Short-Term Forecast of Emergency Departments Visits Through Calendar Selection

Emergency Departments (ED) overcrowding is a common and well-known problem, associated with decreased patient safety, increased mortality rates, and which leads to staff burning out. The ability to predict the ED hourly visits can then relieve overcrowding’s consequences. In this paper (The content of this paper is a contribution to the International Conference on Time Series and Forecasting 2019 (ITISE2019), held in Granada [1].), we present a method which takes into account calendar effects for short-term forecasting of ED visits is presented. Our approach combines a calendar selection rule with a well-known machine learning algorithm belonging to the class of similar shape algorithms, to predict the incoming visit volume for a tunable number of days ahead.