comparing sex inequalities in common affective disorders across countries: great britain and chile
TRANSCRIPT
ARTICLE IN PRESS
0277-9536/$ - se
doi:10.1016/j.so
�Correspond
sity of Bristol,
UK. Fax: +44
E-mail addr
Social Science & Medicine 60 (2005) 1693–1703
www.elsevier.com/locate/socscimed
Comparing sex inequalities in common affective disordersacross countries: Great Britain and Chile
Graciela Rojasa, Ricardo Arayab,�, Glyn Lewisb
aClinica Psiquiatrica, Facultad de Medicina, Universidad de Chile, Avda La Paz 1003, Santiago, ChilebDivision of Psychiatry, University of Bristol, Cotham House, Cotham Hill, Bristol BS6 6JL, UK
Abstract
Most studies throughout the world have found that women report more psychological symptoms than men. Much
less is known about possible variation between countries in the magnitude of these sex differences or the factors
contributing to the increase of risk among women in countries with different levels of development. This study aimed to
compare sex differences for common affective disorders (CAD) between Great Britain and Chile based on two large
urban cross-sectional psychiatric household surveys that used similar methodology. Women in both countries reported
more CAD than men but Chilean women had an increased risk in comparison to their British counterparts, a difference
that became larger as symptom severity increased. Of all the main explanatory variables included in the
analysis––education, employment status, children at home, marital status, and social support—the only statistically
significant interaction that could account for this increased risk was education, with an increasingly larger risk for
women with lower levels of educational attainments in Chile compared to Britain. Education is a powerful socio-
economic indicator that is difficult to revert later in life, especially in countries where opportunities for women are less
forthcoming, and it might act as powerful reminder of social entrapment.
r 2004 Elsevier Ltd. All rights reserved.
Keywords: Women; Mental disorders; Inequalities; Great Britain; UK; Chile
Introduction
Sex differences in the prevalence of common affective
disorders across countries
The commonest affective disorders (CAD), depression
and anxiety, are frequent and disabling in rich as well as
poor countries (Murray & Lopez, 1997). One of the
most consistent findings in psychiatric epidemiology is
that women, especially those living in urban settings,
e front matter r 2004 Elsevier Ltd. All rights reserve
cscimed.2004.08.030
ing author. Department of Psychiatry, Univer-
Cotham house, Cotham Hill, Bristol BS6 6UJ,
-117-954-66721.
ess: [email protected] (R. Araya).
seem to be at an increased risk of suffering from CAD
(Piccinelli & Wilkinson, 2002; Patel, Araya, Ludemir,
Todd & Lima, 1999; Bebbington, 1998). However,
less is known if the magnitude of these sex differences
in the prevalence of CAD is comparable across countries
or if the type of risk factors that might explain this
increased risk among women are similar in countries
with different levels of development. The use of
different methodologies to ascertain the presence of
psychological symptoms, the difficulties of comparing
unadjusted or partially adjusted results, and the
scarcity of large household surveys from developing
countries have somehow hindered the making of valid
comparisons across countries. This is an important issue
that interferes with the possibility of gaining further
d.
ARTICLE IN PRESSG. Rojas et al. / Social Science & Medicine 60 (2005) 1693–17031694
insight into the aetiology of psychiatric disorders
when examining similarities and differences between
countries.
How to explain differences between countries in the
prevalence of common affective disorders among women?
Although little is known of risk factors that could
differentially increase the risk of CAD among women in
some countries, it would seem unlikely that a biological
factor would be able to explain much of these
differences. There are no obvious discrepancies in the
biological make-up of women living in different
countries, other than those brought upon by living in
different social, cultural, and economic realities. Thus if
there were differences in the prevalence of CAD among
women from countries with different levels of develop-
ment, psychosocial factors are more likely to provide an
explanation for these differences. Psychosocial factors
that have been linked to an increased risk of psycholo-
gical morbidity and show unequal distribution between
countries are worth exploring, such as employment or
unemployment, socio-economic differences, the number
of dependent children, multiple roles, or low social
support (Araya, Lewis, Rojas & Fritsch, 2003; Piccinelli
& Wilkinson, 2002; Bebbington, 1998; Weich, Sloggett
& Lewis, 1998).
Several single-country studies, mostly from developed
countries, have investigated the effect of work on
women’s mental health. Most of these studies, but not
all, have been carried out as part of research investigat-
ing the impact of multiple roles (‘role strain hypothesis’)
on the mental health of women (Fokkema, 2002;
Matthews, Power & Stansfeld, 2001; Weich, Sloggett &
Lewis, 2001; Weich et al., 1998; Waldron, Weiss &
Hughes, 1998). As a whole, this research has failed to
provide good evidence in support of the ‘role strain’
hypothesis but, on the contrary, several of these studies
have shown that work was likely to have a positive effect
for the mental health of women, regardless of the
number of roles held simultaneously. Studies carried out
in the developing world (Iran, India, Brazil and Chile)
have failed to find differences in the mental health of
working and non-working women (Ahmad–Nia, 2002;
Araya, Rojas, Fritsch, Acuna & Lewis, 2001; Patel et al.,
1999). It is possible that the more unfavourable living
situation of women in developing countries might have
contributed to dilute any positive effects of employment
on women’s mental health. Other methodological issues,
such as for instance the possibility that psychologically
healthier women were more likely to take on employ-
ment, also interfere with reaching any firmer conclu-
sions. So in spite of the evidence suggesting beneficial
effects of employment it is not possible yet to assume
that work is beneficial for all women and under all
circumstances.
The evidence from developed countries suggesting
that education might be an important risk factor for
mental illness is scant and there is even less support for
an educational effect that could account for the gender
differences in the prevalence of CAD (Bebbington, 1996,
1998). Nonetheless, there are marked differences in
educational attainments within and between countries,
with women consistently achieving lower levels than
men especially in developing countries (World Bank,
2001a, b). There is stronger evidence to show that socio-
economic adversity is associated with the presence of
CAD (Lorant et al., 2003; Araya et al., 2001, 2003;
Weich, Lewis & Jenkins, 2001; Weich & Lewis, 1998;
Lewis et al., 1998); and that people, especially women, in
poorer countries are comparatively under more social
disadvantage than individuals in richer countries (World
Bank, 2001b; United Nations Population Fund, 2000;
Desjarlais, Eisenberg, Byron & Kleinman, 1995). Thus it
might be reasonable to expect that sex differences in
CAD could be larger in poorer than in richer countries.
British studies have also shown that the presence of two
or more young children at home and the lack of a
confiding partner were also associated with an increased
prevalence of depression in women (Bebbington, 1996;
Brown & Harris, 1978). Equally previous research had
consistently shown an inverse association between social
support and CAD, with less support associated with an
increased prevalence of CAD, but sex differences in this
association are less commonly reported (Berkman &
Glass, 2000).
There are many difficulties, though, when trying to
determine the relative importance of individual
psychosocial factors to increase the risk of CAD across
countries. Some authors have argued for the need to
include into the analysis the socio-economic, cultural,
and political context under which these factors
operate (Janzen & Muhajarine, 2003). Contextual
differences cannot be underestimated when comparing
countries with different cultures and levels of socio-
economic development. However, contextual variables
are rarely incorporated into the analysis for this kind of
research in mental health. Among possible reasons for
this omission are (1) difficulties in obtaining reliable and
comparable contextual data across countries, (2) varia-
tion in the relative importance of contextual factors
across cultures and over time within countries, and (3)
methodological difficulties in analysing this kind of data
adequately. We are unaware of any study comparing
mental health between countries that incorporates
contextual as well as individual variables into the
analysis. However, most studies usually consider con-
textual differences when interpreting results and reach-
ing conclusions.
The social context under which sex differences in
CAD present is complex in the developing world. For
instance, many international agencies and governments
ARTICLE IN PRESSG. Rojas et al. / Social Science & Medicine 60 (2005) 1693–1703 1695
see development as closely linked to increasing produc-
tivity through an enlarged, better educated, and
healthier workforce. Women’s participation in the
workforce is rising in almost every developing country
but their share of the better jobs or educational
opportunities still lag far behind men, a gender gap
much wider than in most developed countries (World
Bank, 2001a, b; United Nations Population Fund,
2000). In most developing countries, there has been a
slow introduction of the necessary social changes to
support the incorporation of women into the workforce
(World Bank, 2001a, b; United Nations Population
Fund, 2000). Other important demographic changes
are also occurring such as a decrease in the size of
families, not only in terms of the number of children but
also in the size and ramifications of the extended family,
the traditional provider of social support. The rapid
social transformations that developing countries are
experiencing are likely to have an impact on the mental
health of their populations, an effect that might be more
pronounced among women who seemed to be enduring
more significant changes. Thus it is important to
understand this phenomenon in countries outside the
western world, where most of the world’s population
lives.
This study brings together data from two large and
well-conducted psychiatric household surveys carried
out in Great Britain and Chile (Araya et al., 2001;
Jenkins et al., 1997) using comparable methodologies
and thus allowing comparisons to be made. These
countries represent two contrasting cultures with differ-
ent levels of development. Whilst Great Britain is
considered a high-income country, Chile belongs to the
middle-income, emerging economies group. It is worth
noting, however, that these broad international classifi-
cation systems often fail to capture some aspects of the
social differences between countries. For instance, the
proportion of women participating in the labour force in
Chile is 33%, a figure lower than the average 40% found
in poorer countries (World Bank, 2001b). Equally,
although Chile is regarded as a middle-income country,
it is also one of the ten most unequal countries in the
world as far as income is concerned (World Bank,
2001b).
Our main hypothesis was that women in both
countries have increased risks of CAD in comparison
to men, but Chilean women would have an increased
risk of CAD compared to their British counterparts.
This risk difference between Chilean and British women
would be even more pronounced as symptoms become
more severe. If our hypothesis were correct, we would
undertake an exploratory analysis to find out if there
were differences also in the type of risk factors
associated with CAD in women. We focused this part
of the analysis on social variables that have previously
been related to CAD: educational attainments, working
status, number of children, marital status, and social
support.
Methods
Participants and sampling
This paper used data from the ‘Santiago Mental
Disorders Survey’ undertaken in 1996 in Santiago, Chile
(Araya et al., 2001), and the ‘Great Britain National
Survey of Psychiatry Morbidity’ (Jenkins et al., 1997)
undertaken in 1993. Both studies used similar meth-
odologies. However, the Chilean sample was restricted
to Santiago, capital of Chile, where more than 50% of
the total population lives, whereas the British study used
a nationally representative sample. For the purpose of
this paper, only data collected in urban British settings
was used in order to facilitate the comparisons.
Santiago, Chile: Households within sectors from all
the 35 boroughs of Santiago were randomly chosen with
a probability proportional to the size of the sampling
unit using a three-stage clustered design. A larger
sampling fraction was required in the most affluent
boroughs to allow testing for socio-economic differences
between groups. The sampling framework was the total
adult (aged 16–64) population living in private house-
holds of Santiago, representing 3,217,177 individuals.
Using Kish tables one person per household was chosen
at random. The response rate was 90%. Further details
of the sampling design can be found in previous
publications or requested from the authors (Araya et
al., 2001).
Urban, Great Britain: A stratified, cluster, probability
sample was drawn for the UK excluding the Highlands
and the Islands of Scotland. Two hundred postal sectors
were selected with probability proportional to size and
90 addresses were randomly selected from each sector.
One individual, aged 16–64, in each household was
interviewed. Individuals living in rural or semi-rural sites
as well as ‘proxy’ interviews were excluded. Experienced
interviewers from the Office of National Statistics
classified the households as urban, rural, or semi-rural
according to their pre-established criteria (Meltzer, Gill,
Petticrew & Hinds, 1995). The response rate was 79%
for the overall sample. The main findings have already
been published and further details can be obtained
elsewhere (Jenkins et al., 1997).
Variables
Individuals who met ICD-10 criteria for depressive
episode or anxiety disorders (generalised anxiety, phobia
or panic) were classified as suffering from a Common
Affective Disorder, our main outcome variable. In view
ARTICLE IN PRESSG. Rojas et al. / Social Science & Medicine 60 (2005) 1693–17031696
of the high co-morbidity of these disorders and the
questionable validity of some psychiatric diagnoses we
think it is appropriate to treat these disorders as a single
variable. Diagnostic criteria were assessed using the
Revised Clinical Interview Schedule (CIS-R) (Lewis,
Pelosi, Araya & Dunn, 1992), a structured interview that
has been fully standardised so that it can be adminis-
tered by lay interviewers. This interview has 14 sections,
each covering a specific neurotic dimension with its own
individual score. All scores can be summed to yield a
total score that provides a rough estimate of the severity
of psychiatric symptomatology. The CIS-R in its
English and Spanish versions has been used extensively
in primary care and community studies with validity and
reliability comparable to other commonly used struc-
tured interviews in mental health (Brugha et al., 2000;
Brugha, Bebbington & Jenkins, 1999; Andrews & Peters,
1998; Lewis et al., 1992).
Besides sex (male and female) the following variables
were included in the analysis:
1.
Education summarised in four categories according toincreasing levels of achievements in the British sample
(A-level, GCSE grade A–C, GCSE grade D–F, and
no qualifications) or increasing years of education in
Chile (primary incomplete, primary complete, sec-
ondary incomplete, and secondary complete or
greater). These are the classification systems used by
the Office of National Statistics in both countries
with the purpose of measuring educational level.
Even though these are probably valid ways of
measuring educational achievements in each country,
it does not mean that there is equivalence in the
categories across countries. For instance, we do not
know if these categories are equivalent in terms of the
opportunities arising out of achieving a certain level,
a problem that would persist even if we had used the
number of years of education completed in both
countries. Similarly, the equivalence of categories
between countries for almost all other potential risk
factors included in this study could be questioned on
similar grounds. Educational systems are different
across the world and we have to find ways of
comparing the impact of education on various
outcomes of interest. There will be limitations on
whatever method is chosen to compare different
populations, an issue that invites caution when
interpreting the results. However, we think that, on
balance, our educational categories represent differ-
ent levels of educational achievements in both
countries and allow our samples to be sufficiently
spread in categorical sub-groups.
2.
Working status: Individuals were classified in thefollowing categories: employed, unemployed, house-
keeper, inactive representing students and pensioners,
and permanently unable to work for health reasons.
In order to qualify for the unemployment category,
the individual had to be out of work and actively
seeking it. Individuals temporarily out-of-work for
health reasons were also included as unemployed as
well as those who were working informally and
concomitantly seeking a job.
3.
The number of young children comprised own childrenaged 15 or less living in the household. This variable
was divided into four categories: no kids, one kid,
two kids, and more than two kids.
4.
Marital status divided into married, co-habiting,single, widowed, and divorced/separated.
5.
Social support representing the self-reported numberof people (friends and/or relatives) who could provide
either practical or emotional support if needed
grouped in three categories: low (less than three),
middle (between three and four), and high (more than
four).
In addition, the following variables were also included
for adjustment in the analysis:
1.
Age treated as a continuous variable.2.
The presence of a self-reported physical disease inresponse to this question: ‘Do you have any long-
standing illness, disability, or infirmity?’
3.
The number of units of alcohol consumed weeklysubdivided into five categories: abstained, low,
middle, high, and very high.
Statistical analysis
The association between the dependent variable—
CAD—and sex was examined by calculating crude and
adjusted odds ratios and their 95% confidence intervals
using logistic regression models. Confidence intervals
were calculated using Huber White Robust estimator of
variance, which takes account of the clustered nature of
the samples (Huber, 1981). All models were also
adjusted for household size to take account of the
different sampling fractions. In order to test for an
increased risk of suffering from CAD among Chilean
women, we tested for interactions between country and
sex using likelihood ratio tests (LR tests). We also
investigated if this increased risk could be due to
symptom severity differentials in these two countries
by using different thresholds for caseness according to
CIS-R scores.
Subsequently, we carried out an exploratory analysis
restricting our sample to women only in order to test for
interactions between country and some other variables
of interest that could account for the possible increased
risk among Chilean women. We only tested for
interactions after ensuring there was no co-linearity
between variables. Testing for interactions also allowed
ARTICLE IN PRESSG. Rojas et al. / Social Science & Medicine 60 (2005) 1693–1703 1697
us to explore for an association between women holding
more than one role and the presence of CAD in each
country separately. For instance, we were able to test for
an increase in the risk of being employed among
housewives in Chile, which could have been larger than
the sum of the two independent risks. Although weights
can be applied to adjust for sampling procedures and
obtain population estimates, we decided to use un-
weighted data because our main focus was to establish
associations between groups and countries rather than
making population estimates. All analysis was per-
formed using STATA Version 7.0 (STATA, 2001).
Results
In the British sample, 3434 individuals living in rural
or semi-urban sites as well as 117 ‘proxy’ interviews were
excluded. Thus the total British sample of urban living
interviewees comprised 6556 individuals. In the Chilean
sample, all 3870 completed interviews were from people
living in urban settings and thus were used for the
comparisons. The characteristics of the samples can be
seen in Table 1.
Table 1
Characteristics of the samples by sex and country Santiago, Chile, an
Chile
Women
(N=2332)
Mean age (95% CI) 37.8 (37.2–3
Education (%) Lowest 19
Middle low 24
Middle high 36
Highest 22
Marital status (%) Married 54.8
Cohabiting 4.0
Single 27.4
Widowed 4.9
Separated 8.8
Working status (%) Employed 36.2
Housekeeper 38.1
Pensioner/student 15.7
Retired ill-health 0.8
Unemployed 9.2
Perceived social support (%) High 41.5
Middle 36.2
Low 22.4
Number of young children (%) No child 60.4
One child 17.8
Two children 14.2
4Two children 7.6
There were some important differences in the samples
between and within countries. When comparing the two
countries, the British sample was older than the Chilean
[mean 39 (95% CI 39–40) vs. 37 (36–37); p ¼ 0:000).
Within countries, women were older than men in Chile
whereas in the British sample mean ages were similar for
both sexes. In both countries, males had achieved higher
levels of education than females. Official statistics using
indicators such as enrolment in primary education tend
to show a smaller educational gap between sexes in both
countries (World Bank, 2001b). Comparisons of educa-
tional achievements between men and women within
countries are not affected by the different methods used
to classify education in this study.
Also in keeping with official statistics, a higher
proportion of British women and men were co-habiting
and separated in comparison to their Chilean counter-
parts. The largest differences between the two countries
were to be found in working status and levels of
perceived social support. More British women were
employed whereas more Chilean women were house-
keepers, a finding compatible with other international
datasets (World Bank, 2001b). There were more
students among the Chileans and more people retired
d urban Great Britain. Un-weighted data
Great Britain
Men Women Men
(N=1538) (N=3548) (N=3008)
8.3) 35.4 (34.7–36.1) 39.4 (38.9–39.8) 39.2 (38.7–39.7)
15 36 27
25 11 10
29 26 23
31 26 39
50.8 49.8 52.2
2.4 7.2 7.3
39.6 22.1 29.5
1.8 5.5 1.4
5.4 15.5 9.6
63.7 58.9 71.4
2.0 21.2 1.2
22.2 9.7 7.6
1.2 4.0 6.2
10.9 6.2 13.6
48.5 54.6 61.7
33.9 37.2 29.6
17.6 8.2 8.7
69.3 57.5 67.3
14.4 18.4 13.5
10.7 16.5 13.8
5.7 7.6 5.4
ARTICLE IN PRESSG. Rojas et al. / Social Science & Medicine 60 (2005) 1693–17031698
on health-grounds among the British, possibly reflecting
the age structure of the two samples. Levels of
unemployment were in keeping with seasonally adjusted
official statistics in the British sample. However,
unemployment rate was marginally higher than officially
reported at the time of data collection in Chile
(Ministerio de Planificacion Nacional, 1998). However,
it must be noted that in Chile official reports do not
include as unemployed those working informally but
still seeking employment as we did in this study. A much
larger proportion of the Chilean sample reported low
levels of social support compared to their British
counterpart. Men reported higher levels of perceived
social support than women in both countries. The
number of young children at home in both countries was
similar.
Chilean and British women were significantly more
likely to be suffering from a CAD than their male
counterparts in both countries, even after adjusting for
all other variables in the models (Table 2). Most
importantly, there was a statistically significant interac-
tion between sex and country for CAD before and after
adjustments (LR test w2 ¼ 13:19; p ¼ 0:001 after adjust-
ments in the full model), in which Chilean women
showed an increased likelihood of suffering from a CAD
in comparison to British women.
Chilean women were increasingly more likely to be
psychiatric cases than their male counterparts as the
CIS-R threshold increased; in other words,
treating milder conditions as non-cases only contributed
to accentuate gender differences in Chile. On the
contrary, no major sex differences in the likelihood of
being a case were found with different CIS-R thresholds
for caseness in the British sample. As a result of
these differences between sexes in the severity of
symptoms in each country, the strength of the interac-
tions between sex and country became more prominent
at the most severe end of the symptom severity spectrum
(Table 3).
Table 2
The association between common affective disordersa (CAD) and sex
logistic regression modelling. Un-weighted data
Common affective disorders
Prevalence % (95% CI) Cr
Chile Men 6.8 (5.6–8.2) 1.0
Women 15.3 (13.9–16.8) 2.5
Great Britain Men 8.2 (7.2–9.2) 1.0
Women 11.3 (10.3–12.4) 1.4
aICD-10 depressive, generalised anxiety, panic, and phobia disordebAdjusted by age, marital status, education, employment status,
consumption, and household size.
In view of these statistically significant interactions
showing an increased risk for Chilean women, we
proceeded with the exploratory analysis, restricted to
women only, to investigate possible explanations that
could account for these findings. Among all the
variables examined, education was the only one that
showed a statistically significant interaction with coun-
try, in which Chilean women had an increasingly greater
risk of CAD than British women as levels of educational
attainments decreased (LR test w2 ¼ 12:95; p ¼ 0:005; in
the fully adjusted model). The group of British women
with the lowest educational levels did show a statistically
significant increase in the prevalence of CAD compared
to the best-educated women, but this effect disappeared
in the fully adjusted model. We were unable to find any
other statistically significant interaction (po0:05) be-
tween the other explanatory variables studied and
countries that could account for the increased risk
among Chilean women. The only exception was marital
status and country, in which, contrary to what we
expected, Chilean widows showed a much lower like-
lihood of suffering from CAD than their British
counterparts (LR test w2 ¼ 13:3; p ¼ 0:01 in the fully
adjusted model). In the unadjusted model, we found a
statistically significant interaction (LR test w2 ¼ 11:9;p ¼ 0:02) between working status and country also,
mainly as a result of unemployed Chilean women having
a decreased risk in comparison to their British counter-
parts.
A more descriptive approach investigating risk factors
for women in each country separately revealed more
similarities than disparities across countries (Table 4).
Although differences in risk between countries are better
captured when testing for interactions, studying risk
factors within countries is useful to provide a more
comprehensive picture as to where some of the
differences might lie.
Poor levels of social support, separation, and retire-
ment due to illness showed consistent associations with
by country. Prevalence, crude and adjusted odds ratios using
ude odds ratio (95% CI) Adjusted odds ratiob (95% CI)
0 1.00
2 (2.00–3.18) 2.16 (1.64–2.85)
0 1.007 (1.24–1.74) 1.29 (1.04–1.62)
rs.
children under 15, social support, physical disease, alcohol
ARTICLE IN PRESS
Table
3
The
ass
oci
ation
and
inte
ract
ions
bet
wee
nse
x,co
untr
ies,
and
the
sever
ity
of
psy
chia
tric
sym
pto
ms
(CIS
-Rsc
ore
s)using
diffe
rent
thre
shold
s.A
dju
sted
odds
ratios
using
logistic
regre
ssio
nm
odel
ling.U
n-w
eighte
ddata
Case
sC
IS-R
X6
Case
sC
IS-R
X12
Case
sC
IS-R
X18
Pre
vale
nce
%
(95%
CI)
Odds
ratios
(95%
CI)
a
Pre
vale
nce
%
(95%
CI)
Odds
ratios
(95%
CI)
a
Pre
vale
nce
%
(95%
CI)
Odds
ratios
(95%
CI)
a
Chile
Men
35.2
(32.8
–37.6
)1.0
013.8
(12.1
–15.6
)1.0
05.9
(4.7
–7.1
)1.0
0
Wom
en51.1
(49.1
–53.1
)1.9
4(1
.64–2.2
9)
29.3
(27.4
–31.1
)2.4
8(2
.04–3.0
2)
17.0
(15.5
–18.5
)3.1
8(2
.39–4.2
2)
Gre
atB
rita
inM
en29.8
(28.1
–31.5
)1.0
012.9
(11.7
–14.2
)1.0
06.7
(5.8
–7.6
)1.0
0
Wom
en44.7
(43.1
–46.4
)2.1
1(1
.86–2.3
9)
21.4
(20.1
–22.8
)2.0
5(1
.74–2.4
2)
11.4
(10.3
–12.4
)1.9
3(1
.55–2.4
1)
Countr
y–se
xin
tera
ctio
nb
P(w
2)
0.743(0.11)
0.019(5.52)
0.000(12.16)
aA
dju
sted
by
age,
marita
lst
atu
s,ed
uca
tion,em
plo
ym
entst
atu
s,ch
ildre
nunder
15,so
cialsu
pport
,physica
ldisea
se,alc
oholco
nsu
mption,and
house
hold
size
.bLik
elih
ood
Ratio
test
sw
ere
use
dfo
rte
stin
gin
tera
ctio
ns
infu
lly
adju
sted
model
s.
G. Rojas et al. / Social Science & Medicine 60 (2005) 1693–1703 1699
CAD in both countries. Some other associations with
CAD differed between the countries but most of these
differences did not reach statistical significance
(po0:05). For instance, housekeeping or co-habitation
among British and Chilean women, respectively, were
associated with an increased risk in their own countries
but these factors did not help to account for the
increased risk among Chilean women over and above
the risk in Great Britain.
Discussion
In keeping with most previous literature, women in
both countries showed an increased prevalence of CAD
compared to men, but this risk was much larger for
Chilean women, especially for those with more severe
symptomatology. Of all the variables examined, educa-
tional level showed the only statistically significant
interaction that could somehow account for this
difference in the prevalence of CAD between Chilean
and British women, with less educated Chilean women
showing the largest risk after adjusting for other
variables.
The main strength of this study was the use of similar
methodologies in both sites, including a detailed
psychiatric interview administered to a large and
representative sample with high response rates in both
surveys. Nonetheless, the cross-sectional design used
limited the conclusions about causality. Although this
study compared only urban samples from two countries,
these samples represented contrasting social, cultural,
and economic realities. There are obvious limitations in
terms of the depth of knowledge that can be obtained
about a particular risk factor when conducting large
general household surveys. Although much of the
interview used with both samples was the same, there
were some inevitable differences, for example when
measuring educational achievements, because categories
have to represent meaningful local constructs.
Sex differences, common affective disorders, and
variations between countries
Women in both countries showed an increased risk of
CAD compared to men, but Chilean women showed a
larger risk of suffering from CAD in comparison to their
British counterparts even after adjusting for a wide
range of variables. This finding was even more
pronounced when restricting the analysis to more severe
cases, suggesting that these differences were not due to
an excess of ‘the worried but well’ cases within the
Chilean sample. Thus, it is conceivable that other
developing countries with large gender gaps in education
and socio-economic status will have similar differences
in mental illness.
ARTIC
LEIN
PRES
S
Table 4
The association between CAD, marital and working status, social support, and number of young children among urban women in Chile and Great Britain
Chile Great Britain
Crude ORa Adj. ORa,b Adj. ORa,c Crude ORa Adj. ORa,b Adj. ORa,c
Education Highest 1.00 1.00 1.00 1.00 1.00 1.00
Middle high 2.66 (1.78–3.96) 2.64 (1.77–3.96) 2.18 (1.44–3.29) 1.05 (0.78–1.41) 1.05 (0.79–1.41) 0.91 (0.67–1.22)
Middle low 3.10 (2.05–4.61) 3.23 (2.12–4.93) 2.40 (1.55–3.70) 1.18 (0.79–1.76) 1.15 (0.76–1.73) 0.86 (0.57–1.31)
Lowest 3.85 (2.49–5.96) 4.15 (2.58–6.67) 2.68 (1.64–4.37) 1.81 (1.44–2.28) 1.74 (1.36–2.23) 1.15 (0.89–1.51)
Marital status Married 1.00 1.00 1.00 1.00 1.00 1.00
Cohabiting 2.02 (1.25–3.26) 1.84 (1.12–3.04) 1.57 (0.96–2.56) 1.28 (0.84–1.95) 1.42 (0.91–2.23) 1.36 (0.87–2.14)
Single 0.87 (0.67–1.13) 0.86 (0.60–1.23) 1.06 (0.73–1.54) 1.19 (0.91–1.56) 1.28 (0.92–1.78) 1.10 (0.78–1.55)
Widowed 0.72 (0.40–1.30) 0.86 (0.46–1.60) 0.77 (0.41–1.46) 2.43 (1.58–3.73) 2.17 (1.38–3.41) 1.77 (1.12–2.82)
Separated 1.66 (1.15–2.40) 1.63 (1.11–2.39) 1.66 (1.11–2.47) 1.77 (1.32–2.36) 1.83 (1.37–2.45) 1.23 (0.89–1.72)
Working status Employed 1.00 1.00 1.00 1.00 1.00 1.00
Housekeeper 1.41 (1.08–1.84) 1.49 (1.14–1.94) 1.23 (0.91–1.65) 1.83 (1.38–2.41) 1.82 (1.38–2.41) 1.46 (1.05–2.02)
Inactive 0.74 (0.50–1.09) 0.72 (0.44–1.20) 0.84 (0.50–1.41) 1.17 (0.78–1.77) 1.07 (0.68–1.69) 1.01 (0.63–1.61)
Retired ill-health 4.73 (1.97–11.3) 4.05 (1.47–11.2) 3.46 (1.19–10.0) 6.17 (4.37–8.70) 5.60 (3.83–8.19) 4.30 (2.77–6.68)
Unemployed 1.11 (0.72–1.72) 1.08 (0.69–1.67) 0.95 (0.60–1.48) 2.93 (1.99–4.31) 3.01 (2.02–4.49) 2.48 (1.64–3.76)
Social support High 1.00 1.00 1.00 1.00 1.00 1.00
Middle 1.75 (1.28–2.39) 1.67 (1.22–2.28) 1.52 (1.10–2.10) 1.61 (1.26–2.05) 1.62 (1.27–2.06) 1.42 (1.11–1.82)
Low 3.75 (2.74–5.13) 3.58 (2.59–4.94) 2.99 (2.13–4.19) 3.49 (2.54–4.79) 3.49 (2.51–4.86) 2.57 (1.78–3.71)
Number of young children No child 1.00 1.00 1.00 1.00 1.00 1.00
One child 1.24 (0.91–1.67) 1.13 (0.80–1.58) 0.91 (0.64–1.28) 1.16 (0.81–1.67) 1.46 (0.99–2.14) 1.15 (0.78–1.69)
XTwo children 1.42 (1.04–1.94) 1.26 (0.86–1.84) 1.04 (0.70–1.53) 1.47 (0.96–2.25) 2.12 (1.31–3.42) 1.41 (0.85–2.35)
aAll models estimated using Huber-White robust variance estimator adjusting standard errors for clustering of geographical areas.bAdjusted by age, physical disease, alcohol consumption, and household size.cAdjusted by age, physical disease, alcohol consumption, household size, and other variables in the table.
G.
Ro
jas
eta
l./
So
cial
Scien
ce&
Med
icine
60
(2
00
5)
16
93
–1
70
31700
ARTICLE IN PRESSG. Rojas et al. / Social Science & Medicine 60 (2005) 1693–1703 1701
How to explain the increased risk among Chilean women?
Biological mechanisms are unlikely to explain these
sex differences across countries. Yet it is likely that
Chilean women were exposed to more adverse environ-
mental conditions, increasing the risk for organic injury
or physical illnesses. Against this line of reasoning, we
found that the proportion of women who self-reported
poor general health was similar in both countries.
Incidentally, this argues against the response-bias claim
that Chilean women have higher rates because they are
more likely to acknowledge the presence of symptoms or
illnesses than their British counterparts (Macintyre,
Ford & Hunt, 1999). Nonetheless, we cannot rule out
the possibility of a more specific response bias for
psychological questions, though we attempted to mini-
mise this by using a psychiatric interview rather than a
questionnaire. Thus, on balance, we think it is unlikely
that a response bias can account for the reported
differences.
Our most important finding in this respect was a
statistically significant interaction between education
and country for women, in which Chilean women with
lower educational achievements had a significantly
higher risk of CAD than less educated British women.
Our results showed a robust, dose–response, and
independent inverse association between education and
CAD even after adjusting for all other variables among
Chilean women. Although we found that British women
with the lowest educational level were significantly more
likely to suffer from CAD than the best educated, this
difference disappeared in the fully adjusted model.
Admittedly, we used different methods to measure
educational attainments in the two countries, an
inevitable move in view of the different educational
systems. Although it is not known whether the educa-
tional groupings are equivalent in terms of their impact
on mental health in the two countries, we are confident
that our groupings represent different levels of educa-
tional attainments with practical implications for future
opportunities in each country. Nonetheless, we advise
caution when interpreting this finding.
Bearing in mind these limitations, how could
education account for this greater risk among Chilean
women? Education is a relatively frozen socio-economic
indicator of earlier life that continues to exert its
effect throughout life; for instance, through giving
access to opportunities. It is possible that social
disadvantage in earlier life might have a long-lasting
effect on mental health or that individuals with poorer
education have an increased likelihood of accumulating
more adversity throughout life. Equally it is possible
that lower levels of education might simply be another
indicator of lower socio-economic status and it is this
overall position of social disadvantage that increases the
risk of CAD.
Regardless of the possible explanations for and
limitations to interpreting this finding, the difference in
educational attainment between Chilean and British
women is noteworthy. At a time when the gender gap at
the primary educational level seems to be decreasing,
women are still less likely to receive secondary and
post-secondary education in most developing
countries (United Nations Population Fund, 2000).
Even though the comparability of educational achieve-
ments in the two countries is debatable, our results,
together with those from other studies, should motivate
further research into the impact of education on mental
health.
Can we learn something from the similarities between the
two countries?
Although shared risk factors did not help us to
explain the increased risk among Chilean women, they
provided an interesting insight into situations of risk
that seemed to transcend national boundaries. In
keeping with this predicament, the most important
similarity was that employment did not increase the
likelihood of suffering from CAD in either of these two
countries, a finding which is common with other
research from developed (Fokkema, 2002; Weich et al.,
1998) as well as developing countries (Ahmad-Nia,
2002). So it seems that employment could have more
advantages than disadvantages for women’s mental
health, even in countries where women might be
working under worse conditions (Loewenson, 1999).
Along similar lines, the more difficult conditions under
which women might be working in Chile could have
somehow attenuated the positive effect of employment
when compared with other working status categories.
The association between working women and mental
health is a complex issue and many studies so far
undertaken, including this one, might have not had
adequate depth to be able to explain how, for instance,
holding multiple roles can affect the mental health of
working women. Likewise, it is possible that the more
resourceful and healthy women were more likely to be
employed; thus confounding our results (Emslie et al.,
2002; Matthews et al., 2001). Contrary to this, we found
that the proportion of employed women who self-
reported poor general health was much larger among
British than Chilean women (33% vs. 18%). We also
adjusted our results for the presence of ill health when
testing for interactions. We found no statistically
significant interactions (po0:05) in each country sepa-
rately between working status and being a mother,
regardless of the number of children.
We found that in both countries poorer social support
was strongly associated with an increased likelihood of
suffering from a CAD. However, we found that this
association was also present among men and there were
ARTICLE IN PRESSG. Rojas et al. / Social Science & Medicine 60 (2005) 1693–17031702
no interactions between social support and country
among women, in keeping with other studies (Piccinelli
& Wilkinson, 2002). We also found that in both
countries there was no association between CAD and
the presence of young children at home either, before or
after adjusting for other socio-demographic variables.
The presence of two or more children under the age of
11 had been previously considered a risk factor for
depression among British women (Brown & Harris,
1978; Bebbington, 1996).
Conclusion
The most important finding of this study is the
confirmation that women’s mental health in Chile and
possibly other less developed countries could be at
much-increased risk in comparison to women in more
westernised societies. This is an issue with immense
humanitarian, political, and economic consequences
that has been poorly researched so far. Health research
programmes for developing countries have often had a
narrow focus on reproductive health, a priority that is
not necessarily shared by local communities. For
instance, in Ghana almost three-quarters of women
identified psychosocial but not reproductive health
problems as their most important health concerns
(Avotri & Walters, 1999). The scope of research on
women’s health needs to be broadened; it is no longer
acceptable to conceive women mainly from a reproduc-
tive point of view. Women play much more diverse and
important roles in modern developing societies. It is time
that other more complex social issues affecting the
health and welfare of women, particularly from the
developing world, are properly addressed.
Acknowledgements
We would like to thank Drs. R. Fritsch, J. Acuna, and
M. Horvitz-Lennon for their participation in the field-
work. We would also like to express our gratitude to all
the interviewers who participated in this study and, most
important of all, to the people who took unpaid time to
answer our questions. This study was funded by the
European Community (EC).
References
Ahmad-Nia, S. (2002). Women’s work and health in Iran: a
comparison of working and non-working mothers. Social
Science & Medicine, 54, 753–765.
Andrews, G., & Peters, L. (1998). The psychometric properties
of the Composite International Diagnostic Interview. Social
Psychiatry and Psychiatric Epidemiology, 33, 80–88.
Araya, R., Lewis, G., Rojas, G., & Fritsch, R. (2003).
Education and income: which is more important for mental
health? Journal of Epidemiology and Community Health, 57,
1–13.
Araya, R., Rojas, G., Fritsch, R., Acuna, J., & Lewis, G.
(2001). Common mental disorders in Santiago, Chile.
Prevalence and socio-demographic correlates. British Jour-
nal of Psychiatry, 178, 228–233.
Avotri, J. Y., & Walters, V. (1999). You just look at your work
and see if you have any freedom on earth: Ghanaian
women’s accounts of their work and health. Social Science
& Medicine, 48, 1123–1133.
Bebbington, P. (1996). The origins of sex differences in
depressive disorder: bridging the gap. International Review
of Psychiatry, 8, 295–332.
Bebbington, P. (1998). Sex and depression. Psychological
Medicine, 28, 1–8.
Berkman, L., & Glass, T. (2000). Social integration, social
networks, social support, and health. In Berkman, L., &
Kawachi, I. (Eds.), Social epidemiology (pp. 137–173). New
York: Oxford University Press.
Brown, A. C., & Harris, T. O. (1978). Social origins of
depression: a study of psychiatric disorder in women. New
York: Free Press.
Brugha, T. S., Bebbington, P. E., & Jenkins, R. (1999). A
difference that matters: comparisons of structured and semi-
structured psychiatric diagnostic interviews in the general
population. Psychological Medicine, 29, 1013–1020.
Brugha, T. S., Bebbington, P. E., Jenkins, R., Meltzer, H.,
Taub, N. A., Janas, M., & Vernon, J. (2000). Cross
validation of a general population survey diagnostic inter-
view: a comparison of CIS-R with SCAN ICD-10 diag-
nostic categories. Psychological Medicine, 29, 1029–1042.
Desjarlais, R., Eisenberg, L., Byron, G., & Kleinman, A.
(1995). World mental health: problems and priorities in low
income countries. New York: Oxford University Press.
Emslie, C., Fuhrer, R., Hunt, K., Macintyre, S., Shipley, M., &
Stansfeld, S. (2002). Gender differences in mental health:
evidence from three organisations. Social Science &
Medicine, 54, 621–624.
Fokkema, T. (2002). Combining a job and children: contrasting
the health of married and divorced women in the Nether-
lands? Social Science & Medicine, 54, 741–752.
Huber, P. J. (1981). Robust statistics. New York: Wiley.
Janzen, B. L., & Muhajarine, N. (2003). Social role occupancy,
gender, income adequacy, life stage and health: a long-
itudinal study of employed Canadian men and women.
Social Science & Medicine, 57, 1491–1503.
Jenkins, R., Lewis, G., Bebbington, P., Brugha, T., Farrell, M.,
Gill, B., & Meltzer, H. (1997). The National Psychiatric
Morbidity Surveys of Great Britain: initial findings
from the Household Survey. Psychological Medicine, 27,
775–790.
Lewis, G., Bebbington, P., Brugha, T., Farrell, M., Gill, B.,
Jenkins, R., & Meltzer, H. (1998). Socioeconomic status,
standard of living and neurotic disorder. Lancet, 352,
605–609.
Lewis, G., Pelosi, A. J., Araya, R., & Dunn, G. (1992).
Measuring psychiatric disorder in the community: a
standardised assessment for use by lay interviewers.
Psychological Medicine, 22, 465–486.
ARTICLE IN PRESSG. Rojas et al. / Social Science & Medicine 60 (2005) 1693–1703 1703
Loewenson, R. H. (1999). Women’s occupational health in
globalisation and development. American Journal of In-
dustrial Medicine, 36, 34–42.
Lorant, V., Deliege, D., Eaton, W., Robert, A., Phillippot, P.,
& Ansseau, M. (2003). Socioeconomic inequalities in
depression: a meta-analysis. American Journal of Epidemiol-
ogy, 157, 98–112.
Macintyre, S., Ford, G., & Hunt, K. (1999). Do women ‘over-
report’ morbidity? Men’s and women’s responses to
structured prompting on a standard question on long
standing illness. Social Science & Medicine, 48, 89–98.
Matthews, S., Power, C., & Stansfeld, S. (2001). Psychological
distress and work and home roles: a focus on socio-
economic differences in distress. Social Science & Medicine,
31, 725–736.
Meltzar, H., Gill, B., Petticrew, M., & Hinds, K. (1995). OPCS
Surveys of Psychiatry Morbidity. Report I. The prevalence
of psychiatric morbidity among adults aged 16–64 living in
private households in Great Britain. London: HMSO.
Ministerio de Planificacion Nacional (1998). Encuesta de
caracterizacion socioeconomica nacional. Santiago, Chile:
MIDEPLAN.
Murray, C., & Lopez, A. (1997). Alternative projections of
mortality and disability by cause 1900–2020: global burden
of disease study. Lancet, 349, 1498–1504.
Patel, V., Araya, R., Ludemir, A., Todd, C., & Lima, M.
(1999). Women, poverty and common mental disorders in
four restructuring societies. Social Science & Medicine, 49,
1461–1471.
Piccinelli, M., & Wilkinson, G. (2002). Gender differences in
depression: critical review. Canadian Journal of Psychiatry,
177, 486–492.
STATA (2001). Stata version 7.0. College Station, Texas: Stata
Corporation.
United Nations Population Fund (2000). The state of the world
population 2000: lives together, worlds apart—men and
women in a time of change. New York: United Nations.
Waldron, I., Weiss, C. C., & Hughes, M. E. (1998). Interacting
effects of multiple roles on women’s health. Journal of
Health and Social Behaviour, 39, 216–236.
Weich, S., & Lewis, G. (1998). Poverty, unemployment and the
common mental disorders: a population based cohort study.
British Medical Journal, 317, 115–119.
Weich, S., Lewis, G., & Jenkins, S. P. (2001). Income inequality
and the prevalence of common mental disorders in Britain.
British Journal of Psychiatry, 178, 222–227.
Weich, S., Sloggett, A., & Lewis, G. (1998). Social roles and
gender difference in the prevalence of common mental
disorders. British Journal of Psychiatry, 173, 489–493.
Weich, S., Sloggett, A., & Lewis, G. (2001). Social roles and the
gender difference in rates of the common mental disorders
in Britain: a 7-year, population-based cohort study.
Psychological Medicine, 31, 1055–1064.
World Bank (2001a). Engendering development: through gender
equality in rights, resources, and voice. Washington, DC:
World Bank.
World Bank (2001b). World development report 2000/2001.
Attacking poverty. New York: Oxford University Press.