BAYESIAN HIERARCHICAL MODELING FOR FORECASTING HEALTH RESOURCE ALLOCATION IN GHANA'S RURAL HEALTHCARE SYSTEM

This study addresses the urgent need for accurate health resource forecasting in Ghana’s rural healthcare system, where over 60% of districts experience quarterly stockouts of essential medications and a doctor-to-population ratio as low as 1:11,000. To tackle these challenges, the study applies Bayesian Hierarchical Modeling (BHM), which integrates spatial heterogeneity, temporal variability, and prior data to forecast drug supply, personnel deployment, and equipment distribution. Using secondary data from 2020-2024 across five representative rural districts, the study employed spatial, temporal, and spatio-temporal Bayesian models. Statistical tests revealed significant results: a paired t-test for spatial effects showed t(4) = 6.45, p< 0.01; ANOVA for seasonal personnel deployment yielded F(3,16) = 5.62, p = 0.008; and integrating prior information improved equipment forecast accuracy from 72% to 78% (t(8) = 3.87, p = 0.005). The model achieved 84% forecast accuracy for antimalarials, with RMSE values as low as 3.2. A multiple regression model indicated that equipment efficiency (β = 0.58, p< 0.01), doctor-to-population ratio (β = 0.34, p = 0.04), and seasonal deployment (β = 0.28, p = 0.06) together explained 63% (R² = 0.63) of the variance in resource utilization. A strong correlation (r = 0.93) between equipment efficiency and utilization affirms the model’s predictive power. These findings underscore the necessity of dynamic, localized, and evidence-based forecasting systems. The research recommends embedding BHM into national health information platforms to enable real-time, district-level planning. It also advocates for investments in data infrastructure and personnel training to sustain model-driven resource equity across rural Ghana.