Bayesian mixed effect models to account for environmental modulators of acute malnutrition treatment in children

Abstract Acute child malnutrition is not only a global public health problem influenced not only by very diverse factors, including socioeconomic and dietary aspects, but also by seasonal and geographic factors. The present study is a secondary analysis that attempts to characterize which variables have influenced the Middle Upper-Arm Circumference (MUAC) upon admission and the Length of Stay (LOS) for treatment recovery. The sample of children analysed was 852. Initially, data cleaning and a reduction of the dimensionality of dietary diversity were carried out. A selection of the importance of the variables using the Watanabe Akaike Information Criteria (WAIC) was carried out prior to the adjustment of Bayesian mixed effects models, with the variables of travel time to health site and week of admission as random factors, on the MUAC and LOS variables. Clear differences were seen between both contexts, highlighting significant interactions of travel time in Niger while the seasonal effect stood out in Mali. The MUAC models identified a positive effect of age in both contexts, and in Niger, influences of diet diversity, comorbidities, breastfeeding and vaccination appeared. On the other hand, the LOS models highlighted the severity upon admission, and, in Niger, also factors related to the treatment protocol and the distance to the water source, while in Mali, the quality of water was more decisive. The present study shows the importance of considering acute child malnutrition from a multidimensional and complex approach, where diverse factors (biological, socioeconomic, ecological, etc.) can influence directly or as modulators of the disease and its treatment.