Article Text

Download PDFPDF

Maternal mental health, and child growth and development, in four low-income and middle-income countries
  1. Ian M Bennett1,
  2. Whitney Schott2,
  3. Sofya Krutikova3,
  4. Jere R Behrman2
  1. 1Perelman School of Medicine, University of Pennsylvania, Philadelphia, Pennsylvania, USA
  2. 2The Population Studies Center, University of Pennsylvania, Philadelphia, Pennsylvania, USA
  3. 3Institute for Fiscal Studies, London, England
  1. Correspondence to Dr Ian M Bennett, University of Washington Department of Family Medicine, Box 453696, Seattle WA 98105-6099 ibennett{at}


Objective Extend analyses of maternal mental health and infant growth in low- and middle-income countries (LMICs) to children through age eight years, and broaden analyses to cognitive and psychosocial outcomes.

Design Community-based longitudinal cohort study in four LMICs (Ethiopia, India, Peru and Vietnam). Surveys and anthropometric assessments were carried out when the children were approximately ages 1, 5 and 8 years. Risk of maternal common mental disorders (rCMDs) was assessed with the Self-Reporting Questionnaire (SRQ)-20 (score ≥8).

Setting Rural and urban as well as low- and middle-income communities.

Participants 7722 mothers and their children.

Main outcome measures Child stunting and underweight (Z score ≤2 of height and weight for age), and <20th centile for: cognitive development (Peabody Picture Vocabulary Test), and the psychosocial outcomes self pride and life satisfaction.

Results A high rate of rCMD, stunting and underweight was seen in the cohorts. After adjusting for confounders, significant associations were found between maternal rCMDs and growth variables in the first year of life, with persistence to age 8 years in India and Vietnam, but not in the other countries. India and Vietnam also showed significant associations between rCMDs and lower cognitive development. After adjustment, rCMD was associated with low life satisfaction in Ethiopia but not in the other cohorts.

Conclusions Associations of maternal rCMD in the first year of life with child outcomes varied across the study cohorts and, in some cases, persisted across the first 8 years of life of the child, and included growth, cognitive development and psychosocial domains.


Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


Conditions in early childhood do matter for child growth and development. There is a substantial body of suggestive evidence that maternal mental health affects children in many domains, including nutritional status, health and cognitive and socioemotional development.1–3 Early impacts on physical growth persist through life.4 ,5 While most research has been carried out in high-income countries (HICs), the child outcomes associated with maternal mental health vary across the income status of the nation in which they are assessed; in HICs, these effects are primarily in the domains of psychosocial, behavioural and emotional development, while in some low- and middle-income countries (LMICs), associations with physical growth and illness are also seen; this difference in risk of poor growth outcomes presumably results from the relative caloric/nutritional and public health resources (ie, public water treatment) available in these different settings.1 ,6–11 In experimental studies, interventions to improve maternal depression have been associated with improved paediatric outcomes in HICs and in LMICs, supporting a causal relationship between these maternal mental health and child outcomes.12–14 The factors mediating this relationship are likely complex and may be dependent on the age of the child, since persistent benefit to these children has not been seen.15

Although general patterns across HICs and LMICs are apparent, there is a great need for additional research assessing the relationship of poor maternal mental health to child outcomes in LMICs.1 The range of economic and public health resources varies widely across these nations, making any generalisations tenuous, given the current small level of research among them. In the studies that have been carried out across LMICs, the relationship between maternal mental health, and infant and child outcomes, varies greatly, with significant associations in some countries but not in others; an association between maternal depression and growth abnormalities seen in South Asia has not been seen in sub-Saharan Africa (Ethiopia), for example.1 ,3 ,7 ,16 The mechanisms by which maternal mental disorders may create risk of poor child outcomes are likely to be diverse and complex, including social and physical environment, economic factors and genetics. Differences across LMICs, including cultural practices such as shared parenting, may moderate the effects of low resources and call for distinct intervention strategies.7

In addition to the limited number of studies carried out in LMICs, the assessment of maternal mental health associations with child outcomes has primarily been over a relatively short time period—generally within the first 12–18 months of the life of the infant.1 ,2 ,16 These limitations make it difficult to inform interventions that aim to address child outcomes in these vulnerable populations through maternal interventions. In this study, we examine the associations between risk of maternal common mental disorders (rCMD) and child growth, cognitive development and psychosocial well-being, between infancy and age 8 years, in four LMICs, which vary significantly across a range of cultural and economic contextual factors.15

Our study contributes to the existing literature in a number of important ways. First, the great majority of studies examining the effect of maternal depression on child outcomes are based on samples from HICs.1 ,17 ,18 Although sparse, the available evidence suggests that maternal depression prevalence rates in LMICs are unlikely to be lower than in HICs. In fact, some of the risk factors for depression (for instance, lack of support, negative life events) are more prevalent in poorer settings, and actual prevalence has been estimated to be higher in LMICs.19 Additionally, many of the studies available for developing countries are based on highly selected samples, such as patients in health clinics; or samples from restricted small geographical areas. Our study is based on samples that are approximately representative of national populations in three LMICs and of the states of Andhra Pradesh and Telangana, in India. Further, the focus of this literature is on the relatively brief first-year postpartum period that includes postpartum depression. While this focus is in part motivated by the importance of infancy for long-term development, there is a dearth of evidence on the association of maternal mental health during infancy with health and other outcomes later in childhood. Our data offer the opportunity to examine associations of maternal rCMD in infancy with outcomes not only in infancy but also at ages 5 and 8 years. Finally, the great majority of studies examine correlations between maternal mental health and child well-being (usually using cross-sectional data), without controlling for likely confounders of this relationship. The longitudinal design of the unique panel data set utilised in this study offers opportunities to address some challenges of identifying relationships between maternal rCMDs and different child outcomes at different ages in these four LMICs.


The data used for the current analyses are from Young Lives, a multicountry longitudinal study of child poverty, which tracks approximately 2000 children in each of four LMICs: Ethiopia, India (Andhra Pradesh, Telangana), Peru and Vietnam.15 ,20–22 The research was approved by appropriate ethics committees and conforms to the principles embodied in the Declaration of Helsinki. Informed consent was collected from parents and child at the earliest age possible, and the purpose of the research was clearly explained every time field workers visited a community, with emphasis that Young Lives was a research study and not a development project. Consent was viewed as an ongoing process and is frequently re-assessed with participants. In all four countries, approximately 100 children who were aged 6–18 months in 2002 were randomly selected from each of 20 sites (purposively selected to represent diversity within each country on key socioeconomic, demographic and geographic dimensions, with a pro-poor focus) to make a cohort of approximately 2000 very young children per country (an older cohort was also selected but is not used in this paper). Detailed descriptions of consenting, data collection and interview methodologies have been previously published.20–22 Briefly, adaptation of a common interview for each of the four countries was carried out through a pilot study focused on incorporating local idioms and expressions. Three rounds of data are available (in 2002, 2006–2007 and 2009), with a fourth round (2013–2014) scheduled to be released in late 2015.

Study measures

The primary measures for the analysis in this paper include maternal risk of common mental disorders (rCMDs),1 ,16 child nutritional status, child cognitive development and child psychosocial outcomes. rCMD was measured using the Self-Reporting Questionnaire (SRQ-20), a screening tool developed by the WHO specifically for developing countries and widely used in that context.23–25 It consists of 20 yes/no statements relating to the mental well-being of the mother. As with other instruments used for epidemiological studies of both mental and physical health, this measure is designed to identify risk of mental disorders, with appropriate psychometric properties, and is not a diagnostic tool. The SRQ-20 also identifies risk of depression and anxiety/stress, the most common maternal mental health disorders, but does not distinguish between them. Based on these considerations, we modify the previously adopted language ‘common mental disorder’ (CMD) by adding ‘risk’ (rCMD) to describe a positive SRQ20 screen in order to better reflect the characteristics of the tool.11 This instrument has acceptable levels of reliability and validity in a range of developing countries. Cut-off scores to determine how many yes answers constitute a possible case, balanced against acceptable levels of false positive cases, have been validated against clinical assessments in each of the study countries.23 ,24 These validations indicated a score cut-point of 7/8, to separate risk of non-cases/cases of CMD. We used multiple imputation procedures to fill in missing data on independent variables with the ice command in Stata V.12.1 software (StataCorp LP) and the option of 15 imputations.

The main outcomes of interest are child nutritional status, cognitive skills and psychosocial outcomes. Nutritional status was captured for each of the three rounds of the survey using two anthropometric indicators of long-term growth: (1) stunting, height-for-age z-scores ≤2; and (2) underweight, weight-for-age z-scores ≤2. Cognitive skill was assessed at ages 5 and 8 years using the Peabody Picture Vocabulary Test (PPVT), a widely-used test of receptive vocabulary.26 The PPVT was selected for use in this study, based on pilot studies using several cognitive assessments.27 This instrument measures vocabulary acquisition in persons aged from 2.5 years to adulthood. The test is individually administered, untimed and orally delivered. The task of the test taker is to select the picture that best represents the meaning of a stimulus word presented orally by the examiner. The reliability and validity of this measure was established for each country prior to use in the current study and is described in detail elsewhere.27 Briefly, classical test theory and item response theory were both utilised to estimate reliability indicators. The validity analysis involved assessments of correlations between the PPVT items, and variables such as age and educational attainment previously identified in the literature. Psychometric analyses showed that while the general construct measured by this instrument was the same across the four countries, individual items with the best psychometric properties differed; as a result, while the PPVT is norm-referenced for HICs (but not LMICs) these norms were not used in the current analysis—rather, country-specific distributions were used to determine risks.27 We define low cognitive skill as scores at or below the 20th centile in each round of (country-specific) scores for that cohort.

Psychosocial outcomes were measured using assessments of self-pride and life satisfaction at age 8 years.27 ,28 Self-pride was used instead of the more commonly used psychological construct of self-esteem, which the Young Lives study team conceptualised as being similar in meaning, in order to increase understanding among children.27 ,28 Self-pride was assessed through their agreement with three statements: (1) “I am proud of my shoes or having shoes”, (2) “I feel my clothing is right for all occasions”, and (3) “I am proud of my clothes”. Responses were on a five point Likert-type scale from “totally disagree” (1 point) to “totally agree” (5 points). These scores were summed; having a score in the lowest quintile for the country was used as the risk category for analyses.28 The degree of overlap of self-pride with self-esteem has not been evaluated and so we have chosen to use the concept of self-pride rather than to claim that these measures adequately assess the self-esteem domain.

Life satisfaction is important to consider in poverty research because the poorest are not necessarily the least satisfied, and there is growing recognition that people's own perceptions of their situations should be taken into account when seeking to develop or improve their living conditions.29 The ‘Ladder of Life’ satisfaction tool, based on widely-used validated measures, has been used in global studies of satisfaction and happiness, including in studies of development in India and Ethiopia.30–32 Children were asked to indicate their position between their best-possible and worst-possible lives on a nine-point scale represented visually as a ladder. We again used scoring in the lowest quintile to indicate the risk category.

Potential confounders were controlled for based on conceptual considerations and previous work with which we want to compare the current analyses. Child characteristics included age in months at each of the study assessment periods, and gender (female). We included schooling attainment as a maternal control variable because of the extensive linkage between maternal schooling and maternal mental health and child health status.33 A household wealth index based on the first principal component of household assets, housing quality and service access, was used as a control for household resources.


The analytic approach taken in the current study is framed by a conceptual model linking maternal mental health with child outcomes adapted from previous work with the current study sample (figure 1).16 Descriptive analyses were carried out for all the primary independent and dependent variables. Because of the high prevalence of the outcome condition, we chose to use Modified Poisson regression analyses with robust error variance and with SRQ-20 score of <8 versus ≥8 as the primary independent variable; child outcomes at each of the appropriate ages were regressed on elevated SRQ-20 score to create risk ratios (RR) for assessment of magnitude of association between the independent and dependent variables.34 We then regressed child outcomes on elevated SRQ-20 and the control variables to create adjusted RRs (aRRs).

Figure 1

Conceptual framework and measures.


Table 1 presents the prevalence of women with maternal rCMD, and child stunting, underweight, cognitive development and psychosocial outcomes and the distributions of the other variables included in the conceptual framework (figure 1), for the four study populations. Rates of rCMD ranged from 20% in women from Vietnam to 33% in those from Ethiopia. Though there was variation, we found a high prevalence of stunting and underweight in all populations assessed. Large variations in child growth and development were seen as well across these samples.

Table 1

Description of study population

The unadjusted (RRs) and adjusted risk ratios (aRRs) for the association of maternal rCMD with independent variables are shown in table 2. The crude association of maternal rCMDs with growth measures was significant at age 1 year for the cohorts in India and Peru, and persisted to age 8 years for both of these countries, and was significant for stunting in Vietnam at ages 5 and 8 years. For underweight, the relationship was significant for India and Vietnam, at ages 1, 5 and 8 years. After adjustment for confounding, significant associations were retained for India in stunting (ages 1 and 8 years), and underweight in India (ages 5 and 8 years) and Vietnam (ages 1 and 8 years). Cognitive and psychosocial outcomes varied as well. Significant unadjusted and aRR of low PPVT score (lowest quintile) was seen in India at age 8 years while a significant aRR was seen at age 5 years in Vietnam. For the psychosocial variables, an association between maternal rCMD and low self-pride was seen only in the crude model in Vietnam; a significant association with low life satisfaction was seen in crude and adjusted models in Ethiopia and, in India, only in the crude model.

Table 2

Risk of poor growth, cognitive development and psychosocial outcomes among children of women with probable rCMD (SRQ20 ≥8) at child aged 1 year


In this analysis of the mothers of 7772 children, we found associations between maternal mental health in the first year of life with growth and development as well as the life satisfaction of that child through 8 years of age, with varying patterns across four LMICs. The prevalence of elevated risk of maternal mental disorders is consistent with epidemiological studies of mental disorders globally.35 We found significant associations in adjusted models between exposure to maternal rCMDs in early childhood and increased risk of poor growth and cognitive development in India and Vietnam, not just at the time of the assessment of mental health symptoms, but persisting to later ages (5 and 8 years). The association seen for stunting in Peru, in contrast, does not persist in adjusted models, and no significant associations were found for growth outcomes in Ethiopia. We also found that risk of poor cognitive development in India and life satisfaction in Ethiopia is detected at age 8 years, after exposure to maternal rCMD in infancy, consistent with our model linking maternal mental health and longer-term outcomes for children.

The heterogeneity of associations between maternal mental health and paediatric outcomes across countries is consistent with the existing literature in HICs and LMICs, raising questions about the mechanism by which these variables are linked.1 ,3 While a role for reduced resources in worsening the risk for child outcomes is generally supported by existing research showing worsened outcomes in LMICs than HICs, particularly for growth parameters, the pattern within LMICs is less clear. An overall significant association is found for cognitive and growth outcomes in pooled analyses of data from LMICs, but economic conditions at the national level are not consistently predictors of these outcomes. In the current study, we found evidence for poor growth outcomes in the middle-income countries India and Peru, but not in the third middle-income country considered, Vietnam, or in Ethiopia, the one low-income country assessed. The last result is consistent with other reports, which also found no associations in Ethiopia and proposed protective effects of cultural factors such as shared parenting to account for this difference from other contexts.7 Strong associations between clinical depression among mothers in South Asia, and growth and cognitive abnormalities, have been well documented along with benefits from successful depression treatment, supporting a direct mechanistic linkage between this particular mental disorder and the childhood outcomes.6 ,8–11 In the South Asian context, a role for infectious risk has been identified that contributes to this risk and could be distinct in the Ethiopian setting, though precisely how remains to be explored.10 Interventions focused on supporting child development are effective in reducing the impact of maternal depression on child outcomes, suggesting that a number of distinct intervention strategies may be effective.10 How distinct pathways of risk and protection may influence childhood outcomes in these geographically diverse settings is an important area for future research.

The persistent effects of poor maternal mental health on child growth and cognitive development that we describe are consistent with a life-course epidemiological framework; negative exposures in vulnerable periods of development can have impact through long periods in a child's life.1 The majority of previous studies on this topic have been carried out in HICs and those that have looked at the LMIC setting have focused on outcomes within the first or second years of life. The presence of growth stunting and underweight at the same time points for India and Vietnam, suggests common mechanistic pathways and/or diverse effects of maternal mental health on child growth and development. Poor nutritional status resulting from reduced functioning of the mother in critical periods could account for both outcomes, though distinct channels of infectious, nutritional and psychosocial impacts are plausible. Findings that some of these benefits seen in developmental domains in the South Asian context do not persist when followed to age 7 years of the child, add to the complexity of the relationship, however.15 Such studies highlight the critical need to look at this question longitudinally and over longer periods of child development than has typically been the case.

Our finding that maternal rCMD in infancy was associated with low life satisfaction in children at age 8 years in Ethiopia is novel and of interest. The potential sensitivity of this outcome to exposure to early maternal rCMD provides new insights into possible processes and mechanisms by which maternal mental health can negatively influence child trajectories of well-being and lifetime success. Previous studies assessing negative impacts of maternal depression on children have focused on behavioural outcomes in younger children. In these studies, an increased risk of psychopathology/behavioural abnormalities is seen in the children of women with major depression in the year postpartum—risk that is ameliorated with effective treatment of the mothers.9–11 The assessment of other aspects of psychological well-being, such as life satisfaction, provides an opportunity to more fully understand the consequences of this exposure and potentially provides new avenues for helping children with this exposure. Additional work is needed to confirm and extend these novel findings. Interventions to improve maternal mental health might consider adding measures such as this to assess benefits to children of such programmes. The heterogeneity in growth, cognitive and psychosocial outcomes is of significance, as it is clear that a ‘one-size-fits-all’ approach to addressing these issues will not reflect the variability seen across national contexts. It is unclear what factors might result in the wide variation seen. Exploration of potential protective effects of national conditions is warranted.

There are a number of limitations to the current study. First, the Young Lives study is observational—and while it is longitudinal, any associations identified are weak evidence for causality. However, studies of this type provide a necessary opportunity to explore a complex set of conditions and factors that can lead to significant insight into the economic, social and psychological forces influencing children in poverty in LMICs. Second, there was a significant period of elapsed time between the measure of maternal rCMDs and the developmental and psychosocial outcomes. Only the anthropometric measures were available before age 5 years. However, we were able to detect persistent associations of maternal rCMD with these critical domains of child well-being. Finally, the SRQ20 is a screening instrument and is not able to provide a clinical diagnosis of mental disorders. It is possible that more precise mental health measures would identify greater magnitudes of associations with child outcomes. Nevertheless, we were able to see significant associations of elevated SRQ20 scores with outcomes. Despite these limitations, the study contributes to the literature by investigating the associations between rCMD when children were infants and child growth, cognitive development and psychosocial well-being, between infancy and age 8 years in four LMICs with very different contexts. This information is key to formulating policy for appropriate interventions in these vulnerable settings.1

What is already known on this subject

  • It is known that poor mental health in mothers is associated with poor infant nutritional status and measures of cognitive development in some low- and middle-income countries (LMIC). It is not well known, however, whether these associations persist outside of the infant period and whether the effect of maternal mental health extends to other psychosocial outcomes.

What this study adds

  • This paper extends the time period of assessment from infancy through 8 years of life of the child, and across four distinct low- and middle-income countries. We show that markers of poor nutritional status associated with poor maternal mental health persist in countries where the association is seen in infancy into mid-childhood. In addition, we show that cognitive and psychosocial factors are also affected by poor maternal mental health into childhood. These findings provide important evidence of the significant consequences of maternal mental health on child growth, development and well-being in diverse LMIC settings.


The author thank to Nan M Astone, assigned discussant, and other participants in Session 222 on Parental Characteristics on Child Health and Behavioural Outcomes at the Population Association of America Annual Meetings (Boston, MA, 3 May 2014), for useful comments on an earlier version.



  • Funding This research has been supported by the Bill and Melinda Gates Foundation (Global Health Grant OPP10327313), Eunice Shriver Kennedy National Institute of Child Health and Development (Grant R01 HD070993) and Grand Challenges Canada (Grant 0072-03 to the Grantee, The Trustees of the University of Pennsylvania). The data used in this study come from Young Lives, a 15-year survey investigating the changing nature of childhood poverty in Ethiopia, India (Andhra Pradesh, Telangana), Peru and Vietnam ( Young Lives is core-funded by UK aid from the Department for International Development (DFID) and co-funded from 2010 to 2014 by the Netherlands Ministry of Foreign Affairs. The findings and conclusions contained within are those of the authors and do not necessarily reflect positions or policies of the Bill and Melinda Gates Foundation, the Eunice Shriver Kennedy National Institute of Child Health and Development, Grand Challenges Canada, Young Lives, DFID, or other funders.

  • Competing interests None declared.

  • Patient consent Obtained.

  • Ethics approval University of Pennsylvania Institutional Review Board.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Data sharing statement Data used in the current analysis are available through the Young Lives study.