Using Hospital Admission Predictions at Triage for Improving Patient Length of Stay in Emergency Departments.

This article focuses on developing and evaluating policies for requesting hospital beds early at emergency department (ED) triage to reduce patient length of stay (LOS) by shortening boarding times. The authors propose a framework that uses an Admission Prediction Tool (APT), based on logistic regression of triage data, to estimate each patient's probability of hospital admission and then decide whether to request a bed early, considering ED census and hospital bed availability. They introduce three policies: a simple Emergency Severity Index-based policy (ESIB), a Fixed Threshold policy (FT) using a constant admission probability cutoff, and a Census- and Time-based Threshold policy (CTT) that dynamically adjusts the threshold based on ED crowding and time-dependent parameters. Simulation studies using data from an academic hospital ED in the southeastern United States show that both FT and CTT outperform ESIB, with CTT providing the greatest reductions in average patient LOS, especially during periods of high patient demand, while maintaining manageable levels of false early bed requests. The study highlights the potential operational benefits of coordinated ED-hospital policies but notes cultural and implementation challenges due to changes in traditional hospital admission workflows.