Evaluation of Decision Tree for predicting Patients' Length of Stay After Arthroplasty Surgical Procedures in the Rural Healthcare

Predicted inpatient length of stay (LOS) after total joint arthroplasty (TJA) is important information that sets the expectation for the post-surgical recovery period, which helps hospital management plan, and manage surgical resources as efficiently as possible. In this study, the accuracy of prediction models for predicting patients' LOS in a rural community hospital was conducted in the Northwest United States of America. Data were collected from 181 patients for hip, knee, and shoulder surgeries. A decision tree (DT) prediction model was compared with the National Surgical Quality Improvement Program (NSQIP) calculator for predicting LOS of TJA. The DT model (RMSE = 0.67) provided a more accurate LOS prediction than the NSQIP calculator (RMSE = 1.18). Furthermore, a greater level of interpretability of the decision-making process makes the DT model very applicable in “high stakes” environments like healthcare.