Using geographically weighted regression analysis to cluster under-nutrition and its predictors among under-five children in Ethiopia: Evidence from demographic and health survey

BACKGROUND: Malnutrition among under-five children is a common public health problem and it is one of the main cause for the mortality of under-five children in developing countries, including Ethiopia. Therefore, lack of evidence about geographic heterogeneity and predictors of under-nutrition hinders for evidence-based decision-making process for the prevention and control programs of under-nutrition in Ethiopia. Thus, this study aimed to address this gap. METHODS: The data were obtained from the Ethiopian Demographic and Health Survey (EDHS) 2016. A total of 9,384 under-five children nested in 645 clusters were included with a stratified two-stage cluster sampling. ArcGIS version 10.5 software was used for global, local and ordinary least square analysis and mapping. The spatial autocorrelation (Global Moran’s I) statistic was held in order to assess the pattern of wasting, stunting, and underweight whether it was dispersed, clustered, or randomly distributed. In addition, a Bernoulli model was used to analyze the purely spatial cluster detection of under-nutrition indicators through SaTScan version 9.6 software. Geographically weighted regression (GWR) version 4.0 software was used to model spatial relationships in the GWR analysis. Finally, a statistical decision was made at p-value<0.05 with 95%CI for ordinary least square analysis and geographically weighted regression. MAIN FINDINGS: Childhood under-nutrition showed geographical variations at zonal levels in Ethiopia. Accordingly, Somali region (Afder, Gode, Korahe, Warder Zones), Afar region (Zone 2), Tigray region (Southern Zone), and Amhara region (Waghmira Zones) for wasting, Amhara region (West Gojam, Awi, South Gondar, and Waghmira Zones) for stunting and Amhara region (South Wollo, North Wollo, Awi, South Gondar, and Waghmira zones), Afar region (Zone 2), Tigray region (Eastern Zone, North Western Zone, Central Zone, Southern Zone, and Mekele Special Zones), and Benshangul region (Metekel and Assosa Zones) for underweight were detected as hot spot (high risk) regions. In GWR analysis, had unimproved toilet facility for stunting, wasting and underweight, father had primary education for stunting and wasting, father had secondary education for stunting and underweight, mothers age 35–49 years for wasting and underweight, having female children for stunting, having children eight and above for wasting, and mother had primary education for underweight were significant predictors at (p<0.001). CONCLUSIONS: Our study showed that the spatial distribution of under-nutrition was clustered and high-risk areas were identified in all forms of under-nutrition indicators. Predictors of under-nutrition were identified in all forms of under-nutrition indicators. Thus, geographic-based nutritional interventions mainly mobilizing additional resources could be held to reduce the burden of childhood under-nutrition in hot spot areas. In addition, improving sanitation and hygiene practice, improving the life style of the community, and promotion of parent education in the identified hot spot zones for under-nutrition should be more emphasized.