Abstract

Child stunting, defined as impaired height for age, is a major indicator of severe undernutrition and is more prevalent in Sub-Saharan Africa. Individual child stunting risk factors for childhood stunting are well-studied and known. This study aimed at assessing the viability of combining individual child stunting risk factors into a simple risk factor prediction model that could be used to predict stunting among children aged 5 years or lower. Firstly, a systematic review of risk factors for childhood stunting was conducted. Secondly, using stunting data on nearly 5,000 children aged 5 years or below in the Malawi Demographic Health Survey (MDHS 2015-16) we identified risk factors that were used in the primary multivariate logistics model for child stunting. Thirdly, several reduced models were then obtained depending on the variable selection algorithm that included backward, forward, stepwise, random forest, Least Absolute Shrinkage and Selection Operator (LASSO), and own subjective judgment. Finally, from each reduced multivariable logistic model, a stunting risk score, based on its coefficients, was calculated for each child. The stunting risk prediction models were assessed using discrimination measures including area under-receiver operator curve (AUROC), sensitivity and pecificity. The systematic review produced 68 predictor variables of child stunting, of which 67 were available from the 2016 MDHS dataset, and 27 had complete information. The common risk factors selected by all the variable selection methods include household wealth index, age of the child, household size, type of birth (singleton/multiple births), and birth weight. The best cut-off point on the child stunting risk prediction model was 0.37. The best predictive model was based on risk factors determined by the judgment method, which had AUROC 64% (95% CI: 60%-67%) in the test data. For children residing in urban areas, the AUROC was 67% (95% CI: 58-76%) as opposed to those in rural areas, AUROC =63% (95% CI: 59-67%). The derived child stunting risk prediction model could be useful as a first screening tool to identify children more likely to be at risk of stunting. The identified children could then receive necessary nutritional interventions.

More details

School	: School of Natural and Applied Sciences
Issued Date	: 2023

Download full document