Factors Associated With Child Stunting, Wasting, and Underweight in 35 Low- and Middle-Income Countries.

Key Points Question What are the most important factors associated with child undernutrition, and how do they vary across countries? Findings In this cross-sectional study of 299 353 children aged 12 to 59 months in 35 low- and middle-income countries, household socioeconomic status and parental nutritional status were the leading factors associated with child undernutrition in pooled analyses and in most country-specific analyses. Environmental conditions, health behaviors, disease prevalence, and maternal reproductive care were less frequently associated with child undernutrition, with substantial heterogeneity among countries. Meaning The findings of this study suggest that interventions to improve socioeconomic status and parental nutritional status (eg, education for women and poverty reduction) should accompany food and nutrition programs, but the potential benefits of investing in specific conditions are highly dependent on the context.


Data Source
We drew the most recent data for LMICs from DHSs conducted between 2007 and 2018. Demographic and Household Surveys are nationally representative household surveys that collect detailed nutrition and health information on children, their parents, and households 18 using a multistage, stratified sampling design. The first stage involves the division of each country in geographic areas. Within these subnational regions, populations are stratified by urban or rural area. These primary sampling units or clusters are selected with probability proportional to the contribution of that cluster's population to the total population. In the second stage of sampling, all households within the cluster are listed, and an average of 25 houses are randomly selected for an interview by equal-probability systematic sampling. 18 We excluded earlier survey rounds to avoid inconsistencies in the measurements, collection, and reporting of data required for this study. The study was reviewed by the Harvard T.H. Chan School of Public Health institutional review board and was considered exempt from full review because it was based on an anonymous, public-use data set with no identifiable information on study participants. Our study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.

Study Population and Sampling Size
A total of 35 LMICs had collected data on child anthropometric measures and the factors of interest.
The eligibility criteria for our analytic sample were as follows: children (1) who born singleton, (2) who were aged 12 to 59 months and alive at the time of the survey, (3) with a mother who was not pregnant at the time of survey, and (4) with valid measures on child stunting, underweight, and wasting. We identified 299 353 children from 35 LMICs in the final analytic sample for our primary analysis (eFigure 1 in the Supplement).

Outcomes
The following 3 anthropometric failure outcomes were constructed based on the 2006 World Health Organization child growth standards: stunting, underweight, and wasting. 19 Height-for-age z score, weight-for-age z score, and weight-for-height z score were calculated by comparing the child's measurements with the median value in the reference population of the National Center for Health Statistics International Growth Reference. 20 Stunting was defined as a height-for-age z score less than −2 standard deviations (SDs) of the median, underweight as weight-for-age z score of less than −2 SDs, and wasting as weight-for-height z score less than −2 SDs. 19 Exposures Based on the UNICEF framework, 5 its adaption in the Lancet Maternal and Child Nutrition Series, 4 and previous practices, 9,10,16 we selected 20 factors for our primary analysis and 6 additional factors on paternal characteristics and maternal autonomy for supplementary analyses. We classified these 26 factors associated with child anthropometric failures either directly or via intermediary causes. A total of 9 direct factors were identified, including child nutrition (dietary diversity score, breastfeeding initiation, vitamin A supplements, and use of iodized salt), disease occurrence (infectious disease in past 2 weeks), health behaviors (oral rehydration therapy for diarrhea, care seeking for suspected pneumonia, full vaccination), and living conditions (indoor pollution). The association between each of these direct factors and child anthropometric failures has been documented previously. 17,21,22 The remaining 17 indirect factors included household socioeconomic status (household wealth, maternal and paternal education), parents' nutritional status (maternal and paternal height and BMI), maternal autonomy (for health care, movement, and money), environmental conditions (water source, sanitation facility, and stool disposal), maternal reproductive care (antenatal care, skilled birth attendant at delivery, family planning needs), and maternal marriage age. Prior studies have indicated that household wealth, maternal characteristics, and household environment are strongly associated with child anthropometric failures. 8,23,24 Although only a few studies have investigated the role of paternal nutritional status, we included it in the supplementary analysis owing to potential biological and psychosocial channels between fathers and their offspring. 6,25 We also included maternal reproductive care variables that represent the care mothers received during pregnancy, 26 the risk the child faced during birth, 27 and the families' desired birth spacing and their capacity to reach it. 28 A detailed list and definitions of these factors are presented in Table 1. 9,[29][30][31][32][33][34][35][36]

Statistical Analysis
We assessed the association of each factor with child anthropometric outcomes by first pooling data from all countries and then separately for each country. We included sampling weight, clustering, and stratification variables provided by DHS to ensure that the estimates were representative at the national level and in pooled analyses. 16 We clustered the sample at the level of the primary sampling unit, which allows for interdependence of error terms within clusters and households. 16 In pooled analyses, we reweighted observations by a country's population size and included country fixed effects to account for the unobservable country-level factors. For both pooled and country-specific analyses, we developed 2 sets of logistic regression models for each outcome. First, we ran separate models (single-adjusted models) for each factor in which we adjusted for child's age and sex, birth order, and maternal age at birth. Second, we performed mutually adjusted models (fully adjusted models) in which all factors, as well as child's age and sex, birth order, maternal age at birth, and place of residence (urban vs rural), were considered simultaneously. Based on these models, we compared and ordered the factors according to their coefficient sizes (odds ratios [ORs]). For all factors, the best-off group was set as the reference category to ensure consistency in interpretation of ORs. For factors with multiple categories (ie, household wealth quintile), only the OR corresponding the worst-off group (ie, the poorest quintile) is presented in our results section.
We performed 6 sets of supplementary analyses. First, we included 3 additional paternal characteristics for a subset of 188 290 children from 12 countries that had collected data on fathers.
Second, we stratified children by age (<2 years and Ն2 years) given their different dietary demands. 37 Third, we performed stratified analyses by urban and rural areas. For the second and third analyses, we followed previous practice 38 and used Bonferroni correction to deal with the type I error from multiple testing. Fourth, we reestimated the fully adjusted models after removing source of drinking water, sanitation facility, and household air quality because these indicators had been considered in the construction of household wealth index in DHS. Fifth, we reran the models, adding covariates on children's birth weight and birth interval. As more than half of the children (170 451 of 299 353 [56.9%]) had missing or invalid birth weight or birth interval, only on a subset of 128 902 children was used for this supplementary analysis. Sixth, we added 3 indicators of maternal autonomy for a subset of 142 638 children (47.6%) with available data.
We used Stata version 14.2 (StataCorp) for all analyses. We adopted the MI command for multiple imputations for observations with missing value on 1 or more factors of interest. 39,40 All statistical tests were 2-tailed, and P < .05 was considered statistically significant.

Results
Of 319 566 children who met the inclusion criteria, 20 213 (9.3%) were excluded because of missing Definition Reference category Self-reported Dietary diversity score 9 In quintiles, based on a score ranging from 0 to 8, with a point assigned for consuming grains, roots and tubers, legumes and nuts, dairy products (ie, milk, yogurt, cheese), flesh foods (ie, meat, fish, poultry, and liver or organ meats), eggs, vitamin-A rich fruits, and vegetables as per the 24-hour recall of food intake in the DHS;    Abbreviations: BMI, body mass index (calculated as weight in kilograms divided by height in meters squared); LMICs, low-and middle-income countries; ORT, oral rehydration therapy. remained statistically significant ( Figure 1A). Conditional on all other factors, short maternal height had the strongest association with child stunting, with an OR of 4.7 (95% CI, 4.5-5.0; P < .001), followed by lack of maternal education (OR, 1.9; 95% CI,1.8-2.0; P < .001), poorest household wealth (OR, 1.7; 95% CI,1.6-1.8; P < .001), and low maternal BMI (OR, 1.6; 95% CI,1.6-1.7; P < .001).

Wasting
In single-adjusted models, there were 10 factors significantly associated with higher odds of wasting,  Figure 1C).

Underweight
Short maternal height was most strongly associated with higher odds of underweight (ranked 1st-4th) in 29 countries; however, it ranked 20th in Namibia. Low maternal BMI was also strongly associated with underweight across all 35 countries, ranking between 1st and 6th. The relative rankings for lack of maternal education and poorest household wealth varied largely across countries.

Wasting
Low maternal BMI ranked within the top 5 factors associated with wasting in most countries, except Comoros, Namibia, São Tomé and Príncipe, and Zambia. Short maternal height, poorest household wealth, and lack of maternal education were strongly associated with higher odds of child wasting for some countries but were found to have weaker associations in many other countries. For example,    Short maternal statue indicates maternal height of less than 145 cm; low maternal body mass index (BMI, calculated as weight in kilograms divided by height in meters squared), BMI less than 18.5; child marriage, mother younger than 18 years at marriage; delayed breastfeeding, child was not breastfed within 1 hour of birth; infectious disease, child had infectious disease within 2 weeks before survey. ANC indicates antenatal care; FP, family planning; HH, household; ORT, oral rehydration therapy; SBA, skilled birth attendant.

JAMA Network Open | Global Health
lack of maternal education ranked between 1st and 4th in 12 countries but ranked between 18th and 20th in 8 countries (eFigure 5 in the Supplement). The strength of association for each factor and child wasting also showed large variations across countries (eFigure 6 in the Supplement

Supplementary Analyses
In the first supplementary analysis with paternal height, BMI, and education, we found that paternal factors had weaker associations with child anthropometric failures compared with maternal indicators (eFigure 7 in the Supplement). Short paternal height was associated with stunting with an OR of 1.9 (95% CI, 1. A rich volume of observational studies supports our findings regarding maternal height and BMI, 8,16 but paternal anthropometry remains largely unexplored. The associations between short parental height and child anthropometric status may be attributed to both shared genetic background and common environmental determinants (eg, diet, culture, social class) that first affect parents during their early childhood and subsequently affect the growth of their offspring. 41 The consistent association between maternal BMI and child anthropometric failures may be attributed to intrauterine intergenerational transmission of low maternal BMI during pregnancy, giving infants a high risk of low birth weight and being small for gestational age, which forms the fetal origins of subsequent childhood undernutrition. 34,42 While we did not have data on maternal BMI during pregnancy, BMI at the time of the survey is likely to be associated with previous weight. The influence of maternal BMI on child anthropometric status attenuated only moderately after adding paternal BMI. 34 Our pooled estimates on household wealth and maternal education were comparable with previous multicountry studies. 7,43 Across countries, household wealth had moderate heterogeneity in associations with child stunting and underweight. The relative importance of maternal education Inconsistent findings on the association of oral rehydration therapy for diarrhea with outcomes may be because of the differential prevalence of children very close to the anthropometric failure cutoffs given that only they would be substantially affected by the occurrence of diarrhea and oral rehydration treatment. 49,50 The heterogeneous association between sanitation facility and child undernutrition may be attributed to differences in complementarity of toilet maintenance, including other water and hygiene practices. 51,52

Limitations
There are several limitations to this study. First, factors in the fully adjusted models may be associated with each other and serve as confounders or mediators. Multicollinearity can increase the standard errors of the coefficients and weaken the significance levels, but it does not result in biased estimates. Moreover, the low VIF for all factors presented in the supplementary analysis section indicated low multicollinearity. Second, the use of observational data and cross-sectional analysis limit our capacity to make any causal inferences. Third, some factors analyzed in this study, such as breastfeeding history, care-seeking behavior, and disease history, were self-reported and, therefore, are prone to potential measurement errors.

Conclusions
This systematic investigation of the comparative importance of direct and indirect factors associated with child anthropometric failures suggests the universal importance of improving maternal