A Bayesian Bivariate Spatial Analysis of the Shared and Distinct Determinants of Stunting and Wasting Among Children in Ethiopia: Evidence from the 2019 Mini DHS

Childhood malnutrition remains a major public health challenge in Ethiopia, where stunting and wasting co-exist but may arise from distinct spatial and etiological processes. Analyses focusing on a single outcome may overlook the interdependence of these conditions and their geographic heterogeneity. This study aimed to disentangle the determinants of stunting and wasting among children under five years of age using a Bayesian bivariate spatial modelling framework. Data from 5,405 children included in the 2019 Ethiopia Mini Demographic and Health Survey were analyzed. Stunting and wasting were modelled as correlated binary outcomes using Bayesian bivariate hierarchical geostatistical models implemented through SPDE-INLA, accounting for child, maternal, household, and environmental covariates, non-linear age effects, and spatial dependence. Model performance was assessed using the deviance information criterion, Watanabe-Akaike information criterion, and marginal log-likelihood. The bivariate model identified shared socio-economic and biological determinants. Multiple births, male sex, low maternal education, a higher number of under-five children, and household poverty were associated with increased risks of both outcomes. Female-headed households were associated with lower odds of stunting but higher odds of wasting. Spatial analysis revealed elevated residual stunting risk in the northern and central highlands, whereas wasting hotspots were concentrated in northeastern pastoralist regions. Residual spatial correlation was weak ({rho} = -0.12), indicating largely independent geographic patterns. These findings suggest that effective child nutrition policies in Ethiopia require outcome-specific and regionally tailored interventions addressing both chronic and acute forms of malnutrition.