AI framework cuts ER boarding time by up to 70%

A team of researchers from the University of North Carolina and other institutions have published a preprint on arXiv proposing a proactive bed-request framework for emergency departments (EDs). The core insight is to use predictions about each patient's admission probability and time to disposition to request inpatient beds early, before the final admission decision. They formulate the problem as a Markov decision process and develop three data-driven policies: approximate dynamic programming, reinforcement learning (RL), and a newsvendor-type heuristic.

Using a simulation model fed with data from a large ED, the study shows that proactive aggregate bed requests can reduce average boarding times for admitted patients by 30–70% and average length of stay for all ED patients by 6–15%, while creating only modest idle time for inpatient beds. The newsvendor heuristic provides the most attractive tradeoff between ED performance and inpatient bed idle time, whereas the RL heuristic produces smoother bed-request patterns when stability is crucial. The work demonstrates how prediction tools can improve ED operations and balance delays against idle capacity.