Abstract / Summary
Under-five mortality remains an important indicator of child health in Ethiopia, with substantial geographic and demographic variation. This study examined the timing, determinants, and spatial heterogeneity of under-five mortality using nationally representative birth-history data from the 2019 Ethiopian Demographic and Health Survey (EDHS). The analysis included 5,770 children born during the five years preceding the survey, of whom 342 experienced death before age five. Individual child survival histories were transformed into 165,749 person-month observations. A Bayesian discrete-time survival model with a complementary log-log link was fitted, incorporating age-specific baseline hazards, individual-level covariates, a random effect for 305 primary sampling units (PSUs), and a BYM2 spatial effect for Ethiopia’s 11 regions. DHS sampling weights were incorporated through a normalized Bayesian pseudo-likelihood. Model adequacy was assessed using DIC, WAIC, posterior predictive checks, and prior-sensitivity analysis. Under-five mortality was strongly concentrated in early childhood, with 91.0% of deaths occurring within the first 12 months of life. After adjustment for other covariates, female children had a lower mortality hazard than male children (HR 0.77, 95% CrI 0.60–0.98). Compared with adolescent mothers aged 15–19 years, mortality hazard was lower among mothers aged 20–24 (HR 0.46, 95% CrI 0.27–0.78), 25–29 (0.39, 0.22–0.69), 30–34 (0.33, 0.17–0.64), and 35–39 years (0.37, 0.18–0.75). Education, household wealth, residence, and birth-order categories showed no clear independent association after adjustment. Regional spatial effects varied, with Somali having the highest posterior median spatial hazard multiplier (1.28, 95% CrI 0.93–2.29), although all regional credible intervals included 1. Residual PSU heterogeneity was more pronounced than the structured regional spatial component. Under-five mortality in the 2019 EDHS cohort was predominantly an early-life mortality problem, with lower mortality hazards among female children and children born to mothers aged 20–39 years compared with adolescent mothers. The findings also indicate meaningful residual geographic and cluster-level heterogeneity, although uncertainty around regional effects remains substantial. Ethiopia’s child-survival strategies should maintain strong emphasis on neonatal and early-infancy survival, strengthen maternal and child health support for adolescent mothers, and incorporate region-sensitive mortality surveillance and resource planning, particularly in areas showing higher posterior regional hazard estimates. These priorities should be evaluated through continued population-based monitoring rather than interpreted as causal effects from this observational study.