Predicting Cost Recovery Rate of Inpatient Cases: the Application of Machine Learning Approaches

Abstract

This study aims to predict the cost recovery rate (CRR) of inpatient cases in Indonesian hospital utilizing the powerful approaches of machine learning. The data were collected from hospital insurance claim database of a middle size Indonesian hospital. The number of observations is 15,955 datasets of inpatient cases between 2017 and 2020. The data includes patient profiles (age and sex), length of stay (LOS), severity (SEV), patient class, patient costs, treatment types and reimbursement fees from the insurer. Using logistic regression and decision tree, this study revealed that bed class, admission month, sex, discharge status, SEV, LOS, ICU treatment, surgery, supporting facilitation, radiology, laboratory, blood examination, and rehabilitation are the determinants of inpatient cases' CRR. Meanwhile, the decision tree approach revealed that LOS, bed class and SEV as the primary classifier of the inpatient cases'CRR. The results of this study can be used for the hospital managers to predict and anticipate non-covered patient costs during the first days of hospital admission and in turn minimize financial deficit of the hospital.