Small Area Risk Mapping of Under-five Mortality: A District-level Analysis using Parametric Approach

Abstract

Mortality of under-five children remains a pressing public health concern in Tanzania, with spatial disparities contributed by the intersection of socioeconomic, health and environmental factors. This study employs Bayesian techniques and epidemiological spatial methods to estimate and map district-level under-five mortality, focusing on malaria prevalence, maternal education and source of drinking water. The districts with high risk are identified using georeferenced data of under-five mortality from recent Tanzania Demographic and Health Survey. We estimated the relative risk (RR) of under-five mortality within a hierarchical Bayesian spatial framework using three priors namely Besang-York-Mollie (BYM), Dean & Cressie (DC) and Leroux (BMY2). The model performance was evaluated using Watanabe-Akaike Information criterion (WAIC) to identify the best, from which the relative risk of under-five mortality was estimated and visualized using interactive and static maps. The Leroux (BYM2) model effectively estimated the relative risk compared to BYM and DC. The findings reveal that there are disparities of under-five mortality even across districts of the same region. The top ten districts with higher relative risk of under-five mortality based on model estimates are Ifakara (RR = 3.2), Nyasa ( RR = 3.0), Babati (RR = 2.74), Rungwe (RR = 2.58), Songea (RR = 2.58), Mbarali (RR = 2.55), Rorya (RR = 2.36), Namtumbo (RR = 2.36), Mbulu (RR = 2.11) and Mbogwe (RR = 2.11). The findings advocate for targeted interventions in these districts that may reduce under-five mortality and consequently increase child survival rates in Tanzania.