Bayesian Accelerated Failure Time Model With Spatial Dependency: Application To Under Five Mortality Rate

Abstract

Nigeria is one of the top five in the world with the highest under-five mortality (U5M) rate. The risk of U5M in Nigeria is assumed to vary from one state to another due to diversity in socio-economic and even environmental factors. Thus, this study aimed to quantified the hazard of U5M using Bayesian Accelerated Failure Time (AFT) with spatial dependency. The data for the study were obtained from 2018 Nigerian Demographic and Health Survey (NDHS). The study utilized Bayesian technique based on Markov Chain Monte Carlo (MCMC) technique to obtained the posterior estimates of the parameters. The Loglogistic, Weibull and Lognormal AFT models with and without spatial dependency were considered in this study. Out of these models used, the log-logistic AFT model with Intrinsic Conditional Autoregressive (ICAR) spatial prior performed better than the other models considered in the study. The findings revealed that the survival of under-five children (U5C) was not homogeneous across the states (95% CI: 2.6330, 11.4630). The subject-specific factors such as maternal age at birth, duration of breast feeding, preceding birth intervals, maternal educational qualification, wealth index, region, number of antenatal visits, duration of pregnancy, gender of child, twin status, contraceptive used and toilet facility were the significant risk factors of U5M. Based on these findings, it was recommended among others that the disparities observed across states should be taken into account at the policy level in order to meet the Sustainable Development Goal (SDG) 2030 targets.