Forecasting Inpatient Bed Demand Using Exponential Smoothing with Trend and Seasonal Index

บทคัดย่อ

Forecasting of inpatient hospital bed demand is important for optimizing resource allocation and enhancing healthcare service. This study aims to improve the accuracy of the forecasting model, Exponential Smoothing with Trend, including the Seasonal Index (ETS-SI), to address the complexities of trend and seasonality in hospital admissions data. Anonymized inpatient data from government hospitals in Thailand (2020–2024) was used and categorized by hospital size (small, medium, and large). Then, the ETS-SI model integrated trend and seasonal components to improve forecasting precision and was evaluated against traditional forecasting methods, including Exponential Smoothing (ES), Exponential Smoothing with Trend (EST), and the Seasonal Index (SI). Mean Absolute Percentage Error (MAPE) was used to validate forecasting accuracy. ETS-SI achieved the lowest error rates compared with other methodologies for capturing seasonal and trend-driven variations. Findings emphasize the suitability of ETS-SI for addressing the unique challenges of hospital demand forecasting. The ETS-SI model demonstrated the lowest MAPE among all models, and statistical validation using ANOVA confirmed that its forecasting accuracy was significantly higher (p < 0.05). The ETS-SI model demonstrated superior forecasting accuracy and can be applied to support holistic hospital decision-making—particularly in patient scheduling, staff rostering, and bed resource planning. Future research should explore hybrid models for enhanced forecasting accuracy and broader applicability.