Forecasting emergency department crowding: an external, multicenter evaluation.

Area of Science:

• Emergency Medicine
• Health Services Research
• Data Science

Background:

• Emergency department (ED) crowding is a significant challenge impacting patient care and operational efficiency.
• Accurate forecasting of ED crowding metrics is crucial for effective resource management.
• Existing tools for ED crowding prediction require validation across diverse institutional settings.

Purpose of the Study:

• To apply and validate the ForecastED tool for predicting near-future emergency department (ED) crowding at multiple external institutions.
• To assess the generalizability of the ForecastED tool in forecasting waiting counts, occupancy levels, and boarding counts.
• To evaluate the reliability, calibration, and accuracy of the ForecastED tool across different hospital settings.

Main Methods:

• The ForecastED tool was validated using historical data from five institutions external to the development site.
• A sliding-window design was employed for parameter estimation and forecast validation.
• Forecasts for waiting count, occupancy level, and boarding count were generated for 2, 4, 6, and 8 hours ahead and compared with observed data.

Main Results:

• The ForecastED tool demonstrated successful application across five diverse institutions.
• Forecasts were more reliable and accurate at shorter time horizons (2 hours) compared to longer ones (8 hours).
• While generally reliable, the boarding count forecast showed less reliability at four of the five external sites.

Conclusions:

• The ForecastED tool provides potentially useful forecasts for ED crowding measures at external sites without requiring modifications.
• The tool's generalizability suggests its utility in improving ED operations and resource allocation.
• Future research will focus on real-time validation and integration with ED information systems.