Bayesian Geo-Additive Model to Analyses the Spatial Pattern and Determinants of Childhood Anemia in Ethiopia

Abstract

Background: Anaemia poses significant public health challenges to most developing countries, associated with serious health consequences and affecting about one-fourth of the world’s population, mostly under five-year children. Objectives: This study aimed to Analyze the Spatial Pattern and Determinants of childhood anaemia in Ethiopia using Bayesian Geo-additive approach. Methods: Our study participants were all the children U5 who were confirmed to anaemia from the 2016 EDHS data source. The survey considered 10,641 children U5; of which 7,953 children with complete anaemia levels were included in this study. The outcome variable was defined as the presence or absence of anaemia based on the WHO cut-off points. In this study Moran’s, I was used to investigate the presence of spatial autocorrelation. A geo-additive model which allowed joint analyses of nonlinear effects of some covariates, spatial effects, and other fixed covariates were used. Inference used a fully Bayesian approach via Markov Chain Monte Carlo techniques. Results: Out of 7,953 children U5 years included in this study 4567 (57.4%) were anemic. Based on DIC model selection criteria Bayesian Geo-additive model was found to be appropriate. From the Model, household wealth index, types of toilet facilities, size of child at birth, education levels of mothers, and mother’s anemia status are found to be the significant determinants of childhood anaemia. Child age and mother BMI were found to have a nonlinear relationship with childhood anaemia. Conclusion: Our finding revealed that there was spatial variation in childhood anaemia across the region of Ethiopia with higher prevalence in the eastern and north-eastern parts of Ethiopia. Bayesian Geo-additive models that capture spatial effects fit the data well. Therefore, the concerned body may use the anemia prevalence map as a basis for interventions and resource allocations.