Determinants and regional patterns of under-five mortality in East Africa: a cox frailty meta-analysis approach

Abstract

Mortality rates for children under five continue to pose a serious public health concern in Sub-Saharan Africa, with considerable variation both between and within nations. Comprehending the drivers and regional differences is essential for attaining Sustainable Development Goal (SDG) 3.2. The specific objective of the study was to evaluate the socio-demographic factors of under-five mortality, quantify regional inequalities, and present pooled evidence from Kenya, Rwanda, Tanzania, and Uganda. Using four cycles of Demographic and Health Surveys (DHS) per nation, this study used an extended Cox proportional hazards model with regional frailty variables to account for unobserved variability. A multivariate random-effects meta-analysis was then used to calculate the hazards ratios (HRs) and standard errors in R. Frailty values were plotted over DHS cycles to identify consistent high-risk areas using QGIS. Maternal education, household wealth, breastfeeding, and bigger family size were all protective against child mortality, with female children having a slightly higher risk. The meta-analysis corroborated these correlations, resulting in combined hazard ratios with smaller confidence ranges. Frailty mapping identified chronic mortality hotspots in Tanzania (Lindi, Mara, Pwani), Uganda (West Nile, Acholi), Western Kenya, and Southern Rwanda. Despite lowering rates, under-five mortality varies significantly by area. High-risk areas require tailored interventions that address maternal education, healthcare access, and poverty. DHS drawbacks include cross-sectional design, recollection bias, insufficient contextual information, and the absence of spatial autocorrelation modelling. Future study should use Bayesian hierarchical survival models and machine learning to better knowledge and policymaking.