A comparison of different Machine Learning algorithms for predicting the length of hospital stay for pediatric patients

Abstract

Prolonged length of stay (LOS) is one of the most significant issues that hospitals must face since it can determine an increase of costs and risk of complications and a decrease of patient satisfaction. The capability of accurately predicting LOS is a valuable tool for helping hospital administrators in resource planning, encouraging quality improvement actions and providing support for medical practice. In particular, delayed hospital discharge in pediatric population is something that needs to be carefully considered because of their vulnerability and complexity from the medical viewpoint. Predicting the use of hospital resources and beds in pediatric departments could also help to better dimension hospitalizations management. This work has the aim to determine the predictive factors for the length of stay for patients admitted to the Complex Operative Units of Pediatrics and Pediatric Surgery at the “San Giovanni di Dio e Ruggi d'Aragona” University Hospital of Salerno and to build a classification model of LOS exploiting the potential of Machine Learning. Different algorithms are implemented, and their evaluation metrics are assessed and compared together to develop a prediction model with high performances.