Enhancing Emergency Department Management: A Data-Driven Approach to Detect and Predict Surge Persistence

Area of Science:

• Healthcare Management
• Public Health Analytics
• Time Series Forecasting

Background:

• Effective healthcare planning and resource allocation depend on predicting emergency department (ED) patient attendance.
• Unforeseen health crises necessitate robust systems for timely operational adjustments in EDs.

Purpose of the Study:

• To develop and validate an early warning system for detecting emerging trends and predicting surges in ED attendance.
• To enhance the preparedness and agility of emergency departments through proactive operational planning.

Main Methods:

• Utilized 13 years of historical ED attendance data (1,700,887 data points) from January 2010 to December 2022.
• Developed and validated three models: Seasonal Autoregressive Integrated Moving Average with eXogenous factors (SARIMAX), Exponentially Weighted Moving Average (EWMA) for surge prediction, and a trend persistence model.
• Employed EWMA control charts for drift detection and machine learning models trained on kernel-regressed ED attendance curve slopes for trend persistence prediction.

Main Results:

• The EWMA control chart successfully identified significant COVID-19 events in Singapore.
• The surge prediction model provided early signals of ED attendance trend changes during the COVID-19 pandemic (January 2020 - December 2022).
• The trend persistence model estimated novel trend persistence with a mean absolute error of 7.54 days (95% CI: 6.77-8.79).

Conclusions:

• The study introduced a proactive surge detection framework for emergency healthcare management.
• This framework is crucial for improving the preparedness and responsiveness of emergency departments during health crises.
• The developed system offers timely alerts for proactive operational planning in EDs.