A machine learning approach to predict the hospital length of stay after kidney surgery

Abstract

The analysis of the hospital length of stay could provide a new perspective in the design of strategies to optimize the patients’ management, the clinical operations, and, not least, improve the overall healthcare service quality perceived by both patients, healthcare staff, and healthcare administrators and decision-makers. In the case of kidney injuries, the possibility to predict the length of stay is crucial for ensuring proper management of the patients undergoing surgical interventions. It is therefore significant for clinicians to have the option to anticipate the length of stay of patients and to deal with the main features influencing it to decrease the hospital length of stay. In this work we center around patients undergoing kidney surgery. Information was gathered from more than 3000 cases of kidney surgeries at the national hospital "A.O.R.N. Antonio Cardarelli" of Naples. A machine learning approach has been proposed to classify and predict the LOS of patients undergoing kidney surgery, and the tested algorithms have been assessed and compared in terms of performance metrics. Best predictive models have been identified and described in view of their potential impact in the improvement of the kidney surgery management procedures.