Impact of Maternal Depression on Infant Nutritional Status and Illness: Title and subTitle BreakA Cohort Study FREE

More than 150 million children under the age of five years in the developingworld are malnourished. Malnutrition is implicated in more than half of allchild deaths worldwide,¹ and nutritional deficienciesat all stages of growth have long-term damaging effects on the intellectualand psychological development of children.^{2 - 3} Malnutritionthus represents an enormous waste in potential of millions of children.

Half of these children live in South Asia. Determinants of disproportionatelyhigher rates of malnutrition in this largely food-sufficient region are poorlyunderstood,⁴ the problem having been referredto as the "Asian enigma."⁵ Evidence seems toindicate that as the amount of food available per person increases, its powerto reduce child malnutrition weakens.⁶ Therefore,attention has gradually been turning to factors other than nutritional intake,such as household behaviors and sociocultural practices, which may influencechild health and development.⁷

Depression around childbirth is common, affecting approximately 10%to 15% of all mothers in Western societies.⁸ Recentepidemiological findings suggest that prevalence rates may be almost twiceas high in South Asian mothers.^{9 - 10} Depressionranks among the top 5 disabling disorders worldwide.¹¹ Thereis compelling evidence that maternal depression adversely affects the psychologicaland intellectual development of children.¹² Itis plausible that maternal depression may also put the child's physical healthat increased risk, especially in low-income countries where the environmentis frequently more hostile than in the developed world.¹³ Maternaldepression is also likely to interfere with the emotional quality of childcare, which is a known risk factor for poor growth.^{14 - 15} Thearea is underresearched in low-income developing countries. Three health center–basedstudies have found positive associations between maternal mental health andinfant growth^{16 - 18} buthave limitations of small, selective samples and designs that cannot establishthe direction of association.

The aim of this study was to systematically investigate the associationbetween prenatal and postnatal maternal depression and infant physical outcomesin a representative rural community–based sample in Rawalpindi, Pakistan,using a longitudinal, prospective cohort design.

The study was carried out in a rural subdistrict 60 km southeast ofthe city of Rawalpindi. The study area comprised 10 Union Councils (a UnionCouncil is a geographically defined administrative area consisting of 5-10villages) in the south of the subdistrict, having an area of about 500 km² and a population of 150 000. The area is predominantlyagrarian, food sufficient, and socioculturally homogeneous. The sample consistedof all women aged 17 to 40 years in their third trimester of pregnancy, recruitedfrom the study area during a 4-month period. Subjects were identified by obtainingofficial lists from 120 government-employed Lady Health Workers, who routinelycollect data on new pregnancies, including an estimate of the gestationalage (based on the last menstrual period). Each Lady Health Worker covers apopulation of 1000 (approximately 130 households) and visits about 30 to 35houses per week. As an additional incentive, they were paid a small amountfor every subject identified. In some villages, these data were verified byemploying local traditional birth attendants and key informants to carry outdoor-to-door surveys. Thus, near-full coverage of the study area was achieved.Identified women were included in the study if they were currently married,did not have a physical illness for which they were undergoing treatment,and had had an uneventful pregnancy. Women with severe depression or othermental disorders were excluded. Stillbirths, infants who died before reachingtheir first birthday, and infants born with a congenital abnormality wereexcluded from the study, as were mothers who gave birth prematurely (<37weeks according to the gestational age calculated by the Lady Health Workers).

Informed written consent was obtained from all participants after weexplained the study and provided a detailed information sheet (local healthworkers read out and explained the information to nonliterate participantsand obtained written consent on their behalf). Mental state was assessed usingthe Schedules for Clinical Assessment in Neuropsychiatry (SCAN),¹⁹ developedby the World Health Organization as an internationally validated semistructuredinterview generating International Classification of Diseases,10th Revision (ICD-10)²⁰ diagnoses ofdepressive disorder. All interviews were carried out by 2 trained and experiencedclinicians. Interrater reliability was established prior to the study whenboth interviewers independently assessed 20 women (10 had clinical depression)and agreed on the diagnosis of 19 (κ = 0.90). Interviews were carriedout in the third pregnancy trimester and at 2, 6, and 12 months postnatallyby the 2 interviewers who were blind to the health status of the infants.Disability in mothers at these time points was assessed using the crossculturallyvalidated Brief Disability Questionnaire,²¹ an8-item questionnaire asking subjects whether they had been limited in variouseveryday activities (eg, lifting heavy objects, carrying groceries or goods,climbing stairs, walking uphill, or routine personal and family functions)during the last month. The Brief Disability Questionnaire has a maximum scoreof 16, constituting a nonparametric, hierarchical scale.

Infant growth measurements were carried out using standard anthropometrictechniques and equipment.²² All growth measurements,except birth weight, were carried out by a single trained researcher at 2,6, and 12 months postnatal. Birth weight was measured by trained local healthworkers. Growth data were converted into standard deviations (z scores) for weight and length using Epi Info 2002 software (Centersfor Disease Control, Atlanta, Ga). Infants were classified as underweightor stunted if their weight-for-age z scores or length-for-age z scores were less than −2, on the basis of NationalCenter for Health Statistics/World Health Organization reference data.²³

Infants of both groups (depressed and nondepressed mothers) were seenfortnightly for a year by the same health workers to assess the number ofepisodes of diarrhea and acute respiratory infection (ARI) during the previous2 weeks. Diarrhea was defined as 3 or more unformed stools passed in a 24-hourperiod, and a diarrheal episode was defined as episodes separated by at least3 diarrhea-free days; ARI was defined as the presence of cough, nasal/postnasalsecretions, and rapid breathing, each episode separated by at least 7 days.²⁴ On average, a Pakistani child has 5 episodes of diarrheaand 6 episodes of ARI per year.²⁵ On this basis,these outcomes were classified as 5 or more diarrheal episodes per year oras 6 or more ARI episodes per year. To minimize recall bias, the mother'sreport of infections during the previous 2 weeks was corroborated from anothermember of the household, such as the grandmother, the husband, or an oldersibling. Records of immunization were assessed at 1 year for all infants inthe study, and we classified subjects into those with or without up-to-dateimmunization status. All infant outcomes were assessed by researchers blindto the psychiatric status of the mother.

Depressed and nondepressed mothers were compared on ownership of householdassets derived from the World Bank Assets Questionnaire for Pakistan.²⁶ More subtle socioeconomic differences were assessedby inquiring if the household was in debt and by asking Lady Health Workers,who lived in the same locality and had intimate knowledge of the familiesbeing studied, to rate the household on a 5-point Likert scale ranging from1 (richest) to 5 (poorest). A single dichotomous variable of relative povertywas created by combining these 2 measures, ie, subjects who were both in debtand rated below 3 on the socioeconomic 5-point Likert scale were classifiedas being relatively poor.

Social support was assessed by inquiring about the traditional chilla period: postdelivery confinement of mothers fora period of 40 days, when all domestic responsibilities are taken over byother female family members. The mother is presented fortified dietary itemsas gifts by friends and neighbors and assisted in child care by more experiencedfamily members. The completion of the 40-day chilla periodis a good proxy indicator of social support because it encompasses both familyand community support and is an easily understood concept. The mother wasasked if she had been able to observe chilla forthe full 40 days, who had been the main person supporting her during the period,and whether she enjoyed full rest and dietary supplementation. Based on theresponses to these queries, the observance of the chilla period was rated as satisfactory or not satisfactory, as a proxy measurefor social support.

Maternal financial empowerment within the household was measured byasking the mothers if they were given a lump-sum amount of money for day-to-dayhousehold expenses and whether they could make independent decisions aboutits use. Mothers who answered yes to both questions were classified as financiallyempowered within the household.

Data were also collected on other known variables that could confoundthe association between maternal depression and infant health outcomes. Thesevariables were categorized on an a priori basis as: (1) Female vs male infantsex. In some South Asian communities, the male child is more valued than thefemale child, receiving preferential treatment in terms of nutrition and care.²⁷ The birth of a female child has also been shown tobe a risk factor for maternal depression.⁹ (2)Low birth weight defined as 2500 g or less vs normal birth weight of morethan 2500 g. Low birth weight is an important contributor to continuing malnutritionand infant mortality, especially in the first year of life.²⁸ (3)Duration of exclusive breast-feeding: less than 6 months vs at least the first6 months of life. The World Health Organization recommends exclusive breast-feedingfor the first 6 months.²⁹ (4) Maternal ageof 20 years or younger vs 21 years or older. Younger maternal age is associatedwith increased morbidity and mortality of infants.³⁰ (5)Maternal age of 30 years or older vs 29 years or younger. Older maternal ageis also associated with increased morbidity and mortality of infants.³⁰ (6) Low maternal body mass index at 1-year post partumdefined as 18.5 or lower vs maternal body mass index of higher than 18.5.Low body mass index indicating chronic energy deficiency in mothers is associatedwith poor child nutritional status³¹ and mayalso impact maternal mood. (Body mass index is measured as the weight in kilogramsdivided by the height in meters squared.) (7) Each parent's education: noeducation vs at least 4 years of primary education. Four years of schoolingwas chosen as a cutoff because many Pakistani female children attend primaryschool for 4 years, after which many children from low-income families stopattending. (8) Two or more children younger than 7 years vs fewer than 2 childrenyounger than age 7. In Western societies, having 2 or more young childrento look after is associated with depression in women³² andmay also mean that the mother has less time to attend to the new infant'sneeds. (9) Four or more children vs fewer than 4 children. This cutoff waschosen because we believed that low-income families with 4 or more childrenmight have additional difficulties such as financial constraints and overcrowding.(10) Nuclear family (parents and children only) or extended family (3 generations,ie, one or both parents with married sons and their wives and children). Theextra support from an extended family could have a protective influence onthe infant's health, especially if the mother's care-taking abilities arecompromised because of poor mental health.

Preliminary analyses consisted of descriptive frequencies and χ² and Mann-Whitney U tests to compare differencesbetween depressed and nondepressed mothers in household assets, delivery care,and personal characteristics. Mean differences in weight-for-age z scores and length-for-age z scores at alltime points between the exposed group (infants of prenatally depressed mothers)and the nonexposed group (infants of prenatally nondepressed mothers) wereanalyzed, first by using growth indices as continuous measures. Univariateanalyses were then carried out to estimate relative risks with 95% confidenceintervals (CI) of being underweight (weight-for-age z score<−2) and stunted (length-for-age z score<−2). Univariate analyses were also carried out to estimate the relativerisk of other outcomes (low birth weight, diarrhea, ARI rate, and immunizationstatus). We did not match on variables such as socioeconomic status or educationbecause we could not be sure if these were confounders or part of a causalpathway linking maternal depression to infant growth, and matching would nothave allowed us to study their effects.³³ Multiplelogistic regression was used to simultaneously control for the confoundingeffects of all the variables under study and obtain odds ratios as measuresof association. Population-attributable risk was calculated as the proportionof stunting (a good indicator of malnutrition) in the 12-month-old infantsthat would not have occurred if the effect associated with maternal depressionwere absent. This took into account relative risk adjusted for the full cohortof 632 infants. To assess a dose-response relationship, similar analyses werecarried out on a subcohort of chronically depressed mothers (depressed atall 4 time points) and mothers not depressed at any time point. All analyseswere carried out with Stata.³⁴

Estimates on indices of undernutrition (weight-for-age and height-for-age z scores <−2) in Pakistani children range from38% to 50%. Assuming a conservative estimate of a 20% prevalence of undernutritionamong infants of nondepressed mothers and a 2-fold increase in risk with depressedmothers, a sample size of 91 infants in both groups would be sufficient todetect this with a precision of 0.05 and 80% statistical power.

The research ethics committees of the University of Manchester and RawalpindiMedical College approved the study.

Seven hundred one women in their last trimester of pregnancy (mean,6 weeks from delivery date) were identified. Thirty-one (4%) refused to takepart; of the remaining 670, 14 (2%) suffered from a physical disorder, and24 (4%) had other mental disorders (mostly generalized anxiety) and were excluded.Thus, 632 women were interviewed with the SCAN for an evaluation of theirmental state. Of these, 160 were diagnosed as having an ICD-10 depressive episode, giving a prevalence rate of depressive disorderin the prenatal period of 25%. Each depressed woman was matched with a psychologicallywell woman of similar gestation residing in the same Union Council. Depressedmothers had significantly higher scores on the Brief Disability Questionnairethan nondepressed mothers, both prenatally (median score, 7 vs 2; P<.01) and 12 months postnatally (median score, 8 vs 1; P<.01).

At birth, 8 infants (3%; 4 from the depressed group and 4 from the nondepressedgroup) were born prematurely and were excluded. Two mothers discontinued participationbecause of the severity of depression. Eighteen (6%; 10 from the depressedgroup and 8 from the nondepressed group) had stillbirths or neonatal deaths,and 2 newborns (1 each from the depressed and the nondepressed groups) hadcongenital abnormalities and were excluded. During 1 year, 25 (8%; 14 fromthe depressed group and 11 from the nondepressed group) dropped out. The overallresponse rate was therefore 92% with little differential loss of follow-upbetween the 2 groups.

Depressed and nondepressed mothers were similar in terms of householdand personal characteristics studied. Ninety-eight percent of households inboth groups possessed electricity (χ²₁ = 0.002; P = 1.0), and 41% depressed vs 38% nondepressed owned agriculturalland (χ²₁ = 0.22; P = .60).No differences were found between the 2 groups on principal sources of drinkingwater: manual well (67% in depressed group vs 68% in nondepressed group),well with pump (25% vs 26%), and tap (8% vs 6%) (χ²₂ =0.35; P = .80); or type of toilet facility used:field or pit (53% vs 52%), flush toilet (45% both groups), or other such asbucket (2% vs 3%) (χ²₂ = 0.59; P = .70). No significant differences were found between depressed andnondepressed mothers in the type of delivery care received: birth attendedby medically trained personnel (26% vs 24%), by traditional birth attendants(17% vs 21%), and by family member at home (57% vs 54%) (χ²₂ = 0.78; P = .70). The median age at birthin both groups of mothers was 26 years (interquartile range, 24-30 years).Median body mass index at 12 months postnatal in depressed mothers was 21.4(interquartile range, 19-24.7), and that of nondepressed mothers was 22 (interquartilerange, 19.6-25.4); the difference was not statistically significant (Mann-Whitney U, 1.24; P = .20). About 45% ofmothers in both groups had had no formal schooling, and only 5% in both groupswere employed outside the home.

Figure 1 and Figure 2 show differences in mean weight-for-age and length-for-age z scores of infants of both groups at all times of assessment.The differences increased with the age of the infant, becoming maximum at1 year. Mean weight-for-age z-score difference atbirth was 0.30 (95% CI, 0.01 to 0.59; P<.05),increasing at 1 year to 0.70 (95% CI, 0.48 to 0.92; P<.01).Mean length-for-age z-score difference at 2 monthswas 0.24 (95% CI, –0.01 to 0.48; P = .06),increasing at 1 year to 0.63 (95% CI, 0.38 to 0.88; P<.01).Length was not measured at birth.

The prevalence of underweight (weight-for-age z scoreof less than –2) and stunting (length-for-age z scoreof less than –2) in the sample was 5% and 6%, respectively, at 2 months;18% and 10% at 6 months; and 29% and 24% at 12 months. Because prevalenceat 2 months was low, further analyses were only carried out for measurementsat 6 and 12 months. Univariate analyses (Table 1) showed a significant relative risk of being underweightor stunted at both 6 and 12 months with maternal depression. Other factorsthat had a significant negative impact on infant weight or length includelow birth weight, less than 6 months of breast-feeding, 5 or more diarrhealepisodes per year, lack of parental education, and relative poverty, whereasmaternal financial empowerment had a positive impact. Relative risk from allother variables studied was insignificant (data not shown in table). Multiplelogistic regression (Table 2)showed that after simultaneous adjustment for all variables, the associationbetween infant growth and maternal depression remained significant. The onlyother variables that continued to show a statistically significant associationinclude 5 or more diarrheal episodes per year and relative poverty, whereasassociation with low birth weight, less than 6 months of breast-feeding, lackof parental education, and maternal financial empowerment became insignificant.Association with all other variables also remained insignificant (data notshown in table).

Assuming that the associations are causal and not confounded, the population-attributablerisk for stunting at age 12 months is 30% (95% CI, 19 to 41). The population-attributablerisk is the proportion by which the incidence of stunting would be reducedif maternal depression were eliminated from the population. Fifty-six percentof mothers depressed in the prenatal period were depressed at all points ofassessment in the postnatal period, indicating that the rate of chronic depressionin the sample was high. When these chronically depressed mothers (n = 72)were compared with mothers not depressed at any time point (n = 108), therelative risk for the infants being underweight and stunted increased substantially(at 6 months of age, relative risk for underweight was 5.9; 95% CI, 2.7 to12.8; for stunting at 6 months of age, relative risk was 5.5; 95% CI, 1.9to 16.0; at 12 months age, relative risk for underweight was 3.5; 95% CI,2.2 to 5.6; and for stunting at 12 months of age, relative risk was 3.2; 95%CI, 1.9 to 5.4).

Table 3 shows the associationof maternal depression with other outcomes of interest at birth and at 12months. Relative risks for both low birth weight and excessive diarrheal episodeswere statistically significant; however, there was no association with excessiveARI episodes and only a weak association with immunization status being upto date.

To our knowledge, this is the first longitudinal, population-based studyexploring the impact of maternal depression on infant health in a rural areaof a developing country. Exposure and outcome measures were carefully defined,and the study achieved a high follow-up rate. The main findings are that infantsof prenatally depressed mothers have poorer growth and an increased risk ofdiarrheal infection compared with infants of psychologically well mothers.The study population is comparable to that of the developing world in generalon important human development indicators such as infant mortality rate (56per 1000 live births in the study area vs 63 per 1000 live births for alldeveloping countries) and female literacy rate (53% vs 65%).³⁵ However,the findings would need to be applied with caution in poorer populations facedwith grave food shortages.

By determining the exposure status prior to the development of outcomes,the study ensured that the correct causal temporal sequence was achieved.A higher proportion of infants of prenatally depressed mothers had lower birthweight, indicating that part of the later disadvantage could have been conferredin utero. However, the impact of maternal depression on later growth remainssignificant after adjusting for low birth weight, showing that postnatal depressionhas an independent effect. Longer duration of depression in the mothers confersa greater risk, and infants of mothers who remain well throughout the yearachieve the best outcome, indicating a dose-response relationship. It cannotbe assumed, however, that this dose-response relationship is unidirectional.Poor infant health may be a stressor for mothers and therefore act as a maintainingfactor for their depression.

The effect-size of maternal depression is greater than that of importantknown risk factors for poor growth such as low birth weight, frequent diarrhea,and socioeconomic status, and this effect remains significant after simultaneousadjustment for these factors. Maternal depression in this sample proved tobe a stronger predictor of poor growth than the other known risk factors studied;reductions in the prevalence of maternal depression could result in a reductionin infant growth retardation of up to 30%. Thus, maternal depression couldmake an important and possibly major contribution to poor infant growth outcomesand morbidity in other less resourced countries.

Possible mechanisms by which maternal depression affects infant growthand illness include a less healthy lifestyle and reduced care-seeking in theprenatal period; maternal disability in the postnatal period resulting indeficient physical and emotional care and psychosocial stimulation of theinfant; and associated psychosocial difficulties such as lack of family supportto the mother and lack of financial empowerment, which could impact infantcare.¹³ Infants of depressed mothers were lesslikely to be fully immunized at 12 months compared with infants of nondepressedmothers, possibly indicating a lack of appropriate health-seeking behaviorin depressed mothers. Another finding that may indicate deficiency in careis that although infants of depressed mothers have more diarrheal episodes,rates of ARI are the same (Table 3).Maternal child-care behaviors such as hand-washing before feeding, safe foodpreparation and storage, and obtaining clean drinking water are more likelyto influence diarrhea than ARI.³⁶ It mightbe expected that disability in depressed mothers could influence such preventiveactivities, thus increasing the risk for diarrhea, whereas ARI preventiondoes not have such a direct link with maternal child-care behavior. Socioeconomicstatus and parental education appear to have only weak effects on infant growth,but further research is required to understand the nature of possible mechanismsand interactions with these and other factors such as social support and empowerment,which may moderate the effects of maternal depression.

Vulnerability to undernutrition after the first 6 months of life isvery high, particularly at the time complementary foods are introduced toan immunologically vulnerable infant.²⁸ Althoughthe prevalence of poor-growth outcomes increases with age, the risk from maternaldepression relatively decreases. One possible explanation is less relianceon the mother (other carers may assist in care and complementary feeding,and the infant may start to forage for food), diluting the effect of maternaldepression. Another possible explanation is the gradual recovery of many mothersfrom this self-limiting disorder. While these are plausible explanations,there may be some other unknown factor particular to the study sample thatis responsible for this pattern of infant growth.

The study has implications for both mother and child health. Ordinaryhealth workers can identify depression with relative ease, using simple checkliststhat have excellent validity. The use of these instruments can help identifygroups of mothers whose infants are at a greater risk for poor health. Interventionscould then be targeted to these groups.

Recent approaches to child survival strategies call for a focus on thehousehold as a center for child health activity.⁷ Preventivehousehold behaviors, such as infant-feeding practices, immunization, and home-healthand care-seeking practices, rely heavily on the mother, who is the primarycare provider in most developing societies. The impact of health-promotingactivities is therefore related to the functional capacity of mothers, theirreceptivity to health care messages, and their uptake of the interventionoffered. Attention to the mothers' psychological well-being could increasethe effectiveness of such programs.

The association between maternal mental health and infant physical healthcould help promote women's mental health in the health care agenda. It couldalso provide a window of opportunity to elevate the social status of womenin developing countries and in the process improve their own and their children'sphysical and mental well-being.

Correspondence: Atif Rahman, PhD, University of Manchester Departmentof Child Psychiatry, Royal Manchester Children's Hospital, Hospital Road,Pendlebury, Manchester M27 4HA, England (atif.rahman@ntlworld.com).

Submitted for publication November 27, 2003; final revision receivedFebruary 26, 2004; accepted March 18, 2004.

This study was supported by grant 060177/Z/00/Z from the Wellcome Trust,London, England.

We thank Naeem Tariq, PhD, and Mohammed Pervez, PhD, at the NationalInstitute of Psychology, Islamabad, Pakistan, for their support and assistance.