This study analyzed nationally representative Demographic and Health Survey (DHS) data from five South Asian countries (Bangladesh, India, Pakistan, Nepal, and the Maldives). The DHS uses standardized, cross-sectional, multistage stratified cluster sampling to collect mother and child health indicators. The analysis focused on children aged 0–59 months and employed an observational design using the most recent DHS rounds for each country.
Researchers downloaded the most recent DHS datasets after approval. The initial pooled dataset contained 262,890 child records. Records without valid birth weight (marked as not weighted or “Don’t know”) were removed, leaving 220,499 with valid birth-weight information. Further restriction to children with available height-for-age, weight-for-height, and weight-for-age produced 191,996 records. After excluding cases with missing values in covariates used for matching and adjustment, the final analytic sample comprised 147,332 children. A flowchart in the source documents the exclusion process.
Nutritional outcomes were constructed using WHO growth standards and anthropometry collected at survey time. Three primary indices were used: height-for-age (stunting), weight-for-height (wasting), and weight-for-age (underweight). Each outcome was binary, defined by a z-score ≤ −2 SD versus > −2 SD. AOFOCM (at least one form of child malnutrition) was coded ‘Yes’ if a child had any of the three indicators. MCFOCM (multiple concurrent forms of child malnutrition) was coded ‘Yes’ if two or more indicators were present.
The exposure of interest was low birth weight (LBW) defined by WHO as birth weight <2,500 grams. Birth weight in DHS is recorded from health cards when available or by maternal recall. Covariates included country, maternal age, maternal education, household wealth index, child sex, antenatal care visits (ANC), iron tablet/syrup intake, caesarean delivery, urban/rural residence, maternal media exposure, maternal BMI (Asian categories), birth order, history of terminated pregnancy, and early initiation of breastfeeding (EIBF).
Because the magnitude of associations may vary across settings, the study evaluated heterogeneous effects by estimating associations separately for Bangladesh, India, Pakistan, Nepal, and the Maldives. Exact matching on country was applied during propensity score matching to ensure within-country comparisons.
Initial descriptive statistics and chi-square tests examined bivariate associations between LBW, outcomes, and covariates in the unmatched data to identify potential confounding. To reduce confounding bias inherent to observational data, the authors applied one-to-one nearest neighbor propensity score matching (PSM) without replacement. Propensity scores were estimated via logistic regression predicting LBW from the covariates. No caliper or common support restriction was imposed, but exact matching on country was performed due to the pooled multi-country dataset.
After matching, average treatment effect on the treated (ATT) estimates were obtained in the matched sample using three approaches: Linear Probability Model (LPM), logistic regression, and weighted least squares (WLS). Post-matching regression models adjusted for all covariates as an additional step to address any residual imbalance. Both adjusted and unadjusted analyses were reported within the matched sample.
Matching quality was assessed by standardized mean differences (SMD) for covariates, where SMD < 0.1 indicates acceptable balance. Additional diagnostics included comparison of pseudo values and likelihood ratio test statistics for logistic models fitted to unmatched and matched datasets, and visual inspection of propensity score distributions for overlap between treated and control groups.
The final analytic sample included 147,332 children. The pooled LBW prevalence was reported as 16.23% across the analytic sample. Country-specific LBW prevalences in the final sample were: Bangladesh 13.67%, India 16.32%, Pakistan 18.97%, Nepal 11.94%, and Maldives 12.86%. The majority of participants in the pooled sample were from India (96.69%), with smaller proportions from Bangladesh, Maldives, Nepal, and Pakistan.
Regression analyses applied to the propensity-score matched samples showed statistically significant associations between low birth weight and all five malnutrition outcomes (stunting, wasting, underweight, AOFOCM, and MCFOCM) in the pooled analysis of matched data. The study used multiple modeling approaches (LPM, logistic regression, WLS) in the matched sample and presented both adjusted and unadjusted estimates to assess robustness.
Subgroup analyses examined the association between LBW and each malnutrition outcome within each country. Across the five countries, significant associations were observed for most outcomes. An exception was noted for wasting, where the association with LBW was not statistically significant in Bangladesh and the Maldives in the subgroup analyses. Full country-specific effect estimates and statistical details are reported in the source article.
The authors interpret the findings as evidence that low birth weight is an important determinant of multiple forms of child malnutrition in South Asia and that interventions addressing LBW may help reduce child malnutrition. Strengths of the work include the use of large, nationally representative DHS datasets from five countries and application of propensity score matching to mitigate confounding bias common to observational studies.
Limitations noted in the source include the cross-sectional design of DHS (precluding causal inference), reliance on retrospectively reported birth weight when health card data are unavailable, and the decision not to incorporate DHS sampling weights, clustering, or stratification in the analyses. The authors also highlight cross-country variability as a consideration for context-specific policy responses.
Using propensity score matched analyses of DHS data from five South Asian countries, the study found that low birth weight is associated with higher risk of stunting, underweight, AOFOCM, and MCFOCM in pooled analyses, with heterogeneity in the association with wasting across countries. The authors recommend context-specific interventions targeting determinants of LBW to address child malnutrition in South Asia, while acknowledging the observational nature of the data and between-country differences.