Analysis of determinants of malnutrition among women in Nigeria using stability selection algorithm with Bayesian distributional additive model

Malnutrition remains a pressing public health issue, with a prevalence of over 30% among women causing severe consequences on their health, households and national development. The many types of malnutrition that affect adults are typically evaluated using the body mass index (BMI), a technique that is frequently used to determine whether a person is overweight or underweight for their height. Using data from the Women’s Individual Recode dataset of the Nigerian Demographic and Health Surveys (NDHS) conducted in 2003, 2008, 2013, and 2018, comprising 90,563 non-pregnant women aged 15–49 years, we explored a comprehensive range of malnutrition determinants, including socio-demographic factors, dietary patterns, healthcare access, and environmental conditions. With these derived data, we implemented stability selection, a technique that allows for subsampling in high dimensional selection algorithm, to identify relevant predictors amidst a large pool of potential variables. Additionally, we incorporated a Bayesian additive model for location, shape, and scale, allowing for the exploration of spatial and contextual effects, enabling a nuanced understanding of how geographic location and social determinants intersect to influence women’s nutritional status. Of the variables retained by the stability selection algorithm, the majority were related to the sigma parameter of the response distribution, while only three related to the mean parameter. The findings indicate that women from at least the poorer households are less likely to be malnourished, with higher variability among the richer and the richest households. Our findings regarding spatial effects have intriguingly revealed that malnutrition among women in Nigeria exhibits spatial structures wherein neighboring states across various geopolitical zones of the country demonstrate similar patterns. These findings demonstrate the unequal allocation of resources among various geographic places within the states, especially in the cases of states with notable deviations.