Deep Learning-Based Forecasting of Boarding Patient Counts to Address ED Overcrowding

TL;DR

This paper introduces a deep learning framework that accurately predicts ED boarding counts six hours in advance using operational data, aiding hospitals in managing overcrowding proactively.

Contribution

It develops a novel deep learning approach utilizing operational and contextual data for short-term ED boarding count forecasting without patient-level details.

Findings

TSTPlus model achieved MAE of 4.30 and R2 of 0.79.

Broader input features significantly improved prediction accuracy.

Framework effectively forecasts during extreme overcrowding periods.

Abstract

This study presents a deep learning-based framework for predicting emergency department (ED) boarding counts six hours in advance using only operational and contextual data, without patient-level information. Data from ED tracking systems, inpatient census, weather, holidays, and local events were aggregated hourly and processed with comprehensive feature engineering. The mean ED boarding count was 28.7 (standard deviation = 11.2). Multiple deep learning models, including ResNetPlus, TSTPlus, and TSiTPlus, were trained and optimized using Optuna, with TSTPlus achieving the best results (mean absolute error = 4.30, mean squared error = 29.47, R2 = 0.79). The framework accurately forecasted boarding counts, including during extreme periods, and demonstrated that broader input features improve predictive accuracy. This approach supports proactive hospital management and offers a practical method for mitigating ED overcrowding.