Demand Prediction for Better Hospital Capacity Management.

Area of Science:

• Health Services Research
• Data Science
• Operations Research

Background:

• Effective hospital bed management is essential for patient care and resource optimization.
• Accurate forecasting of patient flow is a persistent challenge in healthcare systems.

Purpose of the Study:

• To evaluate statistical and machine learning models for predicting daily/hourly hospital demand.
• To introduce and assess the Advanced Demand Prediction Tool (ADePT) for healthcare forecasting.

Main Methods:

• Utilized SARIMAX time series model within the ADePT tool.
• Compared ADePT against five other forecasting models using real-world tertiary health service data.
• Evaluated predictions for inpatient admissions, separations, and emergency department presentations.

Main Results:

• ADePT generally outperformed other models for inpatient admissions and separations across various forecast horizons.
• No statistically significant accuracy differences were observed for emergency department presentations.
• ADePT demonstrated high accuracy for smaller patient subgroups, including emergency and elective admissions.

Conclusions:

• ADePT offers a reliable solution for short-term and long-term hospital demand forecasting.
• The tool can significantly improve daily bed management and long-term healthcare capacity planning.
• Accurate forecasting supports efficient resource allocation and enhanced patient care delivery.