Forecasting demand of emergency care.

Area of Science:

• Healthcare Operations Research
• Epidemiology
• Time Series Forecasting

Background:

• Accurate forecasting of hospital bed occupancy is crucial for efficient healthcare management.
• Emergency admissions significantly impact hospital resource allocation and patient flow.
• Predicting fluctuations in demand is essential for maintaining service quality.

Purpose of the Study:

• To develop and validate a model for forecasting daily occupied beds due to emergency admissions in acute hospitals.
• To identify key factors influencing emergency bed occupancy, including environmental and epidemiological variables.
• To assess the relationship between forecast accuracy, hospital volatility, and emergency department (A&E) waiting times.

Main Methods:

• Development of a forecasting model for daily occupied beds.
• Utilizing out-of-sample forecasts to evaluate model performance (RMS error).
• Incorporating air temperature and Public Health Laboratory Service (PHLS) data on influenza-like illnesses as predictors.
• Analysis of Generalized Autoregressive Conditional Heteroskedasticity (GARCH) errors to model volatility and its impact on waiting times.

Main Results:

• The model achieved an RMS error of 3% for 32-day ahead forecasts of emergency bed occupancy.
• Emergency bed occupancy was found to be significantly correlated with air temperature and influenza-like illness data.
• Periods of high volatility, identified by GARCH errors, were associated with increased A&E waiting times.
• Volatility provided earlier warnings of A&E waiting times compared to total bed occupancy.

Conclusions:

• The developed model provides a reliable method for forecasting emergency bed occupancy.
• Environmental and epidemiological factors are important predictors of hospital bed demand.
• Monitoring volatility is critical for proactive management of A&E waiting times, offering a leading indicator over overall bed occupancy.