Comparison of machine learning algorithms to predict length of hospital stay in patients undergoing heart bypass surgery

Abstract

Patients affected by coronary artery obstruction, generally undergo aortocoronary bypass, an open-heart surgery that considerably affect health care expenditure. Since that, the monitoring and government of Aortocoronary bypass performance may be of help in health care management. In this work we compare various machine learning-based classification algorithms, to determine the length of stay for aortocoronary bypass. Data were collected on a group of 116 patients of the “San Giovanni di Dio e Ruggi D'Aragona” University Hospital of Salerno (Italy). Different socio-demographic, clinical, and organizational factors were taken into consideration as input parameters of the model for carrying out the classification analysis. The predictive capability of each of the tested machine learning algorithms was assessed in terms of accuracy and error percentages in the classification and obtained results were compared. Among the adopted algorithms, the Random Forest showed far better performances than the other ones, with an accuracy level of around 97%, thus potentially suggesting the Random Forest as a reliable predictive tool in the determination of the length of hospital stay of healthcare data for patients undergoing coronary artery bypass surgery.