Modeling Infant Mortality in Epidemiology Using Bayesian Hierarchical Model Approach

Abstract

The study proposed hierarchical Bayesian model to simultaneously capture the over-dispersion due to the effect of varying population sizes across states and spatial auto-correlation inherent in the infant mortality at small-area level in Nigeria. A cross-sectional study among 31842 children were extracted from 2013 Nigeria demographic and Health Survey data. Out of which 2886 children died before reaching the age five years. The standardized mortality ratio (SMR) was estimated and mapped to highlight unusual cluster of the child mortality. Poisson regression model with random effects are formulated to capture spatial heterogeneity of geographical inequalities in child mortality. A full Bayesian framework via Markov chain Monte Carlos (McMC) simulation was used to estimate model parameters in WinBUGS. The results showed that economically deprived households, 2.088: 95% CI(1.088, 3.165) were significantly associated with childhood mortality, while unhygienic sanitary and unimproved water source were not significant. The probability maps detected clusters of high prevalence mortality in the northern regions and relatively low prevalence in south-west region of Nigeria. Our approach estimated the geographical variation as well as potential risk factors of infant mortality. The findings can assist relevant agency with information for the initiating public health interventions and child survival.