Crowding in the emergency department in the absence of boarding – a transition regression model to predict departures and waiting time

BACKGROUND: Crowding in the emergency department (ED) is associated with increased mortality, increased treatment cost, and reduced quality of care. Crowding arises when demand exceed resources in the ED and a first sign may be increasing waiting time. We aimed to quantify predictors for departure from the ED, and relate this to waiting time in the ED before departure. METHODS: We utilised administrative data from the ED and calculated number of arrivals, departures, and the resulting queue in 30 min time steps for all of 2013 (N = 17,520). We build a transition model for each time step using the number of past departures and pre-specified risk factors (arrivals, weekday/weekend and shift) to predict the expected number of departures and from this the expected waiting time in the ED. The model was validated with data from the same ED collected March through August 2014. RESULTS: We found that the number of arrivals had the greatest independent impact on departures with an odds ratio of 0.942 (95%CI: 0.937;0.948) corresponding to additional 7 min waiting time per new arrival in a 30 min time interval with an a priori time spend in the ED of two hours. The serial correlation of departures was present up to one and a half hour previous but had very little effect on the estimates of the risk factors. Boarding played a negligible role in the studied ED. CONCLUSIONS: We present a transition regression model with high predictive power to predict departures from the ED utilising only system level data. We use this to present estimates of expected waiting time and ultimately crowding in the ED. The model shows good internal validity though further studies are needed to determine generalisability to the performance in other settings. ELECTRONIC SUPPLEMENTARY MATERIAL: The online version of this article (10.1186/s12874-019-0710-3) contains supplementary material, which is available to authorized users.