Peer-reviewed article
Modeling and mapping under-nutrition among under-five children in Ethiopia: a Bayesian spatial analysis.
Habtewold FG, Arero BG ·
Summary
Written by AI, not yet AICS's view · claude-haiku-4-5-20251001This study used advanced statistical modeling to map under-nutrition among children under five across Ethiopia using data from the 2019 Demographic and Health Survey. The research identified key risk factors including low maternal education, poverty, and poor access to water and sanitation, while protective factors included breastfeeding and higher family wealth. The analysis revealed significant geographic variation, highlighting high-risk regions like Somali, Afar, and parts of Oromia that need targeted support.
Key points
- Lower maternal education, poorer wealth status, and inadequate access to clean water and sanitation were associated with increased under-nutrition cases
- Improved breastfeeding practices and higher maternal education were associated with decreased under-nutrition cases
- Regional disparities were substantial, with Somali, Afar, and parts of Oromia identified as high-risk areas
- Cooking practices and maternal age also significantly influenced under-nutrition outcomes
- The Conditional Autoregressive Besag-York-Mollié model provided the best fit for spatial heterogeneity in the data
Limitations
This is an observational study analyzing existing survey data, so it cannot establish causal relationships or test interventions. The analysis identifies associations and geographic patterns but does not evaluate the effectiveness of specific programmatic responses to address under-nutrition.
Abstract
Malnutrition remains a critical global challenge, characterized by an imbalance between nutrient requirements and consumption. Under-nutrition, a specific form of malnutrition, results from inadequate intake of essential nutrients and has severe implications for young children, especially in developing countries. This study aims to model under-nutrition cases among children under five in Ethiopia, utilizing Bayesian spatial models to identify effective interventions. Four models were considered: Generalized Linear Model (GLM), Generalized Linear Mixed Models (GLMM), Intrinsic Conditional Autoregressive (ICAR), and Conditional Autoregressive Besag-York-Mollié (CAR BYM) with negative binomial distribution. The rationale for employing multiple models stems from the need to compare performance and accuracy in capturing spatial heterogeneity. The data were obtained from the Ethiopian Demographic and Health Survey 2019. The parameter estimation was carried out using Bayesian Markov Chain Monte Carlo (MCMC) through the brms package in R, which interfaces with Stan for efficient sampling. The models were evaluated based on the Watanabe Akaike Information Criterion (WAIC) and Leave-One-Out (LOO) cross-validation, with CAR BYM emerging as the best-fitting model. Spatial modeling revealed that maternal age, breastfeeding practices, access to clean water and sanitation facilities, cooking practices, maternal education, and wealth status significantly influence the number of under-nutrition cases among children under five in Ethiopia. Specifically, lower maternal education, poorer wealth status, and inadequate access to clean water and sanitation were associated with an increased number of under-nutrition cases, while improved breastfeeding practices, rich wealth status and higher maternal education were associated with decreased number of cases. Regional disparities also played a significant role, with the CAR BYM model effectively identifying high-risk regions such as Somali, Afar, and parts of Oromia, identified as areas requiring targeted intervention.