Abstract

Healthcare delivery across Nigeria is shaped by a familiar set of pressures: too few clinicians, thin budgets, patchy infrastructure and data that is often incomplete or trapped in paper registers. This study asks a practical question. Can predictive data analytics, built from the data hospitals already generate, help managers make better decisions about where scarce resources should go? To answer it, we assembled 12,480 anonymised outpatient records collected between January 2023 and December 2025 from four secondary health facilities in two states, and we trained four predictive models such as logistic regression, random forest, gradient boosting (XGBoost) and an artificial neural network to forecast missed appointments and weekly patient demand. XGBoost performed best, reaching an accuracy of 90.2 per cent and an F1-score of 0.885, and it flagged distance to facility and prior attendance history as the two strongest predictors of whether a patient would show up. The demand-forecasting component reduced average weekly staffing error from roughly 22 per cent under the status quo to about 9 per cent. What the numbers point to is modest but real: even lightweight, locally trained models can sharpen resource allocation in settings where every bed, nurse and vial counts. We close with a deployment framework tuned to low-connectivity, low-budget realities, and we are honest about the obstacles, data quality, staff buy-in and the risk of algorithmic bias that will decide whether any of this survives contact with a busy clinic