Modeling the effects of socioeconomic and environmental factors on child malnutrition and contamination risk using generalized estimating equations

Abstract

Malnutrition and environmental contamination remain critical public health problems among children under five years old, particularly in developing environments. This paper investigates the sociodemographic, environmental, and behavioral factors that influence the risk of contamination and child malnutrition using Generalized Estimating Equations (GEE) in order to adjust correlations across repeated observations. Two multivariate binary response variables were taken into consideration: SWUbinary comprised of three binary indicators of whether the child was stunted, wasted, or underweight; and RTbinary comprised of two binary indicators of contamination exposure based on the presence of Relative Light Unit (RLU) and Total Coliform Analysis (TCA) contamination in a household. Using Point-Biserial Correlation and Cramer’s V Statistic, relevant predictor variables were screened and backward stepwise selection was used to determine the best set of predictors from those remaining for each of three correlation structures: unstructured, exchangeable and independent. The best model for each of the two response variables was chosen using the Akaike Information Criterion (AIC). For SWUbinary, the exchangeable correlation model was selected, and for RTbinary, the independent correlation model was the selected model. The results show how important it is to look at both nutrition and environmental factors together when trying to improve the health and well-being of children.