Christian 1998: Depressive Störungen im Alter. Epidemiologie und soziale Bedin- gungen. München: Juventa. Alonso, Jordi; Angermeyer, Matthias C.; Bernet, Sebastian et al. 2004a: Prevalence of mental disorders in Europe: results from the European Study of Epidemiology of Men- tal Disorders (ESEMeD) project. In: Acta Psychiatrica Scandinavica 109 (Suppl. 420): 21-27. [doi: 10.1111/j.1600-0047.2004.00327.x] Alonso, Jordi; Angermeyer, Matthias C.; Lépine, Jean-Pierre 2004b: The European Study of Epidemiology of Mental Disorders (ESEMeD) project: an epidemiological ba- sis for informing mental health policies in Europe. In: Acta Psychiatrica Scandinavica 109 (Suppl. 420): 5-7 [doi: 10.1111/j.1600-0047.2004.00325.x]. Alonso, Jordi; Angermeyer, Matthias C.; Bernet, Sebastian et al. 2004c: Use of mental health services in Europe: results from the European Study of Epidemiology of Men- tal Disorders (ESEMeD) project. In: Acta Psychiatrica Scandinavica 109 (Suppl. 420): 47-54 [doi: 10.1111/j.1600-0047.2004.00330.x]. Amber, M. Gum; Areán, Patricia; Hunkeler, Enid et al. 2006: Depression Treatment Pref- erences in Older Primary Care Patients. In: The Gerontologist 46,1: 14-22. Ayuso-Mateos, Jose Luis; Vázquez-Barquero, Jose Luis; Dowrick, Christopher et al. 2001: Depressive disorders in Europe: prevalence fi gures from the ODIN study. In: British Journal of Psychiatry 179: 308-316. Baron, Reuben M.; Kenny, David A. 1986: The Moderator-Mediator Variable Distinction in Social Psychological Research: Conceptual, Strategic, and Statistical Considera- tions. In: Journal of Personality and Social Psychology 51,6: 1173-1182. Barry, Lisa C; Allore, Heather G.; Guo, Zhenchao et al. 2008: Higher burden of depression among older women – The effect of onset, persistence, and mortality over time. In: Archives of General Psychiatry 65: 172-178 [doi: 10.1001/archgenpsychiatry.2007.17]. Beck, Aaron T. 1987: Cognitive models of depression. In: The Journal of Cognitive Psy- chotherapy: An International Quarterly, 1,1: 5-37. The Association between Age and Depressive Symptoms among Older Men and Women • 121 Beck, Aaron T. 1991: Cognitive therapy: A 30-year retrospective. In: American Psycholo- gist 46,4: 368-375. Beekman, Aartjan T.F.; Copeland, John R.M.; Prince, Martin J. 1999: Review of com- munity prevalence of depression in later life. In: British Journal of Psychiatry 174: 307-311. Berkman, Lisa F.; Berkman, Cathy S.; Kasl, Stanislav et al. 1986: Depressive symptoms in relation to physical health and functioning in the elderly. In: American Journal of Epidemiology 3: 372-388. Blazer, Dan; Burchett, Bruce; Service, Connie et al. 1991: The Association of Age and Depression Among the Elderly: An Epidemiologic Exploration. In: Journal of Geron- tology: Medical Sciences 46,6: M210-215. Börsch-Supan, Axel; Jürges, Hendrik (Eds.) 2005: The Survey of Health, Ageing and Retirement in Europe – Methodology. Mannheim: MEA Eigenverlag. Braam, Arjan W.; Prince, Michael J.A.; Beekman, Aartjan T.F. et al. 2005: Physical health and depressive symptoms in older Europeans. Results from EURODEP. In: British Journal of Psychiatry 187: 35-42. Buber, Isabella; Engelhardt, Henriette 2008: Children’s impact on the mental health of their older mothers and fathers: fi nding from the Survey of Health, Ageing and Retirement in Europe. In: European Journal of Ageing 5,1: 31-45 [doi: 10.1007/ s10433-008-0074-8]. Cairney, John; Krause, Neal 2005: The Social Distribution of Psychological Distress and Depression in Older Adults. In: Journal of Aging and Health 17,6: 807-835 [doi: 10.1177/0898264305280985]. Carta, Mauro Giovanni; Bernal, Mariola; Hardoy, Maria Carolina et al. 2005: Migration and mental health in Europe (the state of mental health in Europe working group: ap- pendix I). In: Clinical Practice and Epidemiology in Mental Health. URL: http://www. cpementalhealth.com/content/1/1/13, 26 April 2010 [doi:10.1186/1745-0179-1-13]. Castro-Costa, Erico; Dewey, Michael; Stewart, Robert et al. 2007: Prevalence of depres- sive symptoms and syndromes in later life in ten European countries. The SHARE study. In: British Journal of Psychiatry 191: 393-401 [doi: 10.1192/bjp.bp.107.036772]. Copeland, John R.M.; Beekman, Aartjan T.F; Dewey, Michael E. et al. 1999a: Depres- sion in Europe. Geographical distribution among older people. In: British Journal of Psychiatry 174: 312-321. Copeland, John R.M.; Beekman, Aartjan T.F; Dewey, Michael E. et al. 1999b: Cross- cultural comparison of depressive symptoms in Europe does not support stereotypes of ageing. In: British Journal of Psychiatry 174: 322-329. Dean, Alfred; Kolody, Bohdan; Wood, Patricia et al. 1992: The Infl uence of Living Alone on Depression in Elderly Persons. In: Journal of Aging and Health 4,1: 3-18. Dewey, Michael E.; Prince, Martin J. 2005: Mental Health. In: Börsch-Supan, Axel; Jür- ges, Hendrik (Eds.): Health, Ageing and Retirement in Europe - First Results from the Survey of Health, Ageing and Retirement in Europe. Mannheim: MEA Eigenverlag: 108-117. • Isabella Buber, Henriette Engelhardt122 Drake, Robert E.; Goldman, Howard H.; Leff, Stephen H. et al. 2001: Implementing Evidence-Based Practices in Routine Mental Health Service Settings. In: Psychiatric Services 52,2: 179-182. EC (European Commission) 2004a: The State of Mental Health in the European Union. European Commission, Health and Consumer Protection DG. URL: http://ec.europa. eu/health/ph_projects/2001/monitoring/fp_monitoring_2001_frep_06_en.pdf, 26 April 2010. EC (European Commission) 2004b: Action for Mental Health. Activities co-funded from the European Community Public Health Programmes 1997-2004. European Commis- sion, Health and Consumer Protection DG. URL: http://ec.europa.eu/health/archive/ ph_determinants/life_style/mental/docs/action_1997_2004_en.pdf, 26 April 2010. EC (European Commission) 2005: Green Paper. Improving the mental health of the pop- ulation: Towards a strategy on mental health for the European Union. European Com- mission, Health and Consumer Protection DG, Brussels, 14 October 2005, COM(2005) 484. Fiske, Amy; Gatz, Margaret; Pedersen, Nancy L. 2003: Depressive Symptoms and Ag- ing: The Effects of Illness and Non-Health-Related Events. In: Journal of Gerontology – Psychological Sciences 58B: P320-328 [doi: 10.1093/geronb/58.6.P320]. Fryers, Tom; Melzer, David; Jenkins, Rachel et al. 2005: The distribution of the common mental disorders: social inequalities in Europe. In: Clinical Practice and Epidemiology in Mental Health. URL: http://www.cpementalhealth.com/content/1/1/14, 25 April 2010 [doi:10.1186/1745-0179-1-14]. Geerlings, Sandra W.; Beekman, Aartjan T.F; Deeg, Dorly J.H. et al. 2000: Physical health and the onset and persistence of depression in older adults: an eight-wave prospec- tive community-based study. In: Psychological Medicine 30: 369-380. Hamilton, Viviane H.; Merrigan, Phillip; Dufresne, Éric 1997: Down and out: estimating the relationship between mental health and unemployment. In: Health Economics 6,4: 397-406. Harris, Katherine M.; Edlund, Marc J.; Larson. Sharon L. 2006: Religious involve- ment and the use of mental health care. In: Health Services Research 41,2: 395-410 [doi: 10.1111/j.1475-6773.2006.00500.x]. Hopcroft, Rosemary L.; Bradley, Dana Burr 2007: The sex difference in depression across 29 countries. In: Social Forces 85: 1483-1507 [doi: 10.1353/sof.2007.0071]. Jang, Yuri; Bergman, Elizabeth; Schonfeld, Lawrence et al. 2007: The Mediating Role of Health Perceptions in the Relation Between Physical and Mental Health. A Study of Older Residents in Assisted Living Facilities. In: Journal of Aging and Health 19,3: 439-452 [doi: 10.1177/0898264307300969]. Jorm, Antony F. 2000: Is Depression a Risk Factor for Dementia or Cognitive Decline? A Review. In: Gerontology 46: 219-227 [doi: 10.1159/000022163]. Jylhä, Marja 2007: Comparisons, cultures, contexts. In: European Journal of Ageing 4,2: 101-102 [doi:10.1007/s10433-007-0051-7]. Kalwij, Adriaan S.; Vermeulen, Frederic 2008: Health and labour force participation of older people in Europe: What do objective health indicators add to the analysis? In: Health Economics 17,5: 619-638 [doi: 10.1002/hec.1285]. The Association between Age and Depressive Symptoms among Older Men and Women • 123 Katz, Ira R. 1996: On the inseparability of mental and physical health in aged persons – Lessons from depression and medical comorbidity. In: American Journal of Geriatric Psychiatry 4: 1-16. Kessler, Ronald C.; McGonagle, Katherine A.; Zhao, Shanyang et al. 1994: Lifetime and 12-month prevalence of DSM-III-R psychiatric disorders in the United States: Results from the National Comorbidity Survey. In: Archives of General Psychiatry 51: 8-19. Korten, Ailsa; Henderson, Scott 2000: The Australian National Survey of Mental Health and Well-Being. In: British Journal of Psychiatry 177: 325-330. Ladin, Keren 2008: Risk of Late-Life Depression Across 10 European Union Countries: Deconstructing the Educational Effect. In: Journal of Aging and Health 2,6: 653-670 [doi: 10.1177/0898264308321002]. Lehtinen, Ville; Michalak, Erin; Wilkinson, Clare et al. 2003: Urban-rural differences in the occurrence of female depressive disorder in Europe – evidence from the ODIN study. In: Social Psychiatry and Psychiatric Epidemiology 38,6: 283-289 [doi: 10.1007/ s00127-003-0631-6]. Lehtinen, Ville; Sohlman, Britta; Kovess-Masfety, Viviane 2005: Level of positive mental health in the European Union: Results from the Eurobarometer 2002 survey. In: Clini- cal Practice and Epidemiology in Mental Health. [http://www.cpementalhealth.com/ content/1/1/9, 25 April 2010] [doi:10.1186/1745-0179-1-9]. Lenze, Eric J.; Rogers, Joan C.; Martire, Lynn M. et al. 2001: The association of late-life depression and anxiety with physical disability: a review of the literature and prospec- tus for future research. In: American Journal of Geriatric Psychiatry 9: 113-135. Litwin, Howard 2002: Examination of the association of age, disability, and mood among Jewish older adults in Israel. In: Aging & Mental Health 6,4: 397-401 [doi: 10.1080/1360786021000007018]. Litwin, Howard; Sapir, Eliyahu V. 2008: Perceived income adequacy among older adults in 12 countries: Findings from the Survey of Health, Ageing and Retirement in Europe. Hebrew University of Jerusalem, Working Paper. Long, J. Scott; Freese, Jeremy 2006: Regression Models for Categorical Dependent Variables Using Stata. Texas: Stata Press. MacKinnon, David P.; Fairchild, Amanda J.; Fritz, Matthew S. 2007: Mediation Analysis. In: Annual Review of Psychology 58: 593-614 [doi: 10.1146/annurev. psych.58.110405.085542]. Mirowsky, John; Ross, Catherine E. 1992: Age and Depression. In: Journal of Health and Social Behavior 33,3: 187-205. Opolski, Melissa; Wilson, Ian 2005: Asthma and depression: a pragmatic review of the literature and recommendations for future research. In: Clinical Practice and Epide- miology in Mental Health. [doi: 10.1186/1745-0179-1-18], URL: http://www.cpemental- health.com/content/1/1/18, 20 July 2010. Prince, Martin J.; Reischies, Friedel; Beekman, Aartjan T.F et al. 1999a: Development of the EURO-D scale – a European Union initiative to compare symptoms of depression in 14 European centres. In: British Journal of Psychiatry 174: 330-38. • Isabella Buber, Henriette Engelhardt124 Prince, Martin J.; Beekman, Aartjan T.F; Deeg, Dorly J.H. et al. 1999b: Depression symp- toms in late life assessed using the EURO-D scale. Effect of age, gender and marital status in 14 European centres. In: British Journal of Psychiatry 174: 339-45. Reischies, Friedel M.; Neu, Peter 2000: Comorbidity of mild cognitive disorder and de- pression – a neuropsychological analysis. In: European Archives of Psychiatry and Clinical Neuroscience 250: 186-193 [doi:10.1007/s004060070023]. Schaan, Barbara 2009: Verwitwung, Geschlecht und Depression im höheren Lebensal- ter. In: Axel Börsch-Supan et al. (Eds.): 50+ in Deutschland und Europa – Ergebnisse des Survey of Health, Ageing and Retirement in Europa [Series ‘Alter(n) und Gesell- schaft’]. Wiesbaden: VS Verlag für Sozialwissenschaften: 115-131. Scogin, Forrest; Rohling, Martin L. 1989: Cognitive Processes, Self-Reports of Memory Functioning, and Mental Health Status in Older Adults. In: Journal of Aging and Health 1,4: 507-520. Stephens, Thomas; Joubert, Natacha 2001: The Economic Burden of Mental Health Problems in Canada. In: Chronic Diseases in Canada 22,1: 18-23. Stordal, Eystein; Krüger, Bjartveit M.; Dahl, Nils H. et al. 2001: Depression in rela- tion to age and gender in the general population: the Nord-Trøndelag Health Study (HUNT). In: Acta Psychiatrica Scandinavica 104,3: 210–216 [doi: 10.1034/j.1600- 0447.2001.00130.x]. Trollor, Julian N.; Anderson, Tracy M.; Sachdev, Perminder S. et al. 2007: Age shall not weary them: mental health in the middle aged and the elderly. In: Australian and New Zealand Journal of Psychiatry 41,7: 581-589 [doi:10.1080/00048670701392817]. Vaupel, James W. 2010: Biodemography of human ageing. In: Nature 464(7288): 536-542 [doi:10.1038/nature08984]. Verropoulou, Georgia; Tsimbos, Cleon 2007: Socio-demographic and health-related fac- tors affecting depression of the Greek population in later life: an analysis using SHARE data. In: European Journal of Ageing 4,3: 171-181 [doi: 10.1007/s10433-007-0060-6]. Winkelmann, Rainer 2003: Econometric Analysis of Count Data. Heidelberg: Springer. Whooley, Mary A.; Kiefe, Catarina I.; Chesney, Margaret A. et al. 2002: Depressive Symptoms, Unemployment, and Loss of Income: The CARDIA Study. In: Archives of Internal Medicine 162: 2614-2620. Wittchen, Hans-Ulrich; Jacobi, Frank 2005: Size and burden of mental disorders in Eu- rope - a critical review and appraisal of 27 studies. In: European Neuropsychopharma- cology 15,4: 357-376 [doi:10.1016/j.euroneuro.2005.04.012]. Yang, Yang 2007: Is old age depressing? Growth trajectories and cohort variations in late-life depression. In: Journal of Gerontology – Social Science 56B, S311-S320. [doi: 10.1177/002214650704800102] Zunzunegui, Maria-Victoria; Minicuci, Nadia; Blumstein, Tzuia et al. 2007: Gender dif- ferences in depressive symptoms among older adults – A cross-national compari- son. In: Social Psychiatry and Psychiatric Epidemiology 42: 198-207 [doi: 10.1007/ s00127-007-0158-3]. The Association between Age and Depressive Symptoms among Older Men and Women • 125 Dr. Isabella Buber-Ennser ( ). Vienna Institute of Demography (Austrian Academy of Sciences), Wittgenstein Centre for Demography and Global Human Capital, A-1040 Vienna, Austria. E-Mail: isabella.buber@oeaw.ac.at URL: www.oeaw.ac.at/vid Prof. Dr. Henriette Engelhardt-Wölfl er. Chair of Population Studies, Otto Friedrich University Bamberg, 96052 Bamberg, Germany. E-Mail: henriette.engelhardt-woelfl er@uni-bamberg.de URL: www.uni-bamberg.de/demografi e Translated from the original text by the Federal Institute for Population Research, for informa- tion only. The reviewed and author’s authorised original article in German is available under the title “Der Zusammenhang zwischen Alter und depressiven Symptomen bei Männern und Frauen höheren Lebensalters in Europa. Erkenntnisse aus dem SHARE-Projekt”, DOI 10.4232/10.CPoS- 2011-02de or URN urn:nbn:de:bib-cpos-2011-02de7, at http://www.comparativepopulationstudies. de. Date of submission: 28.04.2010 Date of Acceptance: 17.01.2010 Comparative Population Studies – Zeitschrift für Bevölkerungswissenschaft www.comparativepopulationstudies.de ISSN: 1869-8980 (Print) – 1869-8999 (Internet) © Federal Institute for Population Research 2011 – All rights reserved Published by / Herausgegeben von Prof. Dr. Norbert F. Schneider Layout and print: Federal Institute for Population Research, Wiesbaden (Germany) Managing Editor / Redaktion Frank Swiaczny Copy Editor / Schlussredaktion Dr. Evelyn Grünheid Scientifi c Advisory Board / Wissenschaftlicher Beirat Jürgen Dorbritz (Wiesbaden) Paul Gans (Mannheim) Johannes Huinink (Bremen) Marc Luy (Wien) Clara H. Mulder (Groningen) Notburga Ott (Bochum) Peter Preisendörfer (Mainz) Board of Reviewers / Gutachterbeirat Martin Abraham (Erlangen) Laura Bernardi (Lausanne) Hansjörg Bucher (Bonn) Claudia Diehl (Göttingen) Andreas Diekmann (Zürich) Gabriele Doblhammer-Reiter (Rostock) Henriette Engelhardt-Wölfl er (Bamberg) E.-Jürgen Flöthmann (Bielefeld) Alexia Fürnkranz-Prskawetz (Wien) Beat Fux (Zürich) Joshua Goldstein (Rostock) Karsten Hank (Köln) Sonja Haug (Regensburg) Franz-Josef Kemper (Berlin) Michaela Kreyenfeld (Rostock) Aart C. Liefbroer (Den Haag) Kurt Lüscher (Konstanz) Dimiter Philipov (Wien) Tomáš Sobotka (Wien) Heike Trappe (Rostock)