Article Text


Stunting risk of orphans by caregiver and living arrangement in low-income and middle-income countries
  1. Jocelyn E Finlay1,
  2. Günther Fink2,
  3. Dana Charles McCoy3,
  4. L Carolina Tavárez4,
  5. Jeanne Chai2,
  6. Goodarz Danaei2,
  7. Majid Ezzati5,
  8. Wafaie Fawzi2,
  9. Mary C Smith Fawzi4
  1. 1Harvard Center for Population and Development Studies, Harvard University, Cambridge, Massachusetts, USA
  2. 2Department of Global Health and Population, Harvard University, Cambridge, Massachusetts, USA
  3. 3Graduate School of Education, Harvard University, Cambridge, Massachusetts, USA
  4. 4Department of Global Health and Social Medicine, Harvard Medical School, Boston, Massachusetts, USA
  5. 5MRC-PHE Centre for Environment and Health, Department of Epidemiology and Biostatistics, School of Public Health, Imperial College London, London UK
  1. Correspondence to Dr Jocelyn E Finlay, Research Scientist, Harvard University, 9 Bow St, Cambridge, MA 01238, USA; jfinlay{at}


Background An estimated 151 million children worldwide have lost one or both parents (UNICEF). Although a considerable literature has documented the impact of parental loss on children's emotional development, to date there are mixed findings regarding the potential effect of orphanhood on children's physical growth. The aim of the study is to examine the association between orphanhood (maternal, paternal or both) on childhood stunting status, accounting for living arrangement.

Methods We combine nationally representative data from 84 Demographic Health Surveys for 49 low-income and middle-income countries (LMICs). Data from 222 690 children aged 0–23 months were included in the analytical sample. We show pooled estimates from multiple analysis models, as well as models examining the role of the surviving parent's residence and household structure.

Results In fully adjusted models, maternal orphans were found to have an increased risk of stunting (RR 1.3, 95% CI 1.2 to 1.4) compared to children with two living parents. The stratified models demonstrated that the greatest risk was observed for maternal orphans whose surviving fathers did not live with the child (RR 1.4, 95% CI 1.3 to 1.6).

Conclusions Early orphanhood constitutes a major risk for children's physical development in LMICs. However, the results suggest that the adverse effects of a parental loss can be mitigated if appropriate support is provided, indicating the potential benefits of the surviving parents staying with their children. The benefits of targeted social policies aimed at single fathers and grandparents supporting orphans may be significant.

  • MCH

Statistics from


Estimates indicate that up to 151 million children globally have experienced the death of one or both parents before the age of 18, with the vast majority of these children living in low-income and middle-income countries (LMICs). The relationship between parental loss and children's social, emotional and cognitive development, as well as children's mental health and school attendance has been widely documented.1–4

A potential mediator in this relationship between parental loss and poor child outcomes, is the nutritional status of children.5–9 Understanding the degree to which orphanhood may be linked with stunting in early childhood may shed light on strategies to improve short-term and long-term outcomes in this vulnerable population. In particular, the first 1000 days of life are a critical period to intervene to prevent negative health and developmental outcomes, such as stunting and impaired cognitive development.10

Although prior research has demonstrated an association between orphanhood and malnutrition,11–15 only a few studies have examined the relationship between orphanhood and linear growth, finding mixed results, with a number of studies primarily from the HIV literature.11 ,12 ,14 ,16–18 Further investigation of this association is needed since a number of factors can place orphans at risk for stunting. The home environment or living situation of the orphaned child can influence his/her nutritional outcome. For example, relatives or neighbours who take in an orphaned child may not have the economic security to ensure his/her well-being,19 particularly for grandparents whose resources may be very limited.20 ,21 In some cases child neglect can occur even when adequate resources are present depending on the context.22 ,23 Recurrent and or chronic infections left unaddressed can also contribute to stunting.24 Young children may experience emotional and behavioural problems,25 potentially resulting in poor appetite or refusal to eat.26 In the case of infants, failure to thrive may also be evident.27

Given the equivocal findings of the association between linear growth and orphanhood, as well as prior evidence of the risk factors as outlined above, the aim of the present study is to further explore the relationship between orphanhood and stunting among children 0–23 months of age. Different forms of parental loss (ie, loss of mother only, loss of father only and loss of both parents) and children's nutritional status before age two are examined. In addition, the living arrangements of children who are orphaned are considered as modifying factors for the association between orphanhood and stunting in LMICs.


Data sources

The data for this study were derived from 84 Demographic and Health Surveys (DHS)28 conducted across 49 countries between 2000 and 2012 (see online supplementary appendix S1). The DHS are nationally representative cross-sectional surveys measuring indicators of population, health and nutrition, with special emphasis on family planning and child health. The target population for this study is all children in the DHS aged 0–23 months who have had their height measured as part of the household roster. Measures have been standardised to ensure comparability across countries and time.

Study population and sample size

The study sample (for the stunting analysis) includes 222 690 children aged 0–23 months at the time of interview. Figure 1 illustrates how the sample was deduced.

Outcome measure

Stunting was the primary outcome considered in this analysis. A child is considered stunted if his/her height-for-age z-score is less than minus two SDs from the sex-specific median height-for-age of the reference population.29

Independent variables

The primary independent variable of interest is orphan status. Orphan status was categorised as four mutually exclusive groups: both parents are alive (reference group), maternal orphans (mother has passed away, but father is alive), paternal orphans (mother is alive, but father passed away) and double orphans (both parents have died).

In order to reduce the risk of confounding, age of the child in months at the time of interview, highest level of education within the household, place of residence, household wealth, head of household and gender of the child were the control variables in this analysis. The DHS household member recode (PR) file has the education level of all members of the household, and the highest attained of any household member is used in this analysis and broken down into four categories: no education or incomplete primary education (reference), completed primary education, completed secondary education and the fourth category of higher education. The DHS wealth index was used to measure household wealth. This measure is calculated as a principal component of a group of assets owned by the household.30 It is a relative measure of wealth within a country and year, not an absolute measure of wealth or income, and is divided into five categories: poorest (reference), poor, middle, rich and richest. Place of residence was defined in terms of whether the household was located in an urban (reference) or rural area as defined by the previous national census. Gender of the child was classified as either male (reference) or female.


Our data and analysis are cross-sectional in nature. Our empirical work is divided into three parts. First, we calculate descriptive statistics for the sample population and include corresponding p values of the statistical difference in prevalence within a characteristic across stunted and non-stunted children (eg, stunted male vs non-stunted male), and also the p values within the stunted group across the characteristics (eg, male stunted vs female stunted). Second, we show unadjusted and adjusted associations between orphan status and stunting. Third, we explore the effects of various household living arrangements for orphans. For children who had one parent die, whether the surviving parent lived with the child was accounted for in an interaction model. For double orphans, the regressions were stratified by head of household, whether it was the grandparent or an ‘other’ person (eg, older sibling, uncle, adoptive parent, not related).

In the regression analysis, the association of orphan status to stunting was estimated using a modified Poisson model with survey specific fixed effects along with the covariates (note, there can be multiple surveys per country over time, so country/year fixed effects would be equivalent). We then consider the role of living arrangement. For one-parent orphans whether the surviving parent lives with the child or not is interacted with the child's orphan status to examine the relative risk of living with compared to living apart from the surviving parent on stunting. We also stratify by head of household, as the child may be with a surviving parent who is not the head of household and the one deciding the allocation of resources within the household. For double orphans, the regressions are stratified by head of household. In all regressions, we accounted for the multistage sampling design and clustered the standard errors at the cluster level to account for the intracluster correlation. The empirical analyses were conducted using Stata/MP 12.0 16 core.

Ethical review

The DHS is a publically available data set distributed by Measure DHS that is administered by ICF International and funded by USAID. This study uses the recoded and de-identified data distributed by Measure DHS. A human subject’s research exemption was obtained from the Harvard T.H. Chan School of Public Health Institutional Review Board (Protocol #21213–101).


Table 1 shows demographic characteristics of the study sample. From the overall sample, male and female children were close to equal proportion, and the children in the sample were spread evenly in terms of age across the 23 months. In addition, more children lived in rural areas (66.4%) and there was a higher fraction of children from the poorest households (24.6%), while the highest level of education attained by anyone in the household was primary (39.2%) or secondary (39.7%) level.

Table 1

Sample breakdown by stunting status and covariates

In this study population, 29.4% of the children were identified as stunted—short for their age and gender. Stunting followed socioeconomic status (SES) gradients—the children living in households from the poorest wealth quintile showed a stunting prevalence of 33.6%, whereas the richest had a prevalence of 22.2%. Children living in rural areas had a higher rate of stunting (32.9%) than their urban counterparts (22.5%; p<0.01). If the household had less than a primary education, the prevalence of stunting was (37.3%) as compared to households with higher education (16.7%; p<0.01). In addition, girls (26.5%) were less likely to be stunted than boys (32.3%; p<0.01). The prevalence of stunting increases with the age of the child; 17.8% of children were stunted in the 0–5 months age range, and 43.2% of children 18–23 months were stunted.

Among 222 690 infants and children less than 24 months of age, 66 were double orphans, 484 were maternal orphans and 2533 were paternal orphans. Of the children who had both parents alive, 66% (n=144 203) lived in the household with both parents. For maternal orphans 25% (n=119) remained in the household headed by their father. However, 64% (n=310) lived in households headed by the grandparent or an ‘other’ person without their surviving father. For paternal orphans, only 5% (n=127) lived without their mother. If both parents died, children either lived in a household headed by a grandparent (52%, n=34) or an ‘other’ relative or person (48%, n=32).

For maternal orphans, 45.2% of these children were stunted compared to those with both parents alive (29.3%; p value<0.01). Paternal orphans also demonstrated a higher risk of stunting (33.4%; p value<0.01). Double orphans had a prevalence of stunting of 39.4%; however, this finding was of borderline significance when compared with children with both parents (p=0.016).

Using a Poisson model to calculate the relative risk of stunting of orphans we found that single orphans had an elevated risk of stunting compared to children whose both parents were alive in adjusted models (table 2). While the relative risk of stunting is above one for double orphans, the CIs are wide due to the low number of cases in this group. The same pattern holds when controlling for survey fixed effects. In the multiple analysis, once wealth, age of the child in months, education, location and child gender are controlled for, the significant effect of a paternal death disappears (RR 0.98, 95% CI 0.93 to 1.03). However, if the mother died and the father survived, then the child remained at an elevated risk of stunting (RR 1.3, 95% CI 1.2 to 1.4) after controlling for confounding (table 2).

Table 2

Unadjusted and adjusted relative risk of orphan status on stunting outcome

We further stratified by the household structure of the orphans, that is, head of household and if child lived with surviving parent when applicable (see table 3). For maternal orphans whose fathers do not continue to reside with them, the relative risk of stunting increases to 1.4 (95% CI 1.3 to 1.6). However, if the father lives with the child, then no elevated risk of stunting was observed as compared to children whose parents were both living (RR 1.1, 95% CI 0.9 to 1.3). For paternal orphans, the association remained null (as compared to the adjusted model in table 2), with no variation related to whether the mother lives with the child (see table 3). Among double orphans that live with their grandparent(s) there is a higher risk of stunting compared to the reference group of non-orphans who live with grandparents (RR 1.6, 95% CI 1.1 to 2.2). If the double orphan lives with an ‘other’ relative or person the risk of stunting is no greater than that of children in the reference group – both parents alive and ‘other’ relative or person is head of household (RR 0.8, 95% CI 0.5 to 1.4).

Table 3

Unadjusted and adjusted relative risk of orphanhood (interacted with father/mother in household) on stunting status, stratified by head of household


The initial findings demonstrated that the rates of stunting were much higher among maternal orphans (45%) as compared to children in which both of their parents were living (29%). A similar finding was observed for those children whose father had died, although the prevalence of stunting was lower (33%). In addition, children who had both parents die also showed a higher level of stunting (39%) relative to non-orphans, although this finding was borderline significant. Interestingly, not all of these findings were sustained in multiple analysis after controlling for confounding variables. In particular, the association of paternal orphanhood and stunting was null (RR 0.98) and not statistically significant. The magnitude of the association was also lower for double orphans, with this result remaining non-significant. However, a positive association between orphan status and stunting was sustained in the multiple analysis model for maternal orphans, indicating a 1.3-fold increase in stunting as compared to non-orphans. Thus, the death of the mother, independent of wealth of the child's household, had a significant effect on the risk of stunting. This is consistent with the fact that the mother is often the primary caregiver and responsible for the nutrition, health and welfare of the child.31

However, in stratified analysis if the father continued to live with the child after the mother died no elevated risk of stunting was observed. This suggests that the father's support could offset the detrimental effects of the mother's death on child nutritional status. However, if the father left the children and did not live with them after the mother died, the children were at a 1.4-fold risk of stunting compared to children whose both parents were alive. This may be in part related to the critical role of poor socioeconomic conditions and food insecurity and the risk of stunting.32 ,33 It may also be related to the father's role in caregiving. A study in rural Vietnam demonstrated that the father taking his child to be immunised or caring for the child when he/she was sick were protective against childhood stunting, although the latter was of marginal significance.34

For double orphans we examined the type of household where the child resided. If the child was living with his/her grandparent, the RR of stunting increased to 1.6 and was significant, in contrast with the non-stratified analysis that demonstrated a null association between being a double orphan and stunting. If the double orphan lived with an ‘other’ relative or person, then the risk of stunting was no greater than if both parents were alive and lived in a household headed by an ‘other’ person.

Overall, these findings indicate that orphan status in general may not increase the risk of malnutrition and head of household as well as whether the surviving parent lives with the child may play a protective role. Household structure may ‘tip the balance’ for increasing or decreasing the risk of stunting for the orphan. The results from the present study are consistent with a risk and resilience framework of early development, whereby certain characteristics of adversity such as orphan status may be offset in the presence of positive environmental ‘buffers’.35 In particular, the findings suggest that children who continue to live with their surviving parent after the loss of either their mother or father were able to show resilience to the potential impact on stunting among single orphans. This is in contrast with double orphans living with a grandparent; the risk of stunting increased by 40–60% if both parents died or only the mother had died.

Although this study is the first of its kind to quantify the relationships between orphanhood and stunting across a large number of LMICs, it is not without qualification. First, it is worth highlighting that the fraction of orphans in our sample was low compared to the aggregate numbers reported by UNICEF that focus on children under 20, and we limit our analysis to the first 1000 days.

In addition to considerations of generalisability as noted above, the study has several other limitations. The date of death of the parent was not recorded, and thus the length of exposure to orphanhood is not known. In addition, since the DHS is a household-based survey, information on orphans living in the street or institutions is not available. The mortality of orphans in this study population is also not known. The orphans we observe are those who survived, and were living in a household at the time of interview. Given that these orphans may have been in better living conditions than those omitted from the analysis, the regression results may underestimate the effect of orphan status on nutritional outcomes. Although we account for socioeconomic status and child characteristics in analyses, it is likely that there are other, non-independent conditions that may both underlie the association between parental loss and stunting, as well as predict the characteristics of households in which orphans reside after their loss. Survivor bias may also play a role in the varying results found in the literature—if children who are orphaned and malnourished are at a greater risk of death a bias towards the null for this association may occur.16 A further limitation of our analysis using cross-sectional data is that we cannot directly identify the casual mechanisms underlying the observed differentials. Given that parental death in general, and maternal death in particular, can be correlated with maternal health status prior to and during pregnancy, some of the observed differences in children's physical stature may be due to growth restrictions experienced in utero. Postdelivery, parental health status and mortality will not only affect parents’ ability to provide a healthy and safe environment for children, but are likely to also affect the mother's ability to breastfeed and to provide appropriate nutrition more generally.

Strengths of this analysis include the breadth of the countries and time period, as well as a large sample size to examine these relationships.

Overall, the findings indicate a critical role of the mother's survival in the event of the child's father dying on childhood stunting. The results also support the significant role of the father remaining with his children in the event the mother dies, in which case there was no increase in risk of malnutrition. Maternal orphans in the absence of the father and double orphans were particularly vulnerable in grandparent-headed households, even after controlling for wealth of the household. These findings suggest that although orphan status can increase the risk of stunting, it is not inevitable and that protective mechanisms, such as the father's presence, may reduce this risk. In the absence of the father, grandparent-headed households caring for orphans may need support from local programmes to potentially offset the risk of stunting—public health and nutrition interventions for this particularly vulnerable population appear to be warranted.

What is already known on this subject

  • Estimates indicate that up to 151 million children globally have experienced the death of one or both parents before the age of 18, with the vast majority of these children living in low-income and middle-income countries (LMICs). Parental loss in early life can compromise nutritional intake for a child, representing one of the pathways by which early life exposure to orphanhood affects later life outcomes. Despite this evidence, the degree to which parental loss affects children's nutritional outcomes early in life remains both understudied and unclear.

What this study adds

  • Overall, the findings support a critical role of the mother's survival in the event of the child's father dying on childhood stunting. The results also support the significant role of the father remaining with his children in the event the mother dies, in which case there was no increase in risk of malnutrition. In the absence of the father, grandparent-headed households caring for maternal or double orphans may need support from local programmes to potentially offset the risk of stunting—public health and nutrition interventions for this particularly vulnerable population appear to be warranted.


The data for this research were collected by The Demographic and Health Surveys Programme (, under a contract from the U.S. Agency for International Development.


View Abstract

Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.


  • Contributors JF and MSF were responsible for conception, analysis plan, analysis and write up. GF was responsible for conception and write up. DC was responsible for conception, references and write up. LCT was responsible for references and write up. JC was responsible for write up. GD and ME were responsible for conception and analysis plan. WF was responsible for conception, analysis plan and write up.

  • Funding Grand Challenges Canada through the Saving Brains Initiative.

  • Competing interests None declared.

  • Ethics approval Harvard T.H. Chan School of Public Health Institutional Review Board (Protocol #21213-101).

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

  • Data sharing statement The data are from the DHS and are available on request from Macro International.

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.