Use of machine learning to predict abandonment rates in an emergency department

Overcrowding is a serious issue that Emergency Departments (EDs) must deal with, since it is leading to longer delays and greater patients’ dissatisfaction, which are directly connected with an increasing number of patients who leave the ED prematurely. Hospital is affected by this aspect in terms of lost revenues from opportunities missed in providing care and adverse outcomes deriving from ED process. For this reason, the ability to control and predict in advance patients who leave ED without any evaluation becomes strategic for healthcare administrators. The purpose of this work is to investigate causes that determine patients who leave the ED without being seen. Machine Learning algorithms are used in order to build and compare different models for LWBS prediction, with the aim of obtaining a helpful support tool for the ED management in healthcare facilities.