Socioeconomic Risk Factors Associated With Acute Malnutrition Severity Among Under‐Five Children Based on a Machine Learning Approach: The Case of Rural Emergency Contexts in Niger and Mali
2025
Luis Javier Sánchez‐Martínez | Pilar Charle‐Cuéllar | Abdias Ogobara Dougnon | Fanta Toure | Antonio Vargas | Candela Lucía Hernández | Noemí López‐Ejeda
ABSTRACT Currently, child acute malnutrition continues to be a serious public health problem, and although its most fatal consequences are well known, its associated factors still need to be studied in more depth in different contexts. The objective of the present study is to determine the association between socioeconomic variables and acute malnutrition severity in rural emergency contexts of Niger and Mali. The present study consists of a secondary analysis of controlled trials. Data related to a total of 1447 treated children (6–59 months of age) were considered, for whom the Variable Selection Using Random Forests (VSURF) algorithm was applied to create interpretation and prediction random forest models (considering 86 variables). In Mali and Niger, the prediction models agree in pointing out aspects related to the water source and the work activity of caregivers as some of the main risk factors for developing severe acute malnutrition. However, the interpretation models highlight important heterogeneity, with the distance to the health center being the greatest exponent of this situation, being the most important factor in Niger while disappearing in Mali. The prediction accuracy in the interpretation model was 68.0% in Niger and 79.80% in Mali, while the prediction model reached similar rates of 63.17% and 75.63%, respectively. Machine learning techniques have proven to be a valid tool to interpret and predict the degree of severity of acute malnutrition based on socioeconomic characteristics, including complex interrelationships. The results obtained point out different aspects to be addressed to prevent and minimize the effects of acute malnutrition.
显示更多 [+] 显示较少 [-]