Using Machine Learning to Create a Hospital Revenue Prediction Model

The healthcare revenue cycle in the United States refers to the financial process that healthcare providers follow to manage patient payments, from the moment a patient makes an appointment to the resolution of any outstanding balances. The complex billing systems, reimbursement variability, regulatory changes, and patient demographics make the revenue predictions for a hospital way more complex than any other industry. This study uses 28 variables related to patient demographics, visit characteristics, and payor attributes to develop a predictive model for understanding and forecasting Payment-to-ratio PCR trends. A descriptive analysis of the first and the last quarters of the data was conducted for 6 variables and their effect was studied. Multiple machine-learning methods were used, and the decision tree regression had the best performance resulting in a testing R-squared equal to 72%.