Saturated Emergencies: forecast the influx six hours in advance to trigger the full capacity plan earlier

In a calibrated simulation on a large American hospital, trigger the full capacity protocol over prediction rather than after exceeding the threshold reduces the time spent saturation.

Faced with overcrowding in emergency rooms, many hospitals have a full-capacity protocol that mobilizes additional beds and staff, but it is usually only activated once the saturation threshold is exceeded. Five researchers tested an anticipatory approach: a machine learning model predicts, hour by hour, the number of people in the waiting room six hours in advance and triggers the protocol before the situation deteriorates.

A simulated year, ten times

The model was integrated into a discrete-event simulation of patient flow, calibrated to the real parameters of a large American hospital, and then compared to conventional activation over 365 simulated days, with ten independent repetitions. Early activation significantly reduces the proportion of time spent above the saturation threshold and the average duration of the protocol, regardless of the threshold chosen. It also reduces the variability of crowding and episodes of extreme congestion.

The length of stay and waiting time generally improve, but the difference is not statistically significant. The authors believe that these results are sufficient to justify a trial in real conditions.