From Forecasts to Decisions: Assessing the Operational Value of Forecasting Accuracy in Emergency Department Staffing Decisions

Accurate forecasting is essential for emergency department staffing decisions, where errors have a serious impact on operating costs and the quality of care. Traditional forecasting evaluation metrics, such as Root Mean Square Error (RMSE), do not capture the effects of expected errors on operations. In this research, we experimentally evaluate how the accuracy of forecasts affects day-to-day staffing decisions by integrating 11 forecasting methodologies (from classic statistical models to advanced machine learning) into a staffing-optimization framework.Using real-world Emergency Department (ED) data, we compare the performance of each forecasting method with operational outcomes that are directly relevant to staffing decisions, including daily staffing imbalance and the number of overstaffing and understaffing days. The results show that forecasts with lower bias, regardless of their overall statistical accuracy, lead to significantly fewer staffing mismatches. Forecasts that closely match actual demand distributions, as indicated by a low Mean Error (ME), generate better operational outcomes than forecasts with similar RMSE but higher bias. In contrast, forecasts that systematically overestimate or underestimate demand, even when statistically accurate, result in considerable staffing imbalances. These findings demonstrate that reducing forecast bias is more important for effective staffing decisions than minimizing aggregate error measures alone.By aligning forecasting and optimization objectives, our findings provide actionable guidance for ED managers to improve resource allocation, maintain balanced staffing levels, enhance patient care, and ensure more efficient use of limited healthcare resources.