FLEXIBLE BAYESIAN PROPORTIONAL ODD SURVIVAL MODEL WITH SPATIAL RANDOM EFFECT AND COVARIATES INTERACTION STRUCTURES: AN APPLICATION TO UNDER-FIVE MORTALITY RATE

Abstract

Despite concerted global efforts to reduce child mortality, 5.3 million under-five deaths were reported in 2018, with Nigeria struggling to meet the Sustainable Development Goal (SDG) 3 target of 25 deaths per 1,000 live births by 2030. This study introduced a flexible proportional odds (PO) model, incorporating spatial random effects and covariate interaction Structures, to elucidate the complex relationships between factors influencing Under-Five Mortality (U5M) in Nigeria. Leveraging 2018 Demographic and Health Survey (NDHS) Data, the study evaluated the extended model against existing models, including those without covariate interaction, those without spatial random effects, those with Independent and Identically Distributed (IID) priors, and those with Intrinsic Conditional Autoregressive (ICAR) spatial priors. The results showed that the extended model: PO survival model with ICAR spatial prior and logistic transformed covariates interaction structure (LTCIS) outperformed existing models based on Log Pseudo Marginal Likelihood (LPML), Deviance Information Criterion (DIC), and Watanabe Akaike Information Criterion (WAIC). Key predictors of U5M include maternal age, breastfeeding duration, preceding birth interval, maternal education, wealth index, region, antenatal visits, pregnancy duration, child's sex, twin birth, mosquito net use, and contraceptive use. Notably, interaction effects revealed that higher maternal education levels amplify the benefits of breastfeeding on U5M. This study demonstrated the effectiveness of the proposed PO model in capturing the intricate dynamics of U5M, highlighting geographic and social clustering patterns that can inform targeted interventions to reduce under-five mortality in Nigeria