Machine Learning Methods for Predicting the Admissions and Hospitalisations in the Emergency Department of a Civil and Military Hospital

Abstract The importance of the Emergency Department (ED) in hospitals was highlighted during the outbreak of the COVID-19 pandemic when ED overcrowding became a critical issue. The management of the service is critical for the effectiveness and efficient operation of the department. In this study, we present the results of the application of different algorithms for forecasting ED admissions (both seven days and four months ahead) and daily hospitalisations. To do this, we have employed the ED admissions and inpatients series from a Spanish civil and military hospital. The ED admissions have been aggregated on daily basis and on the official workers' shifts, meanwhile the hospitalisations series have been considered daily. Over that data we employ two algorithms types: time series (AR, H-W, SARIMA and Prophet models) and feature matrix (LR, EN, XGBoost and GLM models). In addition, we create all possible ensembles among the models in order to find the best forecasting method, even splitting the series by shifts. Our results show that time series algorithms achieve the best performance for almost all cases, but in the short term both time series and feature matrix perform similarly. In general, the ensembles slightly improve the predictions over the single models.