Determinants of child survival in Ghana: A multi-level survival analysis of individual and contextual factors

This study investigates child survival determinants in Ghana using a multi-level survival analysis of the 2006, 2011 and 2018 Multiple Indicator Cluster Surveys (n=13,125). We employed Cox proportional hazards, parametric accelerated failure time (AFT) and gamma frailty models to assess individual and contextual factors influencing under-five mortality, accounting for unobserved heterogeneity. Key findings reveal a wealth-mortality relationship, with middle-wealth households showing lower mortality (HR=0.60, TR=1.44) compared to low-wealth, while rich households exhibited higher risks (HR=1.84, TR=0.64). Maternal education significantly reduces mortality risk (primary: HR=0.69, TR=1.28; secondary+: HR=0.86, TR=1.12). Institutional deliveries lower mortality by 29%–43% (TR=1.29–1.43). Northern regions (Upper West: HR=2.53, TR=0.50; Northern: HR=2.24, TR=0.56) face elevated risks, reflecting structural disparities. Gamma frailty models revealed a significant frailty variance (θ>0), confirming substantial unobserved heterogeneity across regions and indicating that unmeasured, region-specific factors influence child mortality. The Weibull AFT model provided the best fit (AIC=21,999.45), highlighting early-life mortality risks. The policy implications are clear that interventions must be spatially targeted and move beyond simple wealth stratification, requiring specific environmental and infrastructural improvements tailored to high-risk and economically disparate areas of Ghana.