ORIGINAL PAPER

Maria Victoria Zunzunegui Æ Nadia Minicuci Æ Tzuia Blumstein Æ Marianna Noale Æ Dorly Deeg Æ Marja JylhaNancy L. Pedersen, for the CLESA Working Group

Gender differences in depressive symptoms among older adults:a cross-national comparison

The CLESA project

Accepted: 3 January 2007 / Published online: 13 February 2007

j Abstract Objectives To assess country-specificgender differences in depressive symptoms and toexplore if exposures and vulnerabilities vary by gen-der among older men and women from four Europeancountries and Israel. Methods Data on 4,449 subjectsbetween 75 and 84 years old were derived from

CLESA (‘‘Cross-national determinants of quality oflife and health services for the elderly’’. A ratio scoreof depressive symptoms derived form the CESD andGDS scales was regressed on education, marital status,living arrangements, comorbidity and disability andall interactions of these factors with gender andcountry. Results The prevalence of depressive symp-toms is higher in women than in men in everycountry, except Sweden. Women are more likely to beexposed to socio-structural risks, and have poorerhealth and more disability than men in most of thecountries. However, women are not more vulnerableto these risk factors. Conclusions Findings indicatethat the female excess in depressive symptoms re-mains after taking into account the higher prevalenceof socio-structural and health-related risk factors andthat older women are not more vulnerable than oldermen to these known risk factors, suggesting theexistence of additional pathways linked to genderand/or biological sex.

j Key words gender – depressive symptoms – aging– cross-cultural research

Introduction

Evidence shows that prevalence of depression ishigher among women than among men across ageand ethnic groups [5, 16]. These gender differencesare also present when number and severity of symp-toms are examined [34]. Artefactual reasons havebeen discussed previously but they do not fully ac-count for gender differences in depression [34]. Aquestion remains whether women have a higherprevalence of depressive symptoms because they aremore likely to be exposed to risk factors or becausethey are more vulnerable to them.SP

PE

158

The CLESA Working Group consists of:Finland: Jylha M (Principal Investigator), Nikula S, Autio V, Israel:Gindin J (Principal Investigator)*, Blumstein T, Chetrit A, Walter-Ginzburg A, Shamay K (*First Principal Investigator—the lateModan Baruch of the Chaim Sheba Medical Center). Italy: Maggi S(Project Leader), Minicuci N, Noale M, Marzari C. The Netherlands:Deeg DJH (Principal Investigator), Pluijm SMF. Spain: Otero A(Principal Investigator), Garcia de Yebenes MJ, Rodrıguez-Laso A,Zunzunegui MV. Sweden: Pedersen NL (Principal Investigator),Bardage C.

M.V. Zunzunegui (&)Dept. de medecine sociale et preventiveUniversite de MontrealCP 6128, Succursale Centre VilleMontreal (QC) H3C 3J7, CanadaE-Mail: maria.victoria.zunzunegui@umontreal.ca

N. Minicuci Æ M. NoaleNational Council ResearchInstitute of Neuroscience, Padova Aging sectionPadova, Italy

T. BlumsteinGertner Health InstituteChaim Sheba Medical CenterTel Hashomer, Israel

D.J.H. DeegVU University Medical CentreEMGO-Institute LASAAmsterdam, The Netherlands

M. JylhaSchool of Public HealthUniversity of TampereTampere, Finland

N.L. PedersenDept. of Medical Epidemiology and BiostatisticsKarolinska InstituteStockholm, Sweden

Soc Psychiatry Psychiatr Epidemiol (2007) 42:198–207 DOI 10.1007/s00127-007-0158-3

Gender differences in depression decrease in oldage [8, 43]. However, researchers across the worldhave consistently reported that older women are morelikely than men to be exposed to social and economicfactors associated with depression: lower education,lower income, less skilled occupations, greater likeli-hood of widowhood and higher social isolation andmore likely to have greater morbidity and functionallimitations [3, 4, 39, 40, 44].

A higher vulnerability to a risk factor in women ascompared to men would mean that this risk factorholds a certain risk for depression which is greater forwomen than for men. However, some studies havefound higher vulnerability to health and social riskfactors for depression in men [1, 19, 42] while otherstudies have not found any gender differences invulnerabilities to material and social deprivation or todeclining health [6, 7, 45].

Gender is not only an individual attribute but astructural and relational characteristic which leads topower dynamics and identity construction [27].Gender relations are power relations that constructthe interdependent categories of men and women[15]. The meaning of gender varies across societiesand the impact of gender on health will therefore varyaccording to the cultural norms and gender-relatedvalues prevailing in each particular society. We wouldexpect smaller gender health differences in egalitariancultures as compared with societies with segregatedgender roles.

The ‘‘Cross-national determinants of quality of lifeand health services for the elderly’’ (CLESA project)provides an opportunity to test for gender differencesof depressive symptoms in the elderly from differentcultures [30]. The project aimed to investigate thedeterminants of quality of life among older peoplethrough six existing longitudinal studies: TamELSA(Tampere, Finland), CALAS (Israel) ILSA (Italy),LASA (The Netherlands), Aging in Leganes (Spain),SATSA (Sweden). Previous studies have shown higherprevalence of depressive symptoms in Israel [38],Veneto (Italy) [31] and Leganes (Spain) [45] than inthe Netherlands [9, 40] or Sweden [21]. In the baselinedescription of the CLESA study, scores of depressivesymptoms, standardized to the overall mean andstandard deviation, varied from 0.15 in The Nether-lands to 0.34 in Italy [30]. Large gender differenceshave been reported in the studies from Israel [11],Italy [31], Spain [45] while differences between menand women from The Netherlands [10] and Swedenare smaller or disappear when taking into accountsocioeconomic and health factors [21, 41].

Gender roles are more salient in Southern than inNorthern Europe. Southern European women areresponsible to keep the traditional and the religiousvalues of the family. In the current times of socialchange, old women may feel that they have lost thiscentral role in family life [2]. The sample of Israelielderly is a unique population group of multicultural

immigrants whose background stems from traditionalfamily structures, yet for those originating from Eastor Western Europe, ideological forces such as Zion-ism and Socialism played a role of increased genderequality in comparison to southern Europe. Thus,there are overlapping ethnic, familial and ideologicalbases in the mixed cultures of this elderly population[26].

We hypothesized that gender differences indepressive symptoms will be larger in Italy and Spainthan in Sweden and The Netherlands. In addition,exposure to known depression risk factors will behigher among women as compared to men, regardlessof the country but vulnerability to some of the riskfactors will be different. In particular, living alone anddisability will be more strongly associated withdepression in older men than in women of compa-rable age. We aim at answering the following researchquestions: (a) Are gender differences in depressivesymptoms similar in all CLESA countries? (b) Arethere gender differences in exposure to risk factors?Are these differences similar across countries? (c) Domen and women have the same vulnerabilities to riskfactors? Are these vulnerabilities similar acrosscountries?

Materials and methods

The CLESA project involves longitudinal studies performed in sixdifferent countries [7, 14, 20, 22, 28, 33]. Details on the harmoni-zation process of the CLESA study can be found in [30, 32, 35].Data on depressive symptoms were not available in Finland. Thesample for this article includes the 75–84 years old (n = 4,449)since this is the common age group in the five studies with data ondepressive symptoms. A brief description of these studies is pro-vided below.

The Cross Sectional and Longitudinal Aging Study (CALAS)from Israel aims to provide a description of physical health, mentalhealth, everyday functioning, and living conditions of the old–oldsegment of the multicultural elderly population in Israel. In 1989 arandom sample of 1,820 subjects aged 75–94, stratified by age, sexand place of birth was selected from the National PopulationRegister [20].

The Italian Longitudinal Study on Aging (ILSA) considers arandom sample of 5,632 individuals aged 65–84 years, stratified byage and sex, from the registry office of eight municipalities from theNorthern, the Central and the Southern part of Italy. Baseline datacollection was carried out in 1992 on socio-demographical,behavioral and health characteristics, laboratory tests, a physicianexamination and selected diagnostic tests [28].

The Longitudinal Aging Study Amsterdam (LASA) is a longi-tudinal study of changes in well-being and autonomy in an olderpopulation. A random sample of older (55–85 years) persons,stratified for age, sex and level of urbanization was drawn from thepopulation registers in 11 municipalities in The Netherlands. Thebaseline year for the LASA study is 1992–1993 [7].

Ageing In Leganes (AL) is a longitudinal study on the influenceof social relations on the health and function of the communityliving population 65 years and older in Leganes (Spain). In 1993 atwo year age class and sex stratified sample of 1,564 elderly peoplewas obtained from the municipal census. Information was gatheredon functional status and living conditions [14].

The Swedish Adoption/Twin Study on Aging (SATSA) is a lon-gitudinal twin study of normal aging. The questionnaire assessedphysical and mental health status, selected personality dimensions,

199

functional status and living conditions, early rearing and adultenvironments. The study started in 1984 but for the CLESA projectwe considered the 1987 follow-up as the baseline [33].

LASA, ILSA and AL are based on random samples extractedfrom municipal registers; CALAS is based on a random samplefrom the National Population Register and SATSA is based on asample from the Swedish Twin Register.

j Measures

The measure of depressive symptoms used in Israel, the Netherlands,Leganes (Spain) and Sweden was based on the Center for Epidemi-ologic Studies Depression (CES-D) scale [36], in Italy the Italianversion of the Geriatric Depression Scale [13] was administered. Foreach scale, a ratio score (RS) was calculated as a ratio between thescore obtained by each subject and the maximum potential score foranswered items. If there were more than three missing items on theCES-D scale or more than five missing items on the GDS, the RS wasset to missing. RSs were then numbers ranging between 0 (nodepressive symptomatology) and 1 (presence of depressive symp-tomatology). The CES-D and the GDS scales have very similar dis-criminatory ability [37] and similar correlates in two Italianpopulations: Italian Longitudinal Study on Aging and Study on Agingin the Veneto region (results available upon request).

Potential socio-structural risk factors for depressive symptom-atology in the older population were selected according to the lit-erature and the availability of information in the CLESA database.Age, education, lifetime occupation, being married and living alonewere measured using the categories outlined in Table 2. For themultivariate analysis, lifetime occupation was excluded since it isnot well defined for women working as housewives. Living alonewas preferred to being married as being more informative of thesocial network. Comorbidity and disability measures (ADL andIADL) were included as measures of health and function and po-tential risk factors.

j Statistical analysis

To generalize each study sample to the population between75 years old and 84 years old in each country and to take intoaccount the different sampling techniques, a set of country-specificweights was calculated and applied to the raw data.

The association of gender with social structural measures andhealth and disability measures was investigated with the v2-test foreach country. For continuous variables (age, RS) the comparison ofgender mean values within country was evaluated using the GeneralLinear Model (GLM) procedure, after checking for heteroschedas-ticity using the Levine test.

• First, country-specific multivariate linear regression models wereperformed in order to identify predictors of depressive symp-toms, separately for men and women. The models included twodomains: (a) social structure and (b) health status and disability.

• Second, to reach the final model, five hierarchical models wereconstructed as follows:

• from the country-specific multivariate coefficients for eachpredictor, a pooled estimate was constructed by weighting eachcountry-specific coefficient with the inverse of its variance, andthen summing over all five weighted coefficients, divided by thesum of the weights, following Dyer’s method (17). The chi-square test to check homogeneity is based on the differencebetween each country-specific coefficient and the pooled esti-mate. The pooled estimate was used in a chi-square test [17] toverify the null hypothesis that all multivariate coefficients wereequal, i.e. whether the associated factor was homogeneouslydistributed across countries;

• if the hypothesis of homogeneity was rejected, the corre-sponding associated factor was included in the hierarchicalmodel along with the interactions predictor-by-country;

• if the hypothesis of homogeneity was not rejected, the corre-sponding factor was included in the model alone, withoutcountry interaction terms;

• only significant interactions for all levels of the predictor wereaccepted and left in the next model,

• the five hierarchical models were developed in the followingway: Model 1. Country, gender, country*gender; Model 2.Country; gender; significant country*gender interactions,education, education*gender, living arrangements, livingarrangements*gender; Model 3. Country; gender; significantcountry*gender interactions, education; education*gender;living arrangements; comorbidity; comorbidity*gender*coun-try; Model 4. Country; gender; significant country*genderinteractions; education; living arrangements; comorbidity;IADL; IADL*gender*country; Model 5. Country; gender;education; living arrangements; comorbidity; IADL; significantIADL*gender*country interaction; ADL; ADL*gender*country.

• The final model retained only significant terms to obtain themost parsimonious model. Interactions between gender andpredictors test differential gender vulnerability of each riskfactor for depression.

• The Netherlands was taken as the referent country since itpresented the lowest mean values of RS. All analyses wereperformed using release 8.02 of the SAS statistical software.

Results

The distribution of the RS of depressive symptoms bycountry and gender are shown in Table 1. Womenhave higher mean RS than men, and the gender dif-ference is significant in every country except in Swe-den. Mean and median RS are lowest in TheNetherlands and highest in Italy and Israel. Dutchpeople recorded the lowest RS (0.13 for men and 0.17for women) versus the Italians with the highest (0.29for men and 0.39 for women).

Table 1 The distribution of the observed ratio score for depressive symptomsby country and sex

Mean ± SD Median Range

IsraelMen (n = 523) 0.23 ± 0.15 0.20 0–0.70Women (n = 550) 0.32 ± 0.19 0.28 0–0.75p-value ***

ItalyMen (n = 752) 0.29 ± 0.14 0.27 0–0.93Women (n = 1,253) 0.39 ± 0.21 0.37 0–0.96p-value ***

The NetherlandsMen (n = 259) 0.13 ± 0.083 0.09 0–0.63Women (n = 442) 0.17 ± 0.13 0.13 0–0.78p-value ***

Leganes (Spain)Men (n = 153) 0.21 ± 0.07 0.20 0.05–0.63Women (n = 247) 0.28 ± 0.13 0.25 0.07–0.78p-value ***

SwedenMen (n = 108) 0.26 ± 0.13 0.25 0–0.52Women (n = 162) 0.29 ± 0.14 0.28 0–0.86p-value ns

CLESA, community dwelling subjects aged 75–84. Weighted datans: not significant*p £ 0.05; **p £ 0.01; ***p £ 0.001

200

Tab

le2

Mea

nsan

dpe

rcen

tage

sfo

rso

cial

stru

ctur

alm

easu

res,

heal

thst

atus

and

disa

bilit

ym

easu

res,

byco

untr

yan

dse

x

Isra

elIta

lyTh

eN

ethe

rland

sLe

gane

s(S

pain

)Sw

eden

Mal

es(n

=52

3)Fe

mal

es(n

=55

0)M

ales

(n=

752)

Fem

ales

(n=

1,25

3)M

ales

(n=

259)

Fem

ales

(n=

442)

Mal

es(n

=15

3)Fe

mal

es(n

=24

7)M

ales

(n=

108)

Fem

ales

(n=

162)

Soci

alst

ruct

ural

mea

sure

sM

ean

age

±SD

78.9

±2.

978

.7±

3.2

ns78

.8±

2.0

79.2

±2.

6***

78.7

±1.

979

.0±

2.6

ns78

.9±

2.1

78.8

±2.

7ns

78.6

±3.

379

.0±

2.8

nsEd

ucat

ion

(%)

Elem

enta

rysc

hool

not

com

plet

ed51

.370

.5**

*43

.053

.4**

*9.

418

.6**

*87

.2**

0.0

0.0*

*El

emen

tary

scho

olco

mpl

eted

10.6

6.8

28.3

28.4

3446

.773

.811

.586

.997

.9Vo

catio

nal

scho

olor

seco

ndar

yed

ucat

ion

28.5

17.5

11.8

11.4

42.7

27.6

23.1

0.9

6.8

1.3

Colle

geor

univ

ersi

ty9.

65.

216

.96.

813

.97.

13.

10.

46.

30.

8Li

fetim

eoc

cupa

tion

(%)

Hig

her/

low

erno

n-m

anua

l39

.29.

8***

26.5

12.1

***

38.5

25.8

***

14.7

3.1*

**23

.85.

4***

Spec

ializ

ed/(

sem

i)sk

illed

man

ual

3527

.126

.412

.340

.119

47.3

16.3

26.1

3.5

non

spec

ializ

ed/n

on-s

kille

dm

anua

l13

.13.

715

.711

824

.717

.620

.334

.730

.9Fa

rmer

12.7

2.5

31.4

21.2

13.1

319

11.8

15.4

7H

ouse

wife

056

.90.

043

.40.

327

.51.

448

.50

53.2

Mar

ital

stat

us(%

)Si

ngle

orne

ver

mar

ried

1.9

1.5*

**4.

311

.3**

*4.

17*

**1.

84.

7***

5.2

7.9*

**M

arrie

dor

coha

bitin

g75

.823

.873

.625

.369

.430

.476

.225

.376

.630

.5W

idow

ed18

.170

.121

.262

.822

.658

.521

.668

.518

.255

.2D

ivor

ced

orse

para

ted

4.2

4.7

0.9

0.6

3.9

4.1

0.4

1.5

06.

4Li

ving

arra

ngem

ents

(%)

Alon

e16

.855

.5**

*9.

828

.1**

*25

.563

***

623

.5**

*23

.460

.6**

*W

ithsp

ouse

only

56.2

1850

.316

.966

.729

.157

.818

.564

.524

.4W

ithot

hers

2726

.439

.955

7.8

7.9

36.2

5812

.115

Hea

lthst

atus

and

disa

bilit

ym

easu

res

Com

orbi

dity

(%)

No

dise

ase

27.9

20.1

***

96.

3ns

31.2

24.5

ns18

.24.

9***

28.5

28.2

nsO

nedi

seas

e27

.924

.227

.123

.637

.437

.926

.517

.328

.731

.1Tw

oor

mor

edi

seas

es44

.255

.763

.970

.131

.437

.655

.377

.842

.840

.7AD

Lst

atus

(%)

Inde

pend

ent

(0ite

mw

ithdi

s.)

80.9

76ns

64.2

64.2

ns95

.691

.8ns

71.3

59.4

*92

77.3

**M

ild/m

oder

ate

disa

bilit

y(1

,2)

10.1

12.9

24.2

24.3

3.3

7.4

23.5

32.5

5.2

16.8

Seve

redi

sabi

lity

(3,4

)9

11.1

11.6

11.5

1.1

0.8

5.2

8.1

2.8

5.9

IAD

Lst

atus

(%)

Inde

pend

ent

(0ite

mw

ithdi

s.)

70.4

51.7

62.6

45.8

37.3

34.9

50.7

44.9

3857

.3M

ild/m

oder

ate

disa

bilit

y(1

,2)

15.3

26.9

***

30.1

38.3

***

47.1

51.6

ns17

.328

.6*

43.3

24.6

**Se

vere

disa

bilit

y(3

,4,5

)14

.421

.47.

315

.915

.613

.632

26.5

18.7

18.1

CLES

A,co

mm

unity

dwel

ling

subj

ects

aged

75–

84;

wei

ghte

dda

tans

:no

tsi

gnifi

cant

*p£

0.05

;**

p£

0.01

;**

*p£

0.00

1

201

The gender-specific distribution of the socio-structural and health-related risk factors for depres-sive symptoms for subjects aged 75–84 are presentedin Table 2. There are significant differences in thedistribution of socio-structural risk factors acrosscountries, both for men and women (Chi-square p-values < 0.0001). In particular, women have lowereducation than men and they are more likely to bewidowed. Being a housewife is the main occupationfor women in all countries in this age group, whileoccupations for men vary across countries, the high-est percentage being farmers in Italy, non-specialized/non-skilled manual in Sweden, specialized/(semi)skilled manual in The Netherlands and Leganes(Spain), and higher/lower non manual in Israel. InIsrael, The Netherlands and Sweden, more womenthan men live alone. In Italy and Leganes women aremore likely than men to live with others including thespouse if still married. The most common livingarrangement for the Italian and Spanish women isliving with others, usually their children.

There are fewer and less consistent gender differ-ences in the health and disability risk factors fordepression. Overall, women have more chronic con-ditions and disability than men do but differences arenot significant in every country. In Israel and Leganes(Spain), women reported more diseases than men. InLeganes and Sweden, women reported significantlyhigher ADL disability as compared with men. IADLdisability is also significantly higher among women ascompared with men in Israel, Italy, and Leganes(Spain). Only in Sweden is IADL disability moreprevalent among men than among women.

Table 3 presents the gender-specific parameterestimates of the multivariate models to analyze theassociations between potential risk factors and scoresof depressive symptoms in every country. Amongmen, sociostructural correlates of depressive symp-toms were age in Italy, low education in Israel andItaly, and living with spouse was inversely associatedin Israel and The Netherlands; among women, loweducation was positively associated in Israel, Italy andThe Netherlands, while living with spouse for theNetherlands and living with others in Israel and theNetherlands had an inverse effect. Among the healthand disability factors, for men, comorbidity wasassociated in all countries, but Sweden; ADL disabilityis related to depressive symptoms in Israel, Italy andThe Netherlands, while IADL disability is significantin all countries, but Sweden. Among women, associ-ations between health and function with depressivesymptoms are generally weaker, being significant onlyfor IADL disability. Among women, associations be-tween comorbidity and ADL disability with depressivesymptoms are non-significant, except for The Neth-erlands. In Sweden, none of the investigated factorswas associated with depressive symptoms.

Applying Dyer’s approach to test interactionsbased on country-specific models, comorbidity, ADL

and IADL disability were found to be non-homoge-nously distributed across countries, while educationand living arrangements were homogeneous predic-tors (results not shown). These findings led to testingall interactions between gender and risk factors andcountry interactions for comorbidity, IADL and ADL(Table 4). The Netherlands is taken as referencecountry since it has the lowest score in depressivesymptoms.

In the first model, an increased score in thedepression scale is estimated for all countries com-pared to the Netherlands. Women are significantlymore depressed than men in all countries and par-ticularly more depressed than men in Italy and Israel,precisely where the overall prevalence of depressivesymptoms is higher both for men and women. Theexcess in depressive symptoms among men varyfrom 0.17 (parameter estimate) for Italian to 0.14(parameter estimate) for Swedish compared withDutch men. Among women, the corresponding excessis 0.27 (0.17 + 0.04 + 0.06) for Italian and 0.19(0.14 + 0.04 + 0.01) for Swedish compared to Dutchwomen.

In model 2, education and living arrangementswere included and interactions with gender weretested. No significant interactions were found. Loweducation is positively associated with depressivesymptoms and living with either the spouse or withother people is associated with having low depressivesymptoms.

In model 3, we have introduced the effect ofcomorbidity, along with the interactions with coun-try and gender. The presence of co-morbid condi-tions is a risk factor, however it seems to beindependent from the number of conditions (themagnitude of the estimates are similar for the twolevels of the predictor). No interactions were statis-tically significant.

Model 4 includes IADL disability and its interac-tions with country and gender. IADL disability isassociated with depression in all countries and par-ticularly for Italian and Israeli women, while Swedishwomen with severe IADL disability seem to have evenless depressive symptoms than Dutch men or women.The inclusion of IADL disability in model 4 explainsthe increased vulnerability of Italian and Israeli wo-men to depression (Models 1–3), since estimates forcountry*gender interaction terms become non-sig-nificant. However, gender and country differencesremain significant. The likelihood of having depres-sive symptoms increases with the number of activitiesfor which the elderly person, man or woman, isdependant.

According to model 5, those with severe ADL dis-ability were more likely to have depressive symptomsand the intensity of depressive symptoms increaseswith the severity of disability. At this very advancedstage of disability, Italian and Spanish women havefewer depressive symptoms than Dutch women.

202

Tab

le3

Estim

ates

ofth

ese

x-co

untr

y-sp

ecifi

cre

gres

sion

mod

els

ofth

era

tiosc

ore

for

depr

essi

vesy

mpt

oms

Isra

el(n

=37

0)Ita

ly(n

=77

2)Th

eN

ethe

rland

s(n

=51

6)Le

gane

s(S

pain

)(n

=20

5)Sw

eden

(n=

61)

Para

met

eres

timat

eSt

anda

rder

ror

pPa

ram

eter

estim

ate

Stan

dard

erro

rp

Para

met

eres

timat

eSt

anda

rder

ror

pPa

ram

eter

estim

ate

Stan

dard

erro

rp

Para

met

eres

timat

eSt

anda

rder

ror

p

Men

Soci

alst

ruct

ural

mea

sure

sAg

e(c

ont)

)0.

004

0.00

30.

154

0.00

60.

003

0.02

30.

003

0.00

20.

126

0.00

10.

002

0.61

40.

002

0.00

50.

699

Educ

atio

n(e

lem

.or

less

)0.

039

0.01

50.

008

0.03

50.

016

0.02

60.

011

0.01

00.

279

0.00

70.

038

0.85

20.

004

0.04

10.

929

Livi

ngw

ithsp

ouse

only

(vs.

alon

e))

0.05

40.

020

0.00

6)

0.00

90.

026

0.73

4)

0.06

20.

012

0.00

0)

0.00

40.

028

0.90

0)

0.01

60.

035

0.65

2Li

ving

with

othe

rs(v

s.al

one)

)0.

031

0.02

30.

179

0.00

40.

026

0.86

4)

0.03

00.

021

0.14

8)

0.00

20.

029

0.95

40.

079

0.05

50.

156

R-sq

uare

0.04

60.

016

0.06

40.

002

0.06

7

Hea

lthst

atus

and

disa

bilit

ym

easu

res

Com

orbi

dity

(num

ber

ofse

lf-re

port

edco

nditi

ons)

0.03

40.

006

0.00

00.

012

0.00

70.

073

0.02

20.

005

0.00

0)

0.00

20.

005

0.97

20.

025

0.01

40.

084

ADL

mild

disa

bilit

y(v

s.in

depe

nden

t)0.

076

0.03

10.

016

0.04

80.

018

0.00

60.

028

0.03

10.

363

)0.

015

0.01

80.

385

)0.

140

0.09

60.

155

ADL

seve

redi

sabi

lity

(vs.

inde

pend

ent)

0.02

80.

042

0.50

00.

193

0.03

40.

000

0.12

70.

048

0.00

90.

042

0.04

80.

388

0.04

20.

095

0.66

0IA

DL

mild

disa

bilit

y(v

s.in

depe

nden

t)0.

040

0.01

90.

033

0.03

90.

018

0.02

80.

023

0.01

10.

034

0.03

60.

017

0.03

0)

0.00

60.

036

0.86

5IA

DL

seve

redi

sabi

lity

(vs.

inde

pend

ent)

0.09

40.

028

0.00

10.

163

0.03

40.

000

0.02

80.

016

0.07

60.

074

0.01

90.

000

0.09

50.

068

0.16

7R-

squa

re0.

208

0.14

80.

079

0.10

80.

183

Isra

el(n

=31

5)Ita

ly(n

=67

0)Th

eN

ethe

rland

s(n

=49

6)Le

gane

s(S

pain

)(n

=20

1)Sw

eden

(n=

110)

Wom

enSo

cial

stru

ctur

alm

easu

res

Age

(con

t)0.

003

0.00

40.

372

0.00

40.

003

0.25

30.

001

0.00

20.

629

0.00

20.

004

0.56

8)

0.00

30.

005

0.57

1Ed

ucat

ion

(ele

m.

orle

ss)

0.05

10.

022

0.01

90.

076

0.02

20.

001

0.03

40.

013

0.01

00.

007

0.07

50.

928

)0.

040

0.08

40.

636

Livi

ngw

ithsp

ouse

only

(vs.

alon

e))

0.03

80.

024

0.10

90.

015

0.02

50.

559

)0.

037

0.01

40.

008

)0.

010

0.02

80.

720

)0.

012

0.03

20.

712

Livi

ngw

ithot

hers

(vs.

alon

e))

0.06

00.

022

0.00

60.

026

0.02

00.

194

)0.

046

0.02

30.

048

0.01

80.

023

0.44

8)

0.00

60.

044

0.89

6R-

squa

re0.

038

0.02

60.

034

0.01

10.

006

Hea

lthst

atus

and

disa

bilit

ym

easu

res

Com

orbi

dity

(num

ber

ofse

lf-re

port

edco

nditi

ons)

0.00

70.

006

0.24

80.

009

0.00

70.

210

0.02

00.

006

0.00

10.

013

0.00

70.

080

0.01

30.

012

0.26

0AD

Lm

ilddi

sabi

lity

(vs.

inde

pend

ent)

)0.

040

0.02

80.

155

)0.

011

0.02

20.

621

)0.

008

0.02

40.

755

0.08

00.

026

0.00

20.

059

0.05

50.

285

ADL

seve

redi

sabi

lity

(vs.

inde

pend

ent)

0.00

80.

036

0.81

60.

023

0.03

80.

547

0.17

10.

073

0.02

00.

042

0.04

80.

389

0.18

10.

094

0.05

8IA

DL

mild

disa

bilit

y(v

s.in

depe

nden

t)0.

060

0.01

90.

002

0.07

80.

020

0.00

00.

005

0.01

40.

719

0.00

00.

023

0.98

60.

013

0.03

40.

699

IAD

Lse

vere

disa

bilit

y(v

s.in

depe

nden

t)0.

186

0.02

70.

000

0.16

10.

032

0.00

00.

092

0.02

10.

000

0.02

20.

032

0.48

8)

0.03

80.

059

0.51

7R-

squa

re0.

197

0.07

60.

102

0.13

40.

092

Com

mun

itydw

ellin

gsu

bjec

tsag

ed75

–84

,CL

ESA

proj

ect,

wei

ghte

dda

ta

203

The final model, retaining only those variables thatare statistically significant after inclusion of all vari-ables, confirms the differences across countries, withItaly ranking the highest in prevalence and in theexcess of depressive symptoms in women as com-pared with men. Thus, country and gender differencesremain significant as main effects after controlling foreducation, living arrangements, comorbidity andIADL and ADL disability. No interactions betweensocio-structural, health and disability indicators withcountry or gender (except for Italian women) re-

mained significant in the final model suggesting thatthere are no differences in vulnerabilities to risk fac-tors for depressive symptoms between men and wo-men across the countries included in the CLESAstudy.

Discussion

The answers to our three research questions are asfollows: First, women had higher prevalence of

Table 4 Parameter estimates of the regression models

Model 1Country & sex

Model 2Social

Model 3Health

Model 4IADL disability

Model 5ADL disability

Finalmodel

CountryIsrael vs. Netherlands 0.11 0.10 0.08 0.11 0.13 0.13Italy vs. Netherlands 0.17 0.16 0.15 0.18 0.19 0.19Leganes vs. Netherlands 0.09 0.09 0.07 0.09 0.09 0.09Sweden vs. Netherlands 0.14 0.12 0.11 0.12 0.11 0.11GenderWomen 0.04 0.00 0.01 0.02 0.04 0.04Israel*women 0.04 0.04 0.08 0.02ns

Italy*women 0.06 0.05 0.08 0.01ns

Leganes* women 0.02 ns

Sweden*women )0.01 nsEducationElementary vs. higher 0.03 0.03 0.03 0.03 0.03Elementary*women 0.02 0.01 nsLiving arrangementsLiving with spouse only vs. alone )0.04 )0.03 )0.04 )0.04 )0.04Living with others vs. alone )0.02 )0.02 )0.04 )0.04 )0.04Living with spouse only*women 0.02 nsLiving with others*women 0.01 nsComorbidityOne disease vs. none 0.03 0.02 0.01 0.02Two or more diseases vs. none 0.05 0.02 0.02 0.02One disease*Israel*women )0.06One disease*Italy*women )0.01 nsOne disease*Leganes*women 0.03 nsOne disease*Sweden*women 0.03 nsTwo or more diseases*Israel*women 0.00 nsTwo or more diseases*Italy*women )0.04 nsTwo or more diseases*Leganes*women )0.01 nsTwo or more diseases*Sweden*women )0.04 nsIADL statusIADL mild (1–2 item) 0.02 0.03 0.03IADL severe (3+) 0.11 0.10 0.11IADL mild*Israel*women 0.02 nsIADL mild*Italy*women 0.05 0.05 0.05IADL mild*Leganes*women 0.00 nsIADL mild*Sweden*women 0.01 nsIADL severe*Israel*women 0.10IADL severe*Italy*women 0.06 0.06 0.04IADL severe*Leganes*women 0.00 nsIADL severe*Sweden*women )0.11ADL statusADL mild (1–2 item) 0.01 0.00 nsADL severe (3+) 0.12 0.08ADL mild*Israel*women )0.03 nsADL mild*Italy*women )0.01 nsADL mild*Leganes*women 0.00 nsADL severe*Israel*women )0.06 nsADL severe*Italy*women )0.09ADL severe*Leganes*women )0.12

* = ns (parameter not significant)

204

depressive symptoms as compared to men in allcountries, except for Sweden. Gender differences in-creased with the overall prevalence of depression,with the exception of Sweden where gender differ-ences were not significant. Second, the prevalence ofall socio-structural determinants for depression wassignificantly higher in women than in men acrosscountries, except for Sweden. The health status anddisability measures presented a more heterogeneoussituation across countries; there were no gender dif-ferences in Italy and the Netherlands in comorbidityand ADL disability, while in Israel and in Leganeswomen had higher prevalence of comorbidity ascompared with men. IADL disability was more pre-valent among women in Israel, Italy and Leganes yet itwas more prevalent among Swedish men. Third, noimportant gender differences in vulnerability to socio-structural risk factors for depressive symptoms andvery few indications of differential vulnerability topoor health and disability were found. Italian womenseem to be more vulnerable to IADL disability in thefinal model. None of the well known selected riskfactors was associated with prevalence of depressionin these 75–84 years Swedish people. However, in aprevious analysis on the whole Swedish sampleincluding all subjects over 65 years old [18], signifi-cant gender and chronic illness associations werefound. The lack of significance in the current studymay be due to the truncation of the sample and itsresulting small sample size.

Cultures systematize relations between feelings,self-concept, interpersonal communication and rela-tions of power [25]. As an illustration, the associationbetween depressive symptoms and disability is morepronounced in countries such as Italy and Spain withtraditional, family-based value patterns than incountries with more modern individualistic valuepatterns and traditional values are more likely to beobserved in Roman Catholic countries than in Prot-estant countries, which are also more often secular-ized countries in Western Europe [12]. SouthernEuropean older people have expectations on theavailability of family support that are being onlypartially fulfilled due to changes in family structure.In spite of the large proportions of widowed anddisabled older people in Italy and Spain that arecohabiting with their children, the availability ofpublic pensions and the low rates of institutionaliza-tion (less than 3% of people over 65 are living ininstitutions), the old generation in both countriesexpect their children to give them support and takecare of them, particularly when needed due to poorhealth or loneliness. This is in contrast with the el-derly population from The Netherlands or Sweden,who accept that the State has the legal responsibilityof their care [23]. In Israel, the predominance offamily caregivers for older people has been shownespecially among Middle Eastern/North African borndespite the availability of a full range of services [20]

indicating patterns of support similar to southernEurope. Although the above statements are general-izations that do not account for individual differ-ences, cultural norms are forces driving humanbehavior in ways relevant to depression.

Country differences might result from personalitydifferences, e.g. on an axis of stoicism to com-plaining set. People from Northern Europe may bemore stoic and less likely to complain of symptomsthan those from Southern Europe, who may be moreextrovert and more likely to ask for help. In addi-tion, differences in detection of depression and itstreatment may account for differences across coun-tries [29].

Experiences during life-course are shaped bygender. The lack of information on detailed aspectsof earlier life exposures (violence and hard eco-nomic and social conditions during childhood andadolescence, national histories including periods ofwars and the Holocaust), current economic (incomeand assets) and psychosocial (social support, per-sonality characteristics, religiosity) resources andlife styles greatly limits our ability to further explainhealth gender differences. In addition, the cross-sectional nature of this study limits its explanatorypower, since those depressed are a mixture of re-cent onset and chronically depressed which mayobscure the gender differences found. Lastly, Ital-ians responded to a different depression question-naire and we are not able to quantify the extent ofthis possible information bias. Furthermore, theunexplained variance in depression may stem frombiological/genetic factors that were not measured inthis study [24].

Conclusions

This study encompassing four European and one Is-raeli community populations of people between 75and 84 shows that high depressive symptoms aremore frequent among women as compared with menand this excess in women remains after taking intoaccount cross-country differences in education, livingarrangements, and poor health and function.

j Acknowledgements The CLESA Project is supported by EUQoL2000-00664. The TamELSA is supported by the Academy ofFinland, the Juho Vainio Foundation, the Yrjo Jahnsson Founda-tion, and the Medical Research Fund of Tampere University Hos-pital. CALAS was supported by the US National Institute on AgingGrants -R01-5885-03 and R01-5885-06. ILSA was supported by theCNR (National Research Council) and the Italian Department ofHealth—National Institutes of Health. LASA is supported by TheNetherlands Ministry of Health, Welfare and Sports and the VrijeUniversiteit. Aging in Leganes is supported by the Spanish HealthResearch Fund (FIS), the Madrid Regional Research Fund andPrivate Foundations (La Caixa, BBVA). SATSA is supported by theUS National Institute on Aging (AG 04563, 10175) and the SwedishSocial Research Council.

205

References

1. Ahnlund K, Frodi A (1996) Gender differences in the devel-opment of depression. Acta Psychiatr Scand 37:229–239

2. Alberdi I (1999) La nueva familia espanola. Tecnos, Madrid3. Arber S, Cooper H (1999) Gender differences in health in later

life: the new paradox. Soc Sci Med 48:61–764. Barefoot JC, Mortensen EL, Helms MJ, Avlund K, Schroll M

(2001) A longitudinal study of gender differences indepressive symptoms from age 50 to 80. Psychol Aging 16(2):342–345

5. Bebbington PE (1998) Sex and depression. Psychol Med 28:1–86. Beekman AT, Kriegsman DM, Deeg DJ, van Tilburg W (1995)

The association of physical health and depressive symptoms inthe older population: age and sex differences. Soc PsychiatryPsychiatr Epidemiol 30(1):32–38

7. Beekman ATF, Penninx BWJH, Deeg DJH, Ormel J, Braam AW,van Tilburg W (1997) Depression and physical health in laterlife: results from the Longitudinal Aging Study Amsterdam(LASA). J Affect Disord 46(3):219–231

8. Berkman LF, Berkman CS, Kasl S, Freman DH, Leo L, OstfeldAm, Coroni-Huntley J, Brody JA (1986) Depressive symptomsin relation with physical health and functioning in the elderly.Am J Epidemiol 124:372–388

9. Bisschop MI, Kriegsman DM, Beekman AT, Deeg DJ (2004)Chronic diseases and depression: The modifying role of psy-chosocial resources. Soc Sci Med 59:721–733

10. Bisschop MI, Kriegsman DM, Deeg DJ, Beekman AT, van Til-burg W (2004) The longitudinal relation between chronic dis-eases and depression in older persons in the community: theLongitudinal Aging Study Amsterdam. J Clin Epidemiol7(2):187–194

11. Blumstein T, Benyamini Y, Fuchs Z, Shapira Z, Nokivov I(2004) The effect of communal lifestyle depressive symptoms inlate life. J Aging Health 16:151–174

12. Braam AW, Delespaul P, Beekman ATF, Deeg DJH, Peres K,Dewey M (2004) National context of healthcare, economy andreligion and the association between disability and depressivesymptoms in older Europeans: results from the EURODEP. EurJ Ageing 1:26–36

13. Brink TL, Yesavage JA, Lum O, Heersema PH, Adey M, Rose TL(1982) Screening tests for geriatric depression. Clin Gerontol198(1):37–43

14. Beland F, Zunzunegui MV (1995) El diseno y la ejecucion de laencuesta ‘‘Envejecer en Leganes’’. Rev Gerontol 5:215–231

15. Calasanti T (2004) Feminist gerontology and old men. J Ger-ontol B Psychol Sci Soc Sci Series B 59B:305–314

16. Cole MG, Dendukuri N (2003) Risk factors for depressionamong elderly community subjects: a systematic review andmeta-analysis. Am J Psychiatry 160:1147–1156

17. Dyer AR (1986) A method for combining results from severalprospective epidemiologic studies. Stat Med 5:303–317

18. Fiske A, Gatz M, Pedersen NL (2003) Depressive symptoms andaging: the effects of illness and non-health related events. JGerontol B Psychol Sci Soc Sci 58(6):P320–P328

19. Frasure-Smith N, Lesperance F, Jumeau M, Talajic M, BourassaMG (1999) Gender depression, and one year prognosis aftermyocardial infarction. Psychosom Med 61:26–27

20. Ginzburg AW, Guralnik JM, Blumstein T, Gindin J, Modan B(2001) Assistance with personal care activities among the old–old in Israel: a national epidemiological study. J Am Geriatr Soc49:1–8

21. Haynie DA, Berg S, Johansson B, Gatz M, Zarit SH (2001)Symptoms of depression in the oldest old: a longitudinal study.J Gerontol B Psychol Sci Soc Sci 56(2):111–118

22. Jylha M, Jokela J, Tolvanen E, Heikkinen E, Heikkinen RL,Koskinen S, Leskinen E, Lyyra AL, Pohjolainen P (1992) TheTampere Longitudinal Study on Ageing. Description of thestudy. Basic results on health and functional ability. Scand J SocMed Suppl 47:1–4

23. Jylha M, Jokela J (1990) Individual experiences as cultural—across-cultural study on loneliness among the elderly. AgeingSoc 10:295–315

24. Kendler KS, Gatz M, Gardner CO, Pedersen NL (2006) ASwedish national twin study of lifetime major depression. Am JPsychiatry 163:109–119

25. Kleinman A, Goog B (1985) Culture and depression. Universityof California Press, Ltd

26. Litwin H (1995) The social networks of elderly immigrants. Ananalytic typology. J Aging Stud 9:155–174

27. Longino CF (2004) Editorial: are there any concerns aboutgender inclusion men. J Gerontol A Biol Sci Med Sci 59B:304

28. Maggi S, Zucchetto M, Grigoletto F, for the ILSA group (1994)The Italian Longitudinal Study on Aging (ILSA): design andmethods. Aging Clin Exp Res 6:464–473

29. McCracken C, Dalgard OS, Ayuso-Mateos JL, Casey P, Wil-kinson G, Lehtinen V, Dowrick C (2006) Health service use byadults with depression: community survey in five Europeancountries. Evidence from the ODIN study. Br J Psychiatry189:161–167

30. Minicuci N, Noale M, Bardage C, Blumstein T, Deeg DJH,Gindin J, Jylha M, Nikula S, Otero A, Pedersen NL, Pluijm SMF,Zunzunegui MV, Maggi S, for the CLESA working group (2003)Cross-national determinants of quality of life from six longi-tudinal studies on aging: the CLESA project. Aging Clin ExpRes 15(3):187–202

31. Minicuci N, Maggi S, Pavan M, Enzi G, Crepaldi G (2002)Prevalence rate and correlates of depressive symptoms in olderindividuals: the Veneto Study. J Gerontol A Biol Sci Med Sci57(3):M155–M161

32. Nikula S, Jylha M, Bardage C, Deeg DJH, Gindin J, Minicuci N,Pluijm SMF, Rodriguez-Laso A, for the CLESA working group(2003) Are IADLs comparable across countries? Sociodemo-graphic associates of harmonised IADL measures. Aging ClinExp Res 15(6):451–459

33. Pedersen NL, McClearn GE, Plomin R, Nesselroade JR, Berg S,deFaire U (1991) The Swedish adoption twin study of aging: anupdate. Acta Genet Med Gemellol 40:7–20

34. Piccinelli M, Wilkinson G (2000) Gender differences indepression. Br J Psychiatry 177:486–492

35. Pluijm SMF, Bardage C, Nikula S, Blumstein T, Jylha M,Minicuci N, Zunzunegui MV, Pedersen NL, Deeg DJH, for theCLESA working group (2005) Harmonizing a measure of ADL-disability across six countries. J Clin Epidemiol 58:1015–1023

36. Radloff LS (1977) The CES-D scale: a self-reported depressionscale for research in the general population. Appl PsycholMeasurement 1:385–401

37. Robison J, Gruman C, Gaztambide S, Blank K (2002) Screeningfor depression in middle-aged and older Puerto Rican primarycare patients. J Gerontol A Biol Sci Med Sci 57(5):M08–M314

38. Ruskin P, Blumstein T, Walter-Ginzburg A, Fuchs Z, Lusky A,Novikov I, Modan B (1996) Depressive symptoms amongcommunity-dwelling oldest-old residents in Israel. Am J GeriatrPsychiatry 4:208–217

39. Smith J, Baltes JM (1998) The role of gender in very old age:profiles of functioning and everyday life patterns. PsycholAging 13:676–695

40. Sonnenberg CM, Beekman ATF, Deeg DJH, Van Tilburg W(2000) Sex-differences in late-life depression. Acta PsychiatrScand 101:286–292

41. Takkinen S, Gold C, Pedersen NL, Malmberg B, Nilsson SE,Rovine MJ (2004) Gender differences in depression: a study ofolder unlike-sex twins. Aging Ment Health 8(3):187–195

42. Van Grootheest DS, Beekman ATF, Broese van Groenou MI,Deeg DJH (1999) Sex differences in depression after widow-hood. Do men suffer more? Soc Psychiatry Psychiatr Epidemiol34:391–398

43. Veijola J, Puukka P, Lehtinen V, Moring J, Lindholm T, Va-isanen E (1998) Sex differences in the association betweenchildhood experiences and adult depression. Psychol Med28:21–27

206

44. Walter-Ginzburg A, Shmotkin D, Blumstein T, Shorek A (2005)A gender-based multidimensional longitudinal analysis ofresilience and mortality in the old–old in Israel: the cross-sectional and longitudinal study CALAS. Soc Sci Med 60:1705–1715

45. Zunzunegui MV, Beland F, Llacer A (1998) Gender differencesin depressive symptomatology among Spanish elderly. SocPsychiatry Psychiatr Epidemiol 33:195–205

207