Forecasting Accuracy of Pharmaceuticals at Ethiopian Pharmaceutical Supply Services

Accurate forecasting is critical to ensure that life-saving medicines are delivered to those in need. Dynamic and precise forecasting enables evidence-based planning and supports effective decision-making within pharmaceutical supply chains. This study aimed at assessing the forecast accuracy of pharmaceuticals within the Ethiopian Pharmaceutical Supply Services (EPSS). A retrospective time series analysis was conducted using five years of historical data (2018 - 2022) on forecasted and issued quantities of pharmaceuticals. Thirty-three essential pharmaceuticals were sampled purposively for analysis: 25 identified as key products by the Ethiopian Ministry of Health and 8 additional products chosen in consultation with EPSS experts. All selected items were included in the EPSS procurement list and had complete data records. Forecast accuracy was evaluated using point forecast metrics, including Root Mean Squared Error (RMSE), Symmetric Mean Absolute Percentage Error (sMAPE), Mean Absolute Percentage Error (MAPE), and Mean Absolute Scaled Error (MASE). The analysis revealed significant forecast inaccuracies across the selected products. Only 13 of the 33 products (40%) had MAPE values below 50%. Several high-priority items exhibited high MAPE and RMSE values. Furthermore, The MASE results revealed that 13 products (39.4%) exhibited forecast accuracy inferior to that of the naïve benchmark, indicating suboptimal model performance for those items. To explore alternatives, three basic forecasting models Naïve, Mean, and 3-Year Moving Average were tested. The 3-Year Moving Average model consistently achieved the lowest MAPE values for 18 (54.54%) of the products, demonstrating its potential to enhance forecast accuracy. The study revealed that there were high forecast errors in the demand forecasting system of EPSS. The high degree of variability in forecast errors, along with limited model performance compared to naïve methods, underscores the need for a better forecasting techniques and approaches at EPSS.