debt, social disadvantage and maternal depression
TRANSCRIPT
Social Science & Medicine 53 (2001) 441–453
Debt, social disadvantage and maternal depression
Richard Reading*, Shirley Reynolds1
School of Health Policy and Practice, University of East Anglia, Norwich NR4 7TJ, UK
Abstract
Depression is common among women with young children, and is strongly associated with financial adversity. Debtis a common feature of such adversity, yet its relationship with depression has not been examined before. We have used
longitudinal data, collected over six months, on 271 families with young children, to examine this relationship. Multipleregression was used to identify independent predictors of the total Edinburgh Post-natal Depression Scale score from arange of socioeconomic, demographic, social support and child health related variables. Worry about debt was the
strongest independent socioeconomic predictor of the depression score at both initial and follow-up occasions. Toaccount for the possibility of reverse causation, i.e. depression causing worry about debt, alternative regression modelsare reported which show that owing money by itself predicts depression and earlier debt worries predicts depression six
months later. We were unable to show that earlier debt worries independently predicted subsequent depression scoresafter the initial depression score had been taken into account in the analysis. Although debt has not been shown to bean independent prospective predictor of depression, our results suggest it has a central place in the association between
socioeconomic hardship and maternal depression. Evidence from qualitative studies on poverty and from studies on thecauses of depression support this hypothesis. The implications for policy are that strategies to enable families to controldebt should be an explicit part of wider antipoverty measures which are designed to reduce depression andpsychological distress among mothers of young children. # 2001 Elsevier Science Ltd. All rights reserved.
Keywords: Maternal depression; Poverty; Debt; Social disadvantage; Health inequalities; UK
Introduction
Depression is an important cause of morbidity inwomen and is estimated to affect between 10 and 30percent of mothers of young children (Cox, Connor, &
Kendall, 1982; Cox, Murray, & Chapman, 1993; Kumar& Robson, 1984; Cooper, Campbell, Day, Kennerley, &Bond, 1988). This causes suffering among women,
affects relationships within their families, and theirchildren’s developmental progress is impaired in thelonger term (Murray, 1992; Sharp et al., 1995; Murray &
Cooper, 1997). Maternal depression is not purely apostnatal problem; although there are some specificcharacteristics of depression in the months after giving
birth (Murray, Cox, Chapman, & Jones, 1995; Cooper
& Murray, 1995), the similarities in incidence, preva-lence, clinical features and associated factors suggeststhere is little to distinguish depression among mothers ofyoung children regardless of their age (Cooper et al.,
1988; Cox et al., 1993; Murray et al, 1995). Despiteconventional biomedical treatments being effective(Appleby, Warner, Whitton, & Faragher, 1997), there
is little evidence for a purely biological mechanism ofmaternal depression. Instead, combinations of variousfactors have been suggested as precipitating depression
in women who are already psychologically vulnerable.Broadly, these fall into three categories: those indicatingpoor quality or unsupportive relationships, those related
to the pregnancy, life events and acute stressors, andthose associated with socioeconomic disadvantage andfinancial hardship. This study examines the associationbetween depression and a specific aspect of socio-
economic adversity, namely debt.
*Corresponding author. Tel.: (0)1603-287624; fax: (0)1603-
287584.
E-mail address: [email protected] (R. Reading).1On behalf of the CAB and Family Health Study Team.
0277-9536/01/$ - see front matter # 2001 Elsevier Science Ltd. All rights reserved.
PII: S 0 2 7 7 - 9 5 3 6 ( 0 0 ) 0 0 3 4 7 - 6
Causal explanations of maternal depression
The first broad category of precipitating factorsconcern the quality of interpersonal relationships. Manystudies have shown that marital problems, poor quality
couple relationships or absence of a confidante arestrongly associated with depression (Brown & Harris,1978; Kumar & Robson, 1984; Brown, Andrews, Harris,Adler, & Bridge, 1986; Stein, Cooper, Campbell, Day, &
Altham, 1989; Murray et al., 1995; Escriba, Mas,Romito, & Saurel-Cubizolles, 1999; Hope, Power, &Rodgers, 1999; Romito, Saurel-Cubizolles, & Lelong,
1999; Stretch, Nicol, Davison, & Fundudis, 1999).Brown et al. (1986) and Brown and Bifulco (1990) haveshown that lack of support during a crisis, or being ‘‘let
down’’ by the woman’s partner predicts the subsequentonset of depression. Family disruption (Sheppard, 1997)and relationship difficulties with the extended family
also predispose to depression, particularly if theseinclude problems with the woman’s own mother (Kumar& Robson, 1984; Murray et al., 1995; Lambrenos,Weindling, Calam, & Cox, 1996). Lack of social support
is also a risk factor, while good quality social support isa protective factor (Brown et al., 1986). It would seemfrom these studies that the quality of interpersonal
relationships, particularly between the woman and herpartner, is an essential aspect of the causal pathway tomaternal depression.
A second category of precipitating factors are relatedto the pregnancy, life events and psychosocial stresses.With respect to the pregnancy, increased risk ofdepression has been found in association with un-
planned pregnancy (Kumar & Robson, 1984; Warner,Appleby, Whitton, & Faragher, 1996), preterm birth(Kumar & Robson, 1984), stillbirth (Kumar & Robson,
1984; Thorpe, Golding, MacGillivray, & Greenwood,1991) and not breastfeeding (Warner et al., 1996). Riskfactors related to the stress of managing a family include
family size (Murray et al., 1995; Sheppard, 1997), twinsor multiple births (Thorpe et al., 1991), close spacing ofbirths (Thorpe et al., 1991), ill health of the child
(Romito et al., 1999; Escriba et al., 1999), andbehavioural difficulties or developmental delay in thechild (Sheppard, 1997; Stretch et al., 1999). In thiscontext, it is interesting that neither childhood disability
nor the prospect of the child developing a disabilityappear to increase the risk of depression (Lambrenos etal., 1996). Studies on external life stresses have focussed
largely on employment, and complex relationships havebeen found. While many studies have shown thatmaternal or paternal unemployment is likely to increase
the risk of maternal depression (Stein et al., 1989;Lambrenos et al., 1996; Warner et al, 1996; Saurel-Cubizolles, Romito, Ancel, & Lelong, 2000), full time
employment has also been shown to be a risk factor,particularly among lone parents (Brown & Bifulco,
1990; Macran, Clarke, & Joshi, 1996; Baker & North,1999). The explanation appears to be that unemploy-
ment is an indicator of poverty, while full time employ-ment indicates the stress of extra workload which isadded to women’s responsibilities for care. This is
supported by Murray et al. (1995), who showed thatoccupational dissatisfaction increased the risk of depres-sion. Conversely, part time work may be a protectivefactor because of the social support it offers (Brown &
Bifulco, 1990).The third type of precipitating factor is socioeconomic
adversity. This was identified in the pioneering study by
Brown and Harris (1978). Subsequent studies haveshown associations with a range of socio-economicfactors such as low income, financial problems and
money worries, receipt of benefits, maternal andpaternal unemployment, housing tenure, and manualsocial class (Stein et al., 1989; Thorpe et al., 1991;
Murray et al., 1995; Warner et al., 1996; Brown &Moran, 1997; Sheppard, 1997; Graham & Blackburn,1998; Escriba et al., 1999; Romito et al., 1999). Theinterpretation of most of these studies is that socio-
economic adversity, however it is measured, is notsimply an additional contributory factor but has aspecific and pervasive influence.
The body of research on depression among lonemothers has been important in unravelling the effects ofeconomic hardship from other possible causes. A high
proportion of families headed by a lone parent live inpoverty (Judge & Benzeval, 1993; Oppenheim & Harker,1996). Women heading a lone parent family have amuch greater risk of depression than mothers with a
partner (Macran et al., 1996; Sheppard, 1997; Benzeval,1998). There are a number of possible explanationsapart from the direct effect of poverty; greater stress, less
social support, the effect of unemployment, and socialselection, but most of the studies set up to unravel thecauses have concluded that financial hardship is the
most important underlying feature (Macran et al., 1996;Benzeval, 1998; Baker & North, 1999; Hope et al, 1999).Although social support and psychological stress
undoubtedly have an effect, they are simply part of theweb of disadvantage endured by women living inpoverty.Two carefully controlled longitudinal studies enable
this web of disadvantage to be teased out. The first, alarge population based study among residents of NewHaven, Connecticutt (Bruce, Takeuchi, & Leaf, 1991),
examined the role of poverty in determining risk ofvarious psychiatric conditions, carefully controlling forother factors such as sex, age, race and history of
psychiatric illness. Previously well adults living inpoverty had over twice the risk of depression than thosenot living in poverty, with the population attributable
risk being 10% (i.e. 10% of new cases of depression weredirectly attributable to the effect of poverty). This study
R. Reading, S. Reynolds / Social Science & Medicine 53 (2001) 441–453442
therefore demonstrates that poverty has a direct causalinfluence on development of new episodes of depression.
The second was a detailed multifactorial study of thedeterminants of depression among the original Islingtonsample of inner city mothers (Brown & Moran, 1997).
Their model of a causal pathway began with pre-existingvulnerability factors such as childhood adversity andlack of self-esteem. Poor social support and lack of asupportive partner contributed more proximally to the
underlying vulnerability. Onset of depression wastriggered by severe life events within the family,particularly those involving humiliation or entrapment.
However, over and above all these, the most importantfactor was financial hardship. This was both thestrongest independent predictor of depression and it
also adversely affected all the other predisposing andrisk factors.
Socioeconomic disadvantage and depression
One potential criticism of this conceptual frameworkof the causes of depression is that socioeconomic
hardship is simply another cause of stress and does notneed to be distinguished on its own. However, there arethree reasons for considering it separately. First,
measures of socioeconomic disadvantage consistentlypredict depression, whereas other aspects of stress showinconsistent and distinctly different associations between
different studies. Several of the studies quoted above,particularly those with more robust methods, show theoverarching influence of poverty and socioeconomic
adversity, and how this can influence other aspects ofstress and social relationships (e.g. Brown & Moran,1997). Also, from a theoretical point of view poverty is amore chronic and insidious cause of stress than the other
factors which are either acute, contingent on thepregnancy or arise as a result of chance or unexpectedly(for instance a sick child, twin pregnancy or a severe life
event).The second reason for considering poverty as a
specific risk factor is that it signifies depression among
mothers as a feature of the social inequalities in healthfound in all modern developed countries (Whitehead &Diderichsen, 1997; Acheson et al., 1998). Given the
adverse effects on family relationships and childbehaviour and development, this in turn provides apossible mechanism for the social inequalities in childmental health and well being.
The third reason is that it has important policyimplications. Maternal depression is not just a problemfor primary health care and mental health services, it
requires a social policy response. To an extent this hasbeen recognised in the government’s drive to end childpoverty but attempts to tackle depression and other
aspects of maternal health have concentrated on‘problem’ groups, for instance teenage mothers, lone
parents, and those who have not returned to work.A variety of approaches are being tried by the
current UK government, targeted on these vulnerablegroups. However, if financial hardship is the primaryproblem, then many of these policies may be missing
the mark. It is more appropriate to address the problemof poverty than to implement policies aimed atreducing the numbers of high risk women, forexample lone mothers. Furthermore, interventions
targeted on specific groups may miss large numbers ofother women who suffer the same underlying socialdisadvantages and who therefore run a similarly high
risk of depression.
Debt and depression
Families may move into and out of poverty over time,particularly after the birth of an infant when changes
may occur in income, housing circumstances, demandson material and financial resources, employment andbenefit eligibility. One of the more ubiquitous features ofsocial disadvantage among families with young children
in recent years has been the presence and effect of debt(Oppenheim & Harker, 1996; Kempson, 1996). Low-income families headed by young people, particularly
those with larger numbers of children, run a high risk ofdebt (Berthoud & Kempson, 1992).The relationship between debt and maternal depres-
sion has not been studied in a quantitative way. Giventhat debt seems such an overarching feature of lifeamong families living in or at the margins of poverty,
this is an issue which deserves examination. Qualitativestudies of the experience of poverty provide richevidence that anxiety about debt may trigger orpotentiate depression among mothers of young children
(Kempson, Bryson, & Rowlingson, 1994; Kempson,1996), particularly in circumstances in which theireconomic resources may have changed.
As part of a study on the effects of Citizen’s AdviceBureau services on the health of mothers and youngchildren, we have data which provides an opportunity to
examine this hypothesis. We collected extensive socialand demographic information on families with childrenunder one year of age, including two Edinburgh
Postnatal Depression Scale (EPDS) questionnairesabout six months apart, data indicating access to socialnetworks, and measures of child and family stressors. Interms of the model of the causes of depression discussed
above, we are able to examine in detail the associationsbetween socioeconomic disadvantage and depression.We have used regression modelling to investigate the
following research questions. Firstly, to confirm thatmaternal depression is associated with socioeconomicdisadvantage among our sample. Secondly, to identify
which aspects of socioeconomic disadvantage have thestrongest relationship with depression, with a particular
R. Reading, S. Reynolds / Social Science & Medicine 53 (2001) 441–453 443
emphasis on the part played by debt and worries aboutdebt. Thirdly to investigate whether debt and worries
about debt prospectively influence the onset or worsen-ing of depression over time.
Methods
Setting and participants
We recruited families over a nine month period fromsix urban general practices in Norwich, UK during the
period July 1997 to February 1998. These practicesserved a mixed population which tended to be moder-ately deprived. Families were invited to participate in the
project if they had an infant under one year of age at thebeginning of the study period. Participants completedtwo postal questionnaires on demographic, social and
health information about the family; one at recruitment(Time 1) and another at the end of the study (Time 2).These were on average around six months apart butvaried between 40 days and 14 months. Information
about the study was delivered to 919 eligible families.Two hundred and sixty-one families responded at Time1, and 219 responded at Time 2. Overall, 271 families
responded at either Time 1 or Time 2, while 209 familiesresponded at both Time 1 and Time 2.
Measures
Demographic data
These variables measure certain aspects of stressassociated with family care. They included family sizeand structure, age of the index child and of themother, and whether the family was headed by a lone
parent.
Socioeconomic data
These variables measure socioeconomic stress. Theyincluded family income, whether the family were inreceipt of benefits (other than child benefit), employment
status of adult members of the household, car owner-ship, housing tenure (only in the second questionnaire),and overcrowding (i.e. more than one person per room).In addition, we asked about money worries on a scale of
one to four (a question devised by Graham & Black-burn, 1998), whether the family owed any money, and ifso how worried were they by this debt (on a scale of 0–
6). The question on owing money was changed slightlybetween the two questionnaires as a result of someambiguity in the first one. Those without any debts were
automatically coded as having no worries about debt.
Social support and networks
These variables are a measure of social support andaccess to social networks although not of the quality of
close personal relationships. They include whether a liftwas available in an emergency, whether the family had
had someone babysit for free in the last month andwhether they were part of a babysitting arrangementwith any other families..
Child healthThese variables are a measure of stresses associated
with the pregnancy, and care of the child. They
included whether the child had been breastfed at birth,whether the child had attended hospital (again thewording of this question changed slightly between
questionnaires, the first questionnaire asked about anyhospital attendance in the previous two months, thesecond asked about admissions in the previous six
months), attendance at the general practice in theprevious two weeks, child sleeping problems, and avariety of questions adapted from the Warwick Child
Health and Morbidity Profile (Spencer & Coe, 1996) ongeneral health, minor illnesses, behaviour problems, andaccidents.. These latter questions were only asked in thesecond questionnaire.
Maternal depressionThe main outcome measure was the score on
the Edinburgh Postnatal Depression Scale (EPDS)(Cox, Holden, & Sagovsky, 1987). This was collectedat Time 1 and Time 2. The EPDS is a well-established,
10-item measure of depression in women, developedfor use in community settings. It has good reliabilityand validity (Cox et al., 1987; Cox, Chapman, Murray,
& Jones, 1996; Murray & Carothers, 1990). The EPDScan be scored as a categorical or continuous variable.Values of 13 or greater indicate a high risk of asignificant depressive disorder. However, in this
study the overall numerical score was used rather thana cutoff value, in order to retain as much usefulstatistical information and because there appears to
be no clear distinction between a ‘‘normal’’ and a‘‘depressed’’ score (Murray & Carothers, 1990; Coxet al., 1996).
Statistical methodsLinear regression modelling by ordinary least-squares
regression was used from the statistical package SPSSfor Windows version 8. First, unadjusted relationshipsbetween the possible explanatory variables and theEPDS score at Time 1 and Time 2 were calculated using
simple linear regression.A correlation matrix between explanatory variables
was calculated. Correlation coefficients greater than 0.5
in magnitude were taken as evidence of colinearity.Multiple regression models were built, entering onlythose variables with a significant unadjusted relationship
with the EPDS score. Variables were added andremoved from the models in a stepwise fashion using
R. Reading, S. Reynolds / Social Science & Medicine 53 (2001) 441–453444
inclusion criteria of p50:05 and rejection criteria ofp > 0:1. Different models were built in order to examine
various assumptions. Details of the rationale for thesedifferent models are given in the results section.Numbers in the different models vary slightly because
of cases with missing data which we excluded from thoseanalyses.The study received ethical approval from the local
research ethics committee.
Results
The families were from a socially diverse populationand Table 1 shows socioeconomic, demographic andhealth related data from the sample. All of the lone
parents were mothers. At the return of the firstquestionnaire 60/243 (25%) of women had an EPDSscore of 13 or higher indicating a high likelihood ofsignificant depressive illness; at the second questionnaire
Table 1
Distributions of the socioeconomic, demographic, social network and child health variables used in the analyses at the two data
collection pointsa
Time 1 Time 2
Socioeconomic variables
Mean family income per week £286
(25%, 50%, 75% values) (133, 250, 370)
Families owing money 69% 67%
Families on benefits 38%
Housing tenure (Families not owning own home) 43%
Mother unemployed 57%
Father unemployed 10%
Overcrowding (more than one person per room) 22%
No access to car 25%
Debt worries=0 (no worries) 48% 44%
Debt worries=1 9% 10%
Debt worries=2 11% 11%
Debt worries=3 10% 13%
Debt worries=4 12% 11%
Debt worries=5 4% 7%
Debt worries=6 (extremely worried) 7% 5%
Demographic variables
Lone parent families 19%
Mean maternal age at Time 1 29.1
(25%, 50%, 75% values) (25, 29, 33)
Mean age of child in months 6.7 12.5
(25%, 50%, 75% values) (2.6, 5.6, 10.8) (8.0,12.0,17.6)
Families with single child 51%
Families with two children 32%
Families with more than two children 18%
Social network variables
Access to a lift if needed 70%
Access to babysitting network 33%
Free babysitting in past month 58%
Child health variables
Child breast fed at birth 72%
Childs general health very good 49%
Child had more than two illnesses in past month 19%
Child attended hospital in past two months at Time 1 15%
Child been admitted to hospital in past six months at Time 2 9%
Child attended GP in past two weeks 34% 23%
Child has problems sleeping 12% 12%
Mean EPDS score (25%, 50%, 75% values) 8.4 (4,7,12) 7.4 (4,7,11)
aNote. Total number of cases at Time 1 varies between 251 and 271 because of missing data, and at Time 2 numbers vary between
208 and 219.
R. Reading, S. Reynolds / Social Science & Medicine 53 (2001) 441–453 445
this had dropped to 31/208 (15%). There was a smallreduction in scores between the first and second
questionnaires among women who had returned both(change in EPDS=�0.5, SD=4.0), but the main reasonfor the apparent reduction in prevalence was a lower
return rate of the second questionnaire among womenwho scored high values on the first (65% of womenscoring 13 or higher on the first EPDS returned thesecond questionnaire compared with 84% of women
scoring less than 13, w2 ¼ 10:3, p ¼ 0:001).A number of socioeconomic, demographic, social
support and health related factors were associated with
the EPDS score. Unadjusted associations with the EPDSscore at Time 1 are shown in Table 2 and at Time 2 inTable 3. There was restricted data on the child’s health
at Time 1 so only those variables which were availablewere included in analyses of the EPDS at Time 1.At Time 1 there were significant associations between
the EPDS scores and the socioeconomic factors offamily income, receipt of benefits, housing tenure, accessto a car and worries about debt, the demographic factorof lone parenthood, and the social support related factor
of availability of a lift. The direction of these associa-
tions was as expected with social disadvantage, loneparenthood, and lack of access to a lift all being
associated with higher scores.At Time 2 a similar pattern of associations was seen.
Socioeconomic factors that were significantly associated
with EPDS score were family income, owing money,receipt of benefits, housing tenure, access to a car,worries about debt, and worries about money. Thesignificant demographic factors were lone parenthood
and the number of children in the family. The socialsupport factor associated with the EPDS score wasagain availability of a lift. A number of health
related factors were associated at this time includingwhether the child had attended hospital in the twomonths prior to the first questionnaire, whether the child
had problems sleeping at Time 2 and the initial EPDSscore.Many of the socioeconomic variables were correlated
with each other, although rarely strongly enough to beconcerned about colinearity. Pearson correlation coeffi-cients greater than 0.5 in magnitude were only foundbetween housing tenure and income (r ¼ �0:51), loneparenthood and receipt of benefits (r ¼ 0:55), car
Table 2
Unadjusted simple linear regression equations between possible explanatory variables and Edinburgh Post-natal Depression Scores at
Time 1
Explanatory variable Constant Regression
coefficient
SE of
coefficient
Sig. (p) R2
Socioeconomic variables
Family income (£ per week) 9.6 �0.004 0.002 0.005a 3.3%
Family owes money at Time 1 8.0 0.7 0.75 0.352 0.4%
Family on benefits 7.3 3.0 0.69 50.001a 7.3%
Tenure (non home owners) 7.2 1.7 0.72 0.018a 3.0%
Mother unemployed 7.7 1.2 0.69 0.075 1.3%
Father unemployed 8.0 �1.7 1.15 0.145 1.1%
Overcrowded household 8.2 1.2 0.83 0.167 0.8%
No access to car 7.8 2.6 0.80 0.001a 4.2%
Debt worries at Time 1 (0–6) 7.0 0.9 0.17 50.001a 10.8%
Demographic variables
Lone parent family 7.8 3.7 0.86 50.001a 7.1%
Maternal age (years) 9.9 �0.06 0.07 0.405 0.3%
Age of child (days) 7.8 0.003 0.002 0.217 0.7%
Number of children (1,2,3+) 8.5 �0.03 0.40 0.934 50.1%
Child’s gender (male=1) 8.5 0.004 0.69 0.995 50.1%
Social network variables
Access to lift 9.7 �1.9 0.74 0.011a 2.7%
Access to babysitting network 8.6 �0.6 0.72 0.445 0.2%
Free babysitting in past month 8.6 �0.3 0.69 0.694 0.1%
Child health variables
Breast fed at birth 8.9 �0.7 0.76 0.371 0.3%
Attended hospital in past 2 mo 8.2 1.7 0.97 0.085 1.2%
Attended GP in past 2 weeks 8.4 �0.03 0.72 0.962 50.1%
Problems sleeping 8.3 0.9 1.07 0.382 0.3%
ap50:05. These variables are included in the later multiple regression analyses.
R. Reading, S. Reynolds / Social Science & Medicine 53 (2001) 441–453446
ownership and receipt of benefits (r ¼ �0:51), housingtenure and receipt of benefits (r ¼ 0:59), and between theworries about debt and money worries (r ¼ 0:61 at Time1 and 0.69 at Time 2).
Table 4 shows the results of the multiple regressionanalysis on the EPDS score at Time 1. All the variables
found to have a significant unadjusted association withthe EPDS score were entered as independent variables.
Table 3
Unadjusted simple linear regression equations between possible explanatory variables and Edinburgh Post-natal Depression Scores at
Time 2
Explanatory
variable
Constant Regression
Coefficient
SE of
coefficient
Sig. (p) R2
Socioeconomic variables
Family income (£ per week) 8.5 �0.004 0.002 0.033a 2.2%
Family owed money at Time 1 6.3 1.5 0.75 0.051 1.9%
Family owes money at Time 2 5.8 2.4 0.72 0.001a 5.1%
Family on benefits 6.6 2.4 0.70 0.001a 5.2%
Tenure (non home owners) 6.6 1.8 0.68 0.009a 3.3%
Mother unemployed 7.0 0.7 0.68 0.302 0.5%
Father unemployed 6.9 1.0 1.27 0.437 0.4%
Overcrowded household 7.1 1.2 0.91 0.186 0.4%
No access to car 7.0 1.8 0.83 0.031a 2.2%
Debt worries at Time 1 (0–6) 6.1 0.8 0.17 50.001a 9.6%
Debt worries at Time 2 (0–6) 5.8 0.9 0.17 50.001a 13%
Money worries at Time 2 (1–4) 2.0 2.3 0.40 50.001a 13.7%
Demographic variables
Lone parent family 7.0 2.0 0.89 0.024a 2.4%
Maternal age (years) 9.5 �0.07 0.064 0.256 0.6%
Age of child (days) 6.0 0.004 0.002 0.064 1.6%
Number of children (1,2,3+) 5.2 1.2 0.42 0.003a 4.0%
Child’s gender (male=1) 7.2 0.4 0.69 0.531 0.2%
Social network variables
Access to lift 9.1 �2.3 0.76 0.002a 4.3%
Access to babysitting network 7.4 0.01 0.71 0.989 50.1%
Free babysitting in past month 6.7 1.0 0.70 0.135 1.1%
Child health variables
EPDS at Time 1 2.0 0.67 0.06 50.001a 45.2%
Breast fed at birth 8.0 �0.8 0.78 0.301 0.5%
Attended hospital in 2 mo before Time 1 7.0 2.4 0.97 0.016a 2.9%
Admitted to hospital in 6 mo before Time 2 7.3 1.1 1.16 0.353 0.4%
Attended GP in 2 weeks before Time 1 7.1 0.8 0.74 0.281 0.6%
Attended GP in 2 weeks before Time 2 7.5 �0.5 0.81 0.528 0.2%
Problems sleeping at Time 1 7.2 0.8 1.01 0.445 0.3%
Problems sleeping at Time 2 7.1 2.7 1.01 0.009a 3.3%
Childs general health very good at Time 2 7.7 �0.8 0.68 0.254 0.6%
More than 2 childhood illnesses in past 3 months at Time 2 7.1 1.5 0.87 0.091 1.4%
ap50:05. These variables are included in the later multiple regression analyses
Table 4
Multiple regression models explaining Edinburgh Post-natal Depression Score at Time 1a
Model R2 adjusted Variables included Regression
coefficient
SE Significance (p)
1 16.5% Constant 6.39 0.43 50.001
(n ¼ 231) Debt worries at Time 1 (0–6) 0.86 0.17 50.001
Lone parent 3.28 0.83 50.001
aVariables entered: Income, receipt of benefits, car owner, debt worries at Time 1, lone parent, access to lift.
R. Reading, S. Reynolds / Social Science & Medicine 53 (2001) 441–453 447
Debt worries and being a lone parent were thetwo significant variables, together accounting for
16.5% of the variance in the EPDS score. Debt worriesranged from values of 0–6, and so the relative effect ofbeing most worried about debt was an increase in the
EPDS score of around 5 points. This model did notinclude the variable of whether the family owed money,because it was not associated at the unadjusted level(Table 2). The close correlation between lone parent-
hood and being in receipt of benefits meant that thesetwo variables were virtually interchangeable in theregression equation.
We then examined predictors of the EPDS score at thetime of the second questionnaire (Time 2). Model 1 inTable 5 shows a multiple regression analysis in which all
the Time 1 and Time 2 variables which were individuallyrelated to the Time 2 EPDS score were entered with theexception of the Time 1 EPDS score and Time 1 debt
worries. Model one accounts for 19.1% of the varianceand worries about debt was again the strongest predictorwith neither receipt of benefits nor lone parenthoodcontributing. Two health variables, sleeping problems in
the child and the child attending hospital before the firstquestionnaire, and one social support variable, access to
a lift, were also independently associated with the EPDSscore.Both of these analyses are predominantly cross-
sectional, so the strong contribution of debt worriescould be an artefact of reverse causation, i.e. depressedwomen are worrying more about their debt. Therefore,in model 2 we removed debt worries from the
independent variables. Owing money and the receipt ofbenefits became significant independent predictors. Thechild having attended hospital dropped out of the
equation. This model explained 13% of the variance.Thus when we remove ambiguities around the questionof worry, the state of being in debt and living in poverty
(i.e. being in receipt of benefits) are still associated withdepression. Reverse causation could still apply butwould imply that depression resulted in women getting
into debt rather than just worrying about it.We then introduced a longitudinal perspective. In
model 3 we repeated model one but replaced currentdebt worries with debt worries at Time 1. This model
Table 5
Multiple regression models explaining Edinburgh Post-natal Depression Score at Time 2a
Model R2
adjusted
Variables included Regression
coefficient
SE Significance
(p)
1 19.1% Constant 6.49 0.73 50.001
(n ¼ 188) Debt worries at Time 2 (0–6) 0.88 0.17 50.001
Child sleeping problem at Time 2 2.65 1.05 0.013
Access to lift �1.55 0.74 0.036
Child attended hospital in 2 months prior to Time 1 1.87 0.90 0.039
2 12.9% Constant 6.49 0.84 50.001
(n ¼ 189) Receipt of benefits 1.58 0.75 0.036
Owing money at Time 2 2.13 0.70 0.003
Child sleeping problem at Time 2 2.71 1.07 0.013
Access to lift �1.82 0.77 0.018
3 15.4% Constant 7.28 0.73 50.001
(n ¼ 186) Debt worries at Time 1 (0–6) 0.73 0.18 50.001
Child sleeping problem at Time 2 2.71 1.06 0.011
Access to lift �1.83 0.76 0.017
4 49.4% Constant 1.49 0.52 0.005
(n ¼ 174) EPDS at Time 1 0.62 0.06 50.001
Debt worries at Time 2 (0-6) 0.45 0.15 0.003
Child attended hospital in 2 months prior to Time 1 1.62 0.76 0.035
aNote. 1 Variables included: Income, owing money at Time 2, receipt of benefits, tenure, car owner, debt worries at Time 2, lone
parent, number of children, access to lift, child attended hospital in two months prior to Time 1, childs sleeping problem at
Time 2.
2 Variables included: Income, owing money at Time 2, receipt of benefits, tenure, car owner, lone parent, number of children,
access to lift, child attended hospital in two months prior to Time 1, childs sleeping problem at Time 2.
3 Variables included: Income, owing money at Time 2, receipt of benefits, tenure, car owner, Debt worries at Time 1, lone
parent, number of children, access to lift, child attended hospital in two months prior to Time 1, childs sleeping problem at
Time 2.
4 Variables included: Income, owing money at Time 2, receipt of benefits, tenure, car owner, debt worries at Time 2, lone
parent, number of children, access to lift, EPDS at Time 1, child attended hospital in two months prior to Time 1, childs
sleeping problem at Time 2.
R. Reading, S. Reynolds / Social Science & Medicine 53 (2001) 441–453448
shows that earlier debt worries were a strong predictorof later depression, explaining 15% of the variance.
Finally, in model 4 we added the EPDS score at Time1 to the other variables in model one. Previous depressedmood is known to predict subsequent depression,
particularly over such a relatively short timescale as thisstudy. As expected, the earlier EPDS score explains amajor part of the variance of the later EPDS score.There continued to be a strongly significant contribution
to the model from worries about debt at Time 2, with abarely significant contribution from the child havingattended hospital.
Despite demonstrating the close associations betweendepression and various measures related to debt atdifferent times, we were unable to confirm an indepen-
dent longitudinal association. This would requireworries about debt at Time 1 to predict the EPDS atTime 2 controlling for the effect of the EPDS at Time 1.
When this was done the EPDS at Time 1 swallowed upthe explanatory effect of debt at Time 1.
Discussion
We have confirmed a close relationship betweenfinancial hardship and depressed mood in mothers of
infants. Worries about debt appeared to be the strongestpredictor of this relationship but being in debt was alsosignificantly associated with the depression score along
with being in receipt of benefits once worries about debthad been removed from the equation. Worry about debtat the initial assessment predicted a high depression
score subsequently, but we were unable to confirm acausal relationship because when we controlled for theinitial depression score no further variance in the laterdepression score was explained by worries about debt at
the first assessment.The results beg more questions than they answer.
They do not support a simple cause and effect relation-
ship, yet they imply that debt is an important aspect ofthe link between financial adversity and depression.Rather than a linear pathway there may be a circular
one; worries about debt contribute to making depressionworse, while depression causes women to worry moreabout their debt and deal with them less effectively.Regardless of the direction of the effect, our results
demonstrate that being in debt to the extent that itcauses worry is intimately related to maternal depres-sion. This has implications for our understanding of
depression in women, of inequalities in health amongwomen and children, and of policies to address these.
Strengths and weaknesses of the studyThe strengths of our study include the detailed
socioeconomic data collected and the longitudinal
nature of the study. We have constructed a series ofmodels to explain the depression score, selectively
omitting or adding variables in an attempt to show thestrength and direction of the effects of different
variables. The comprehensive range of socioeconomicvariables reduces the possibility that those measuringdebt are being confounded by other aspects of social
disadvantage. This is the first study we are aware ofwhich examines the relationship between debt andmaternal depression from a quantitative viewpoint.The data were collected recently and therefore provide
a contemporary picture of the circumstances faced byfamilies of young children in the UK. The populationstudied was from six unremarkable urban general
practices, and the families were from a wide range ofsocial backgrounds.We believe the results are generalisable to the
wider UK population. We have compared our studypopulation with representative data from the 1991 UKcensus. Comparisons with national data are subject
to inaccuracy because definitions may vary, for instanceour definition of the number of rooms in a housevaried somewhat from the census measure, and becauseextracting data for families with young children from
the standard census tables is not always possible.Nevertheless, there are close similarities. For instance,25% of families with dependent children aged 0–4 years
in Great Britain do not own or have access to acar (OPCS, 1993), exactly the same percentage as inour sample. Eighteen percent of children aged 0–4 years
in Great Britain live in a lone parent family (OPCS,1994), which may be compared with 19% of familiesin our sample being headed by a lone parent. Elevenpercent of children aged 0–4 in Great Britain live in
households with more than one person per room(OPCS, 1994), compared with 22% of families in oursample. Finally, 38% of families in our sample were
in receipt of benefits which compares closely withrecent estimates of the proportion of households withchildren living in conditions of poverty (Oppenheim &
Harker, 1996).There are a number of potential weaknesses to the
study. The sample was relatively small which limited our
ability to demonstrate complex multivariate relation-ships. The sample may have been biased because lessthan a third of those approached by letter responded.We have compared the responders with all women
eligible to be included on a limited range of measuresavailable from the computerised child health informa-tion system. These were age, number of children, lone
parent status, and birthweight and gestation of the indexchild. No significant differences were found, indeed thesample of responders showed remarkably similar char-
acteristics to the group as a whole. Therefore, despite thelimitations of response rate, the sample was comparableto both national representative data and to the practice
populations. Of more relevance to the validity of theresults was the differential loss to follow up of the more
R. Reading, S. Reynolds / Social Science & Medicine 53 (2001) 441–453 449
depressed women, and of the more deprived women.Fewer lone parents responded to the second question-
naire which may explain why lone parenthood was notfound to be an explanatory factor at this time, whereas itwas at the first questionnaire, and it is consistently found
in other studies.Other weaknesses exist because the primary purpose
of the study was to pilot a trial of Citizen’s AdviceBureau services rather than to investigate the social
causes of maternal depression. Thus, the question onwhether any money was owed was slightly different inthe second questionnaire, the data on previous health
and pregnancy history was limited, social support wasnot well covered and the number of study participantswas less than optimum for such a study. We have
already commented that the question about debt worriesis likely to be confounded by affective state and is notnecessarily an objective measure of financial impact. On
the other hand, this question may represent aspects ofthe experience of debt and financial hardship whichwould be missed by simply using a measure of theamount of money owed.
Debt and depression, how are they related?There are a small number of other research studies on
the association between debt and depression. In onestudy of males attempting suicide (Hatcher, 1994)severity of debt was related to both the severity of the
suicide attempt and to the severity of symptoms ofdepression and hopelessness. Graham and Blackburn(1998) have reported that among mothers receiving
income support, depressed mood was closely associatedwith both the length of time they had been in receipt ofbenefits and the degree of money worries they had.Graham and Blackburn’s question on money worries
was very similar to ours on debt worries, so much so thatwe included it in our second questionnaire and showed avery high correlation between the answers (r ¼ 0:69,p50:001). Debt or money worries and being in receiptof benefits account for depression similarly in bothGraham and Blackburn’s and our studies.
A recent study from Ohio, USA, measured theassociation between credit card debt and self reportedhealth (Drentea & Lavrakas, 2000). They found that
both the amount of debt } measured by the debt/income ratio } and debt stress } measured in a wayvery similar to our question about debt worries } werestrongly associated with self-reported general health
after controlling for a wide range of demographic andsocioeconomic variables. This study was also of a cross-sectional design and the authors make the same caveats
about interpreting the direction of the effects as we do.Nevertheless, the consistency of our findings with theirsis notable and highlights the need to consider debt as an
important socioeconomic variable in studies investigat-ing health inequalities. As Drentea and Lavrakis state
‘‘Understanding debt is one way of understandinghealth inequalities . . .. [because] debt may help us to
understand financial strain and hardship compared tomerely measuring income’’ (p. 527).Results from qualitative studies of the experience of
poverty are relevant to this understanding. Kempson(1996) summarises a number of studies which describethe pervasive and debilitating effect of debt. Theseinclude fear of disconnection, homelessness, or harass-
ment and feelings of stigmatisation, shame, despair,depression and suicidal intent. The links between debtand the precipitation of depression seem fairly obvious
and plausible.An alternative interpretation of our results could be
that worry about debt is a general measure of social
disadvantage and the findings simply reflect a suscept-ibility to depression among poorer women. The questionon debt worries was the only one which enabled a
subjective answer to socioeconomic circumstances to begiven; perhaps this picks up aspects of deprivation whichare not reflected in the more categoric measures.Kempson et al. (1994) describe a similar spiral of stress,
depression, and oppression felt by families in financialhardship, regardless of whether they cope by jugglingever increasing debts or by keeping out of debt but
cutting back on essentials. Both ways of dealing withpoverty have similar costs in terms of their mental health‘‘Poor people face a Hobson’s choice between anxiety
about their budgets or anxiety about their debts’’ (p.286).Our results also need to be considered in the light
of Brown and Moran’s study on determinants of
depression in women living in poverty (1997). Depres-sion was found to occur in women living in conditionsof financial hardship who also experienced a severe
event of humiliation or entrapment. Examples ofhumiliating events included separations, put downsand delinquent behaviour in close family members.
Examples of entrapment consisted of ongoing severeevents such as being told a paralysed and bedriddenhusband would not improve. In other words, the events
were described as occurring in the context of closepersonal ties. We would argue that the experience ofdebt, particularly problem debts, is an event which is,by its very nature, both humiliating and entrapping
and it occurs in circumstances of financial hardshipi.e. it meets all the criteria for an aetiological factoridentified by Brown and Moran. Furthermore, our
question on worries about debt may have tapped intothis sense of humiliation or entrapment that debtengenders.
Implications for policy and further researchThe identification of debt as a specific health risk is
not an unpredictable finding. It supports the many UKgovernment initiatives which currently focus on relieving
R. Reading, S. Reynolds / Social Science & Medicine 53 (2001) 441–453450
child poverty as a key contribution to reducing healthinequalities (Department of Health, 1999). Reducing
child poverty by, for example, the working families taxcredit, the minimum wage and tax incentives forchildcare, should reduce the number of families getting
into debt, and enable others to manage their debts moreeffectively, yet few of these policies specifically addressthe issue of debt. One initial step could be to include ameasure of the burden of debt in the list of indicators of
poverty and social exclusion which are compiled by boththe government and independent policy analysts (De-partment of Social Security, 1999; Howarth, Kenway,
Palmer, & Miorelli, 1999). More actively, preventingfamilies getting into debt and helping families out ofdebt are feasible social policy objectives which would fit
comfortably within the range of other antipoverty andfamily support policies pursued by the government.Quite apart from the possible health consequences we
have highlighted, debt which cannot be redeemed orserviced benefits no-one in society apart from the creditindustry (Berthoud & Kempson, 1992). At a local level,health and social service managers and policy makers
could encourage links between professionals who havecontact with families, such as social workers and healthvisitors, and debt counselling and advice services in the
statutory and voluntary sectors. Health and localauthorities are being encouraged to develop localpolicies which address health inequalities, this type of
collaboration provides an opportunity for a practicalresponse.The second set of implications are for further
research. The results from this and related studies
(Graham & Blackburn, 1998; Drentea & Lavrakas,2000) have highlighted worries about debt and financesas an important factor which may help to explain the
pathway for social differences in health. None of ourstudies have confirmed this and there is a need forcareful prospective studies over a sufficiently long period
of time to examine the strength, direction and specificityof the association between debt and depression. Thereare also opportunities for studies of interventions to
address debt, as well as further aetiological studies. Forexample, our data came from a pilot study of the effectsof Citizen’s Advice Bureau services in primary care.There have been several previous studies of this type of
intervention but few have measured health outcomes(Paris & Player, 1993; Veitch, 1995; Abbott & Hobby,1999; National Association of Citizen’s Advice Bureaux,
1999). Our results provide some theoretical support forsuch studies, as we can postulate a pathway between anaspect of social and financial hardship (i.e. debt) and an
objective measurable health outcome (i.e. maternaldepression). There may also be scope for comparingthe effects on maternal mood of psychological interven-
tions against depression with social interventions direc-ted against debt and financial hardship.
Acknowledgements
The CAB and Family Health Study group consists ofthe two principal authors; John Appleby, healtheconomist; Mary Mclean, project coordinator; Janet
Watts and Sarah Kember, health visitor team leaders;Jackie Wheatley, CAB adviser; Marion Wright, assistantdirector of Norwich CAB; Sarah Steel, paediatrician.The study was funded by the Research and Develop-
ment directorate of the Eastern NHS executive. Thestudy depended on the active collaboration of all theparticipating families, the health visitors, practice
managers and general practitioners in the six practiceswe studied. We are grateful to Mildred Blaxter, KathyBranson, Michelle Bussutil, Bridget Coupland, Richard
Gilham, Sue Lindsey, SamMugford, Shirley Pearce, LeeShepstone, Nav Shinh, and Jill Tanner for help, adviceand support. The comments of three anonymous
reviewers have helped us greatly improve the paper.Copies of the questionnaires may be had from thecorresponding author.
References
Abbott, S., & Hobby, L. (1999). An evaluation of the health and
advice project: Its impact on the health of those using the
service. Report No. 99/63, Health and Community Care
Research Unit, University of Liverpool, Liverpool.
Acheson, D., Barker, D., Chambers, J., Graham, H., Marmot,
M., & Whitehead, M. (1998). Independent inquiry into
inequalities in health. Stationary Office, London.
Appleby, L., Warner, R., Whitton, A., & Faragher, B. (1997). A
controlled study of fluoxetine and cognitive-behavioural
counselling in the treatment of postnatal depression. British
Medical Journal, 314, 932–936.
Baker, D., North, K., & ALSPAC Study Team (1999). Does
employment improve the health of lone mothers? Social
Science and Medicine, 49, 121–131.
Benzeval, M. (1998). The self-reported health status of lone
parents. Social Science and Medicine, 46, 1337–1353.
Berthoud, R., & Kempson, E. (1992). Credit and debt: The PSI
report. Policy Studies Institute, London.
Brown, G. W., Andrews, B., Harris, T., Adler, Z., & Bridge, L.
(1986). Social support, self-esteem and depression. Psycho-
logical Medicine, 16, 813–831.
Brown, G. W., & Bifulco, A. (1990). Motherhood, employment
and the development of depression. A replication of a
finding?. British Journal of Psychiatry, 156, 169–179.
Brown, G. W., & Harris, T. (1978). Social origins of depression:
A study of psychiatric disorders in women. London: Tavi-
stock.
Brown, G. W., & Moran, P. M. (1997). Single mothers, poverty
and depression. Psychological Medicine, 27, 21–33.
Bruce, M. L., Takeuchi, D. T., & Leaf, P. J. (1991). Poverty and
psychiatric status. Longitudinal evidence from the New
Haven epidemiologic catchment area study. Archives of
General Psychiatry, 48, 470–474.
R. Reading, S. Reynolds / Social Science & Medicine 53 (2001) 441–453 451
Cooper, P. J., Campbell, E. A., Day, A., Kennerley, H., &
Bond, A. (1988). Non-psychotic psychiatric disorder after
childbirth: A prospective study of prevalence, incidence,
course and nature. British Journal of Psychiatry, 152,
799–806.
Cooper, P. J., & Murray, L. (1995). The course and recurrence
of postnatal depression: Evidence for the specificity of the
diagnostic concept. British Journal of Psychiatry, 166,
191–195.
Cox, J. L., Chapman, G., Murray, D., & Jones, P. (1996).
Validation of the Edinburgh postnatal depression scale
(EPDS) in non-postnatal women. Journal of Affective
Disorders, 39, 185–189.
Cox, J. L., Connor, Y., & Kendall, R. E. (1982). Prospective
study of the psychiatric disorders of childbirth. British
Journal of Psychiatry, 140, 111–117.
Cox, J. L., Holden, J. M., & Sagovsky, R. (1987). Detection of
postnatal depression. Development of the 10-item Edin-
burgh postnatal depression scale. British Journal of Psy-
chiatry, 150, 782–786.
Cox, J. L., Murray, D., & Chapman, G. (1993). A controlled
study of the onset, duration and prevalence of postnatal
depression. British Journal of Psychiatry, 163, 27–31.
Department of Health (1999). Reducing health inequalities: An
action report. Department of Health, London.
Department of Social Security (1999). Opportunities for all:
Tackling poverty and social exclusion. Department of
Health, London.
Drentea, P., & Lavrakas, P. J. (2000). Over the limit: The
association among health, race and debt. Social Science and
Medicine, 50, 517–529.
Escriba, V., Mas, R., Romito, P., & Saurel-Cubizolles, M. J.
(1999). Psychological distress of new Spanish mothers.
European Journal of Public Health, 9, 294–299.
Graham, H., & Blackburn, C. (1998). The socio-economic
patterning of health and smoking behaviour among mothers
with young children on income support. Sociology of Health
and Illness, 20, 215–240.
Hatcher, S. (1994). Debt and deliberate self-poisoning. British
Journal of Psychiatry, 164, 111–114.
Hope, S., Power, C., & Rodgers, B. (1999). Does financial
hardship account for elevated psychological distress
in lone mothers?. Social Science and Medicine, 49,
1637–1649.
Howarth, C., Kenway, P., Palmer, G., & Miorelli, R. (1999).
Monitoring poverty and social exclusion. York: Joseph
Rowntree Foundation.
Judge, K., & Benzeval, M. (1993). Health inequalities: new
concerns about the children of single mothers. British
Medical Journal, 306, 677–680.
Kempson, E. (1996). Life on a low income. York: Joseph
Rowntree Foundation.
Kempson, E., Bryson, A., & Rowlingson, K. (1994). Hard
times? How poor families make ends meet. Policy Studies
Institute, London.
Kumar, R., & Robson, K. M. (1984). A prospective study of
emotional disorders in childbearing women. British Journal
of Psychiatry, 144, 35–47.
Lambrenos, K., Weindling, A. M., Calam, R., & Cox, A. D.
(1996). The effect of a child’s disability on mother’s mental
health. Archives of Disease in Childhood, 74, 115–120.
Macran, S., Clarke, L., & Joshi, H. (1996). Women’s health:
dimensions and differentials. Social Science and Medicine,
42, 1203–1216.
Murray, D., Cox, J. L., Chapman, G., & Jones, P. (1995).
Childbirth: Life event or start of a long-term difficulty.
British Journal of Psychiatry, 166, 595–600.
Murray, L. (1992). The impact of postnatal depression on
infant development. Journal of Child Psychology and
Psychiatry, 33, 543–561.
Murray, L., & Carothers, A. D. (1990). The validation of the
Edinburgh post-natal depression scale on a community
sample. British Journal of Psychiatry, 157, 288–290.
Murray, L., & Cooper, P. J. (1997). Effects of postnatal
depression on infant development. Archives of Disease in
Childhood, 77, 99–101.
National Association of Citizens Advice Bureaux (1999).
Prescribing advice: Health and inequality. National Associa-
tion of Citizens Advice Bureaux, London.
Office of Population Censuses and Surveys, 1993 (1991).
Census: Housing and availability of cars. London: HMSO.
Office of Population Censuses and Surveys, 1994 (1991).
Census: Children and young adults. London: HMSO.
Oppenheim, C., & Harker, L. (1996). Poverty : The facts (3rd
ed). London: Child Poverty Action Group.
Paris, J. A. G., & Player, D. (1993). Citizens’ advice in general
practice. British Medical Journal, 306, 1518–1520.
Romito, P., Saurel-Cubizolles, M-J., & Lelong, N. (1999). What
makes new mothers unhappy: Psychological distress one
year after birth in Italy and France. Social Science and
Medicine, 49, 1651–1661.
Saurel-Cubizolles, M-J., Romito, P., Ancel, P-Y., & Lelong, N.
(2000). Unemployment and psychological distress one year
after childbirth in France. Journal of Epidemiology and
Community Health, 54, 185–191.
Sharp, D., Hay, D. F., Pawlby, S., Schmucker, G., Allen, H., &
Kumar, R. (1995). The impact of postnatal depression on
boys intellectual development. Journal of Child Psychology
and Psychiatry, 36, 1315–1336.
Sheppard, M. (1997). Depression in female health visitor
consulters: social and demographic facets. Journal of
Advanced Nursing, 26, 921–929.
Spencer, N. J., & Coe, C. (1996). The development and
validation of a measure of parent-reported child health and
morbidity: The Warwick child health and morbidity profile.
Child: Care, Health and Development, 22, 367–379.
Stein, A., Cooper, P. J., Campbell, E. A., Day, A., & Altham,
P. M. E. (1989). Social adversity and perinatal complica-
tions: their relation to postnatal depression. British Medical
Journal, 298, 1073–1074.
Stretch, D. D., Nicol, A. R., Davison, I., & Fundudis, T.
(1999). The nature of mother and toddler problems. (ii)
Patterns of disturbance and disadvantage. European Child
and Adolescent Psychiatry, 8, 92–106.
Thorpe, K., Golding, J., MacGillivray, I., & Greenwood, R.
(1991). Comparison of prevalence of depression in mothers
of twins and mothers of singletons. British Medical Journal,
302, 875–878.
Veitch, D. (1995). Prescribing citizens advice: An evaluation of
the work of the citizens advice bureau with health and social
services in Birmingham. Birmingham District Citizens
Advice Bureau, Birmingham.
R. Reading, S. Reynolds / Social Science & Medicine 53 (2001) 441–453452
Warner, R., Appleby, L., Whitton, A., & Faragher, B. (1996).
Demographic and obstetric risk factors for postnatal
psychiatric morbidity. British Journal of Psychiatry, 168,
607–611.
Whitehead, M., Diderichsen, F. (1997). International evidence
on social inequalities in health. In: F. Drever, M. Whitehead
(Eds.), Health inequalities: Decennial supplement. Series DS
No 15. Stationary Office, London (pp. 44-68).
R. Reading, S. Reynolds / Social Science & Medicine 53 (2001) 441–453 453