B1371
Title: Forecasting emergency department length of stay and hospital admissions during the pandemic
Authors: Siddharth Arora - University of Oxford (United Kingdom) [presenting]
James Taylor - University of Oxford (United Kingdom)
Abstract: The aim is to deploy predictive modelling in an emergency department (ED) to help improve patient outcomes and assist hospitals to make informed interventions for resource allocation/expansion to meet the challenges posed by the pandemic. Firstly, we investigate the impact of the pandemic on ED patient-flow, by focussing on a multitude of outcomes of interest to the service provider, such as attendances, ambulance arrivals, emergency admissions to the hospital, length of stay, and the reason for attendance. Secondly, we forecast the total ED length of stay (LOS), as communicating these estimates can help reduce patient drop-out rates and improve patient satisfaction. Finally, we predict the risk of emergency admission to the hospital, to assist EDs to prioritize patients. For each low-acuity patient, personalized and probabilistic estimates of their LOS and admission risk are provided at their time of registration at the ED. Our forecasting methodology, which is based on machine learning, is adapted to account for the reorganization of ED triage protocol during the pandemic. We envisage the findings of this study could potentially help facilitate patient risk stratification and case management in the ED, which could also have implications for hospital capacity planning.
