---
title: "Low birth weight and child malnutrition in South Asia: propensity score matched analysis"
id: "plos-one-7-exploring-the-association-between-low-birth-weight-and-different-forms-of-child"
canonical_url: "https://medichelpline.com/clinical-feed/plos-one-7-exploring-the-association-between-low-birth-weight-and-different-forms-of-child"
content_type: "clinical_feed_article"
specialty: "Pediatrics"
source_name: "PLOS ONE (Medicine)"
source_url: "https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0358877"
published_at: "2026-09-21T14:00:00.000Z"
evidence_level: "Journal Feed"
license: "CC-BY-NC-4.0 / Informational Use"
---
# Low birth weight and child malnutrition in South Asia: propensity score matched analysis
## Provenance & Clinical Metadata
- **Canonical URL:** https://medichelpline.com/clinical-feed/plos-one-7-exploring-the-association-between-low-birth-weight-and-different-forms-of-child
- **Specialty:** [Pediatrics](https://medichelpline.com/clinical-feed/pediatrics.md)
- **Primary Source:** PLOS ONE (Medicine)
- **Source URL:** [Original Journal Publication](https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0358877)
- **Published At:** 2026-09-21T14:00:00.000Z
- **Evidence Rating:** Journal Feed
## Executive GIST (TL;DR)
- This multi-country study used Demographic and Health Survey data from Bangladesh, India, Pakistan, Nepal, and the Maldives to examine associations between **low birth weight (LBW)** and five measures of child malnutrition: **stunting**, **wasting**, **underweight**, at least one form of child malnutrition (AOFOCM), and multiple concurrent forms of child malnutrition (MCFOCM). - The analytic sample included 147,332 children with complete data after exclusions for missing birth weight and other covariates; overall LBW prevalence in this sample was 16.23% with country-specific prevalences reported. - Country prevalences of LBW in the final sample were: Bangladesh 13.67%, India 16.32%, Pakistan 18.97%, Nepal 11.94%, Maldives 12.86%. - LBW was defined per WHO as birth weight <2,500 g using recorded birth weight or maternal recall in DHS surveys. - Outcomes were constructed from WHO growth standards using height-for-age, weight-for-height, and weight-for-age z-scores with binary cutoffs at −2 SD to define stunting, wasting, and underweight; AOFOCM and MCFOCM combined these indicators. - The study applied one-to-one nearest neighbor **propensity score matching (PSM)** with exact matching on country to reduce confounding; propensity scores were estimated via logistic regression and matching was done without replacement. - Matching quality was evaluated using standardized mean differences (SMD), pseudo values and LR tests from logistic models, and visual inspection of propensity score overlap; SMD < 0.1 was used as the threshold for good balance. - Average treatment effect on the treated (ATT) estimates were obtained in the matched samples using Linear Probability Models, logistic regression, and weighted least squares; both adjusted and unadjusted post-matching analyses were performed. - Regression analyses of matched samples found statistically significant associations between LBW and all five malnutrition outcomes overall; subgroup analyses by country showed significant associations across countries except for wasting in Bangladesh and the Maldives where associations were not statistically significant. - The authors emphasize context-specific interventions targeting **LBW** to reduce child malnutrition in South Asia while noting the observational, cross-sectional nature of DHS data and cross-country variability. - Data sources were DHS standard surveys: Bangladesh 2022, Nepal 2022, India 2019–21, Pakistan 2017–18, Maldives 2016–17. The final analytic dataset excluded cases with missing birth weight, anthropometry, or covariates. - The study did not incorporate DHS sampling weights, clustering, or stratification in analyses and received no specific funding; authors declared no competing interests.
## Clinical Analysis & Structured Key Points
Exploring the association between low birth weight and different forms of child malnutrition: A multi-country propensity score matching analysis in South Asia | PLOS One Browse Subject Areas ? Click through the PLOS taxonomy to find articles in your field. For more information about PLOS Subject Areas, click here . Article Authors Metrics Comments Media Coverage Reader Comments Figures Figures Abstract This study investigated the association between low birth weight (LBW) and various forms of child malnutrition across five South Asian countries (Bangladesh, India, Pakistan, Nepal, and the Maldives). Data from the most recent demographic and health surveys were analyzed. Propensity score matching was applied to evaluate the association between LBW and five forms of child malnutrition: stunting, wasting, underweight, at least one form of child malnutrition (AOFOCM), and multiple concurrent forms of child malnutrition (MCFOCM). Subgroup analyses were performed to assess variability in effects across the countries studied. Analysis of 147,332 complete cases revealed an overall LBW prevalence of 16.23%, with country-specific prevalences of 13.67% (Bangladesh), 16.32% (India), 18.97% (Pakistan), 11.94% (Nepal), and 12.86% (Maldives). Regression analyses of matched samples revealed significant associations between LBW and all five forms of malnutrition. Subgroup analyses also revealed a significant association between LBW and all five forms of malnutrition across the five countries, with the exception of wasting, for which no statistically significant association was observed in Bangladesh and the Maldives. These findings emphasize the importance of context-specific interventions targeting LBW to reduce child malnutrition in South Asia, while taking the observational design and cross-country variability into consideration. Citation: Hussain MP, Rahman MM, Kader MR, Talukder MRI, Khan MTF, Ahmmed F (2026) Exploring the association between low birth weight and different forms of child malnutrition: A multi-country propensity score matching analysis in South Asia. PLoS One 21(9): e0358877. https://doi.org/10.1371/journal.pone.0358877 Editor: Md. Moyazzem Hossain, Jahangirnagar University, BANGLADESH Received: October 14, 2025; Accepted: September 7, 2026; Published: September 21, 2026 Copyright: © 2026 Hussain et al. This is an open access article distributed under the terms of the Creative Commons Attribution License , which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Data Availability: All the data were downloaded from the DHS website ( https://dhsprogram.com/data/available-datasets.cfm ) after authorization was received on the data request. The corresponding survey names for the datasets from the five countries used in this study are Bangladesh 2022, Nepal 2022, India 2019–21, Pakistan 2017–18, and Maldives 2016–17. All five datasets are classified as Standard DHS surveys. Funding: The author(s) received no specific funding for this work. Competing interests: The authors have declared that no competing interests exist. Introduction A healthy child is indispensable for building a future generation that is productive, efficient, and able to positively contribute to society and global development [ 1 , 2 ]. However, the different forms of malnutrition, including stunting (too short for age), wasting (too thin for height), overweight, and underweight, pose a significant threat to child health and may even lead to mortality [ 3 – 5 ]. In 2024, child malnutrition increased for the sixth year in a row in the world’s most vulnerable regions [ 6 ]. According to the 2025 edition of the UNICEF/WHO/World Bank Group Joint Malnutrition Estimate, in 2024, 150.2 million children under five years of age were affected due to stunting, 42.8 million suffered from wasting, and 35.5 million were overweight [ 7 ]. These figures revealed that the 2025 World Health Assembly (WHA) global nutrition targets and the 2030 Sustainable Development Goal (SDG) 2 targets were not on track to be met worldwide [ 7 ]. The scenario is particularly concerning in South Asia [ 8 – 10 ], which has the highest wasting prevalence of any sub-regions globally and accounts for half of all children affected by this condition [ 8 ]. South Asia also has one of the highest proportions of children affected by stunting, alongside some African regions [ 11 ]. It is therefore very crucial for South Asia to address the factors associated with malnutrition in the region. Prior research has identified numerous socioeconomic factors and child health-related issues that may pose a risk for malnutrition in children under the age of five. Among them, some notable factors include maternal nutritional status, maternal education, wealth index, mother’s age at birth, paternal occupation, age of the child, antenatal care, short birth spacing, birth order in the family, etc. [ 12 – 20 ]. Apart from these, low birth weight is indicated as an important determinant of malnutrition in numerous studies [ 21 – 25 ]. South Asia has a higher percentage of low birth weight compared to the other regions, accounting for close to half of all low birthweight newborns in 2020 [ 26 ]. Therefore, assessing the association between low birth weight and malnutrition in South Asian countries has been a key interest for researchers aiming to improve child health and development [ 21 , 23 , 27 , 28 ]. Low birth weight is considered a useful indicator not only for child malnutrition but also for poverty, maternal health, and healthcare delivery [ 29 ]. According to the World Health Organization (WHO), low birth weight is officially defined as birth weight of less than 2500 grams or approximately 5.5 pounds [ 30 ]. Compared to children with normal birth weight, those born with low birth weight experience increased rates of subnormal growth, health issues, and poorer neurodevelopmental outcomes. Moreover, the likelihood of adverse outcomes rises as birth weight falls [ 31 ]. In 2020, around 20 million babies were born globally with low birth weight [ 32 , 33 ], and more than 40% of these births occurred in South Asian countries [ 26 , 33 ]. Low birthweight affects a large percentage of all live births in South Asia, which is one of the highest rates in the world [ 26 , 33 ], making it a particularly concerning issue in the countries of South Asia. More specifically, across India, Pakistan, and Bangladesh, low birth weight is consistently linked to a higher risk of childhood malnutrition, with increased odds of stunting, wasting, and underweight conditions [ 21 , 23 , 27 ]. It is important to note that there are many factors associated with malnutrition that also may affect the weight of children at birth, especially those related to social and maternal health. For example, mothers’ education, mothers’ age at birth, wealth index, and antenatal care (ANC) are also mentioned as potential predictors of low birth weight in previous studies [ 34 – 38 ]. Considering these issues, in this study, our motivation is to observe the association between low birth weight and malnutrition in children under five years of age. Previous studies have employed various statistical models and techniques to identify the significant determinants of child malnutrition across different regions. The most common and widely used method for this type of research is the multiple binary logistic regression model, while some studies also used the ordinal logistic regression model [ 12 , 24 , 39 – 42 ]. One of the limitations of using these methods in observational studies is that they are often susceptible to confounding bias, which arises when a risk factor for the outcome also affects the treatment of interest [ 43 ]. To overcome this problem, propensity score matching (PSM) can be utilized as an alternative method [ 20 , 44 , 45 ]. The PSM assesses the association of the treatment variable with the outcome [ 20 ] by balancing the covariates between treated and untreated groups. Our study addresses that issue. We utilize the propensity score matching approach in malnutrition research by examining the association between LBW and various forms of child malnutrition across five South Asian countries to minimize confounding bias. In addition, this study investigated the association between LBW and five different forms of child malnutrition in the literature utilizing five nationally representative data sets from five South Asian countries. Unlike most existing studies, which are typically limited to a single country or a narrower set of outcomes, our analysis provides a broader and more comprehensive assessment while minimizing confounding bias. Methods Study design and participants This study used data from the most recent Demographic and Health Survey (DHS) conducted in five South Asian countries, including Bangladesh, Nepal, India, Pakistan, and Maldives. The DHS are nationwide household surveys that use standardized questionnaires to collect data on a wide range of health and nutrition indicators, with a primary focus on mother and child health [ 46 ]. The DHS program employed an identical survey design across all countries, consisting of a cross-sectional approach and a multistage stratified cluster random sampling strategy, ensuring that the data were nationally representative of each country’s population [ 46 , 47 ]. For further detailed information regarding the survey methods, we refer to the DHS reports for the countries under the study [ 48 – 52 ]. Data extraction and preparation The most recent DHS datasets from five South Asian countries, including Bangladesh (BDHS 2022), Nepal (NDHS 2022), India (IDHS 2019-21), Pakistan (PDHS 2017-18), and Maldives (MDHS 2016-17) were extracted from the DHS program’s official database following approval of a data-use request [ 53 ]. The initial combined datasets across these countries included health information for 262,890 children. We removed cases that did not contain children’s birth weight data, including those marked as ‘Not weighted’ or ‘Don’t know’, as well as missing entries. After removing these, we were left with 220,499 children’s records with valid low birth weight information. From this subset, we focused on additional health information such as height-for-age, weight-for-age, and weight-for-height, which were available for 191,996 children. After further excluding cases with missing values in various covariates, the final dataset had a total sample of 147,332 children. A detailed overview of the exclusion process is presented in the flowchart in Fig 1 . Download: PNG larger image TIFF original image Fig 1. Flow chart explaining how the sample was extracted for final analysis. https://doi.org/10.1371/journal.pone.0358877.g001 Outcome variables This study assessed nutritional status using three anthropometric measurements: height-for-age, weight-for-height, and weight-for-age, based on World Health Organization (WHO) growth standards [ 54 ]. These anthropometric measurements were collected at the time of the survey among children under five years of age (0–59 months). Using these indices, we constructed five binary response variables: stunting, wasting, underweight, AOFOCM (at least one form of child malnutrition), and MCFOCM (multiple concurrent forms of child malnutrition). The height-for-age index measures a child’s height compared to what is expected for their age, according to the standard growth chart [ 54 ]. Children whose height-for-age was more than 2 standard deviations (SD) below the median of the reference population were categorized as “stunted” (Yes/No). Similarly, wasting and underweight were determined using weight-for-height and weight-for-age indices, respectively, with binary categorization (Yes/No) based on the −2 SD cutoff [ 54 ]. Finally, the outcome variable AOFOCM was defined as ‘Yes’ if a child exhibited at least one positive indicator (stunting, wasting, or underweight), and ‘No’ otherwise [ 55 ]. In contrast, MCFOCM was defined as ‘Yes’ if a child exhibited at least two of the aforementioned malnutrition indicators, and ‘No’ otherwise [ 55 ]. Treatment and covariates Low birth weight (LBW) was used as the treatment variable. Following WHO guidelines, LBW was defined as a binary variable, with children weighing less than 2,500 grams classified as ‘Yes’ (abnormal weight) and those weighing 2,500 grams or more classified as ‘No’ (normal weight) [ 30 ]. In the DHS surveys, birth weight is obtained retrospectively, based on either recorded birth weight (from health cards) or maternal recall at the time of the survey. To control potential confounding effects, several other variables were also included in the analyses. These variables include country (Bangladesh, India, Maldives, Nepal, Pakistan), mother’s age ( =35), mother’s education (No education, Primary, Secondary, Higher), wealth Index (Poorest, Poor, Middle, Rich, Richest), sex of child (Male, Female), ANC ( =4), intake of iron tablet/syrup (No, Yes), C-section (caesarean section) (No, Yes), place of residence (Urban, Rural), exposed to media (No, Yes), BMI (body mass index) (Underweight, Normal, Overweight, Obese), birth order (1–2, 3–4, Higher), terminated pregnancy (No, Yes) and EIBF (early breast feeding) (No, Yes). The variable “exposure to media” was determined by the mother’s engagement with watching television, reading newspapers or magazines, or listening to the radio. Mothers who reported exposure to at least one of these media sources were classified as exposed to media [ 20 ]. As noted by one of the reviewers, it is important to elaborate on the relevance of this covariate. Maternal exposure to media is particularly relevant in this context because previous studies have shown that it is associated with a reduced risk of child malnutrition, including stunting, wasting, and underweight [ 56 – 58 ]. Moreover, exposure to media has been associated with greater maternal healthcare utilization in the south Asian countries after adjusting for several socioeconomic factors [ 59 ]. BMI was categorized according to the recommended measures for Asian people: underweight (<18.5 kg/m 2 ), normal (18.5 kg/m 2 to < 23 kg/m 2 ), overweight (23 kg/m 2 to < 27.5 kg/m 2 ), and obese (≥ 27.5 kg/m 2 ) [ 60 ]. Variable to test heterogeneous effects Although the primary objective of this study was to assess the association between LBW and various forms of child malnutrition, the magnitude of this association may vary across countries, an aspect known as heterogeneous effects. To assess this, we examined whether the association between LBW and each form of child malnutrition differed across five national subgroups: Bangladesh, Nepal, India, Pakistan, and the Maldives. Statistical analysis To evaluate the association between the treatment variable (LBW) and the outcome measures (various forms of child malnutrition), a comprehensive set of statistical methods was employed. Descriptive statistics, including frequencies and percentages, were used to summarize the sample characteristics. Chi-square tests were initially performed to examine the bivariate association between LBW and the outcome variables in the unmatched dataset. Additionally, chi-square tests were also conducted between LBW and each of the covariates to assess the relationship between them and identify any potential confounding effects of the covariates on the treated variable. Significant associations in these comparisons imply the evidence of confounding, which justifies the need for matching in order to reduce confounding bias and estimate the association of LBW with child malnutrition outcomes. All statistical analyses were performed without incorporating the DHS sampling weights, clustering, or stratification. To control confounding effects, we applied a one-to-one propensity score matching (PSM) approach and constructed a matched sample of treated (LBW) and control (non-LBW) children. PSM aims to reduce confounding bias by balancing the baseline characteristics between the treated and control groups, so that valid conclusions can be made regarding the impact of a treatment on the outcome measures [ 61 ]. To estimate the propensity scores, we utilized logistic regression as the distance function, which is the most commonly used approach [ 62 ]. The model produces an estimated propensity score (PS) for each individual, representing the predicted probability of being treated given their observed covariates [ 63 ]. Once the PSs are calculated, the matched pairs between treated and control subjects were formed using their PSs. The most popular methods for forming matched pairs based on propensity scores are: nearest neighborhood matching (NNM) and optimal matching (OM) [ 62 , 64 ]. However, Gu and Rosenbaum [ 63 ] showed that NNM outperformed OM in terms of balancing covariates, and therefore, we employed the NNM algorithm without replacement to match treated individuals with control individuals based on their PSs. NNM selects patients one by one from the treated group. For each treated subject, it finds an observation in the control group whose PS is closest to that treated subject, and includes the observation in the matched sample [ 62 ]. No caliper or common support restriction was applied. Given the multi-country nature of the data, exact matching on country was performed to ensure that comparisons were made within the same country. Once the matched sample is found, we assessed the quality of matching using three approaches. First, we assessed the standardized mean difference (SMD) for all covariates between the matched treatment and control groups, where SMDs less than 0.1 indicate a good matching [ 62 ]. Second, we compared the pseudo values and likelihood ratio (LR) test statistics from logistic regression models fitted to the unmatched and matched datasets. The lower pseudo value and the insignificant goodness of fit test suggest evidence of good covariate balance [ 65 ]. Finally, we also examined the distribution of PSs between treated and control groups to visually assess the effectiveness of the matching process. The more closely the two distributions overlap, the more similar characteristics the treated and control groups have, which is indicative of more effective matching [ 20 ]. After ensuring the data were successfully matched, we estimated the average treatment effect on the treated (ATT) for each outcome variable using three modeling approaches: Linear Probability Model (LPM), Logistic Regression, and Weighted Least Squares (WLS) [ 61 , 66 , 67 ]. Each model was applied to the matched dataset while adjusting for all covariates to obtain robust estimates of the treatment effect. The post-matching regression models were used as an additional adjustment step to account for any small residual imbalance after matching and to improve the precision of the estimated association within the matched sample. In addition, we conducted unadjusted analyses within the matched sample and checked if the results differ from the adjusted analysis. We also considered the possibility of heterogeneous effects for different countries. Therefore, we performed subgroup analyses by country and estimated association between LBW and all forms of child malnutrition for different countries. Data management and curation were performed using Stata, while the propensity score matching was implemented using the R package “MatchIt” [ 68 ]. Results Background characteristics of the respondents Frequency and percentage distributions of the sociodemographic or potential confounding variables, treatment, and outcome variables are reported in the supplementary Table S1 in S1 File . Among the 147,332 participants, the majority were from India (96.69%), with smaller proportions from Bangladesh (1.04%), Maldives (1.12%), Nepal (0.78%), and Pakistan (0.37%).
## Related Clinical Research

- [SEROCoV-KIDS cohort: impact of COVID-19 on children and adolescents in Geneva, Switzerland](https://medichelpline.com/clinical-feed/bmj-open-21-cohort-profile-the-serocov-kids-population-based-cohort-study-and-biobank-on.md)
- [Pediatric Hospital Medicine FTEs and Staffing Models: Challenges and Gaps](https://medichelpline.com/clinical-feed/pubmed-42764174.md) (DOI: 10.1542/hpeds.2026-009408)
- [Time‑Varying Predictors for Clinical Prediction in Pediatric Hospital Medicine](https://medichelpline.com/clinical-feed/pubmed-42764170.md) (DOI: 10.1542/hpeds.2026-009395)
- [Physical activity, sedentary behavior, and sleep after childhood cancer: device-based findings fro](https://medichelpline.com/clinical-feed/medrxiv-2-physical-activity-sedentary-behavior-and-sleep-in-everyday-life-after-childhood.md)
- [Airway microbiome–metabolome signature linked to pediatric severe asthma](https://medichelpline.com/clinical-feed/plos-one-11-a-microbiome-metabolome-signature-associated-with-pediatric-severe-asthma.md)

## Navigation
- [← Back to Pediatrics Feed](https://medichelpline.com/clinical-feed/pediatrics.md)
- [← All Clinical Specialties](https://medichelpline.com/clinical-feed.md)
## Medical & Regulatory Disclaimer

> [!CAUTION]
> MedicHelpline content is structured for research, educational, and professional discovery purposes. It does not constitute individual medical advice, clinical diagnosis, or treatment recommendations.
> Always verify dosing, contraindications, and regulatory alerts against official product labeling and primary regulatory sources before clinical decision-making.