Abstract
Child mortality is a critical measure of healthcare performance, with Sustainable Development Goal 3.2 targeting reductions in neonatal and under-five mortality to 12 and 25 deaths per 1,000 live births, respectively, by 2030. However, Ghana's child mortality rate remains high at 32 deaths per 1,000 live births, with significant regional disparities driven by healthcare access and economic inequalities. This research develops risk maps to identify high-risk regions and determinants, providing valuable insights for targeted interventions to reduce child mortality. We analyzed under-five mortality (U5MR) in Ghana (2022) using data from the 2022 Ghana Demographic and Health Survey. Crude death rates (per 100 live births) were calculated and smoothed with Global Empirical Bayesian (GEB) and Local Empirical Bayesian (LEB) methods. Spatial dependence and clustering were evaluated using Global Moran's I, Local Indicators of Spatial Association (LISA), and Getis-Ord statistics, and SaTScan statistic. Spatial associations between U5MR and its determinants were assessed using bivariate LISA. Predictor importance was quantified using Gradient Boosting, a nonparametric ensemble learning approach, with interpretability enhanced by SHapley Additive exPlanations (SHAP), including feature importance rankings, partial dependence, and waterfall plots. The Bayesian spatial model, Besag-York-Mollié (BYM2) was also used to further study spatial pattern of child mortality risk in Ghana. In 2022, Ghana's national under-five mortality rate (U5MR) was 6.4 per 100 live births, with marked regional disparities. Oti, Northern, and Savannah recorded the highest risks, while Greater Accra and Ahafo had the lowest. Spatial analyses (GEB, LEB, SaTScan, LISA, Getis-Ord Gi*) consistently identified high-risk clusters in the northern belt, with a 27% elevated risk in a cluster spanning Oti, Northern, Bono East, and Volta. ANC coverage was the strongest predictor, explaining over 75% of U5MR variation, while caesarean rates also showed strong spatial associations. Persistently high U5MR in high-coverage regions points to gaps in quality of care and underlying structural inequities. Economic and governance factors were less influential, highlighting the critical need for targeted maternal and child health interventions in high-burden areas. The persistent regional disparities in under-five mortality in Ghana, particularly the elevated risks in the northern belt, highlight urgent equity challenges in maternal and child health. Despite national progress, high-risk regions remain underserved in terms of both access to and quality of health services. Antenatal care coverage emerged as the most influential determinant of mortality risk, reinforcing the need to prioritize maternal service delivery in disadvantaged areas. Policy efforts should move beyond national averages to implement region-specific strategies strengthening primary healthcare systems, addressing quality-of-care gaps, and tackling structural barriers such as rural inaccessibility. A more targeted and spatially-informed approach is essential to achieving equitable child survival outcomes across Ghana.