Machine Learning Model Developed to Aid in Patient Selection for Outpatient Total Hip Arthroplasty
Name: Jessica Ding
School: Vagelos College of Physicians and Surgeons, Class of 2023
Mentors: Thomas R. Hickernell, MD and Annette Wu, MD, MPH, PhD
Abstract
It is important to select appropriate patients for outpatient total hip arthroplasty (total hip replacement) (THA). Machine learning models are powerful predictive tools and can use patient comorbidities and demographic factors to aid in patient selection. This study aimed to determine whether machine learning models can assist with preoperative patient selection for outpatient total hip arthroplasty. Elective THA cases during 2010-2018 from the American College of Surgeons National Quality Improvement Program (ACS-NSQIP) database were used to develop artificial neural network (ANN) models, which are powerful tools for making predictions. The models predicted whether THA patients had same-day discharge, short hospital length of stay (short LOS) (< 2 days), or prolonged length of stay (long LOS) (> 2 days). Variables significantly associated with the outcomes were identified with multiple linear regression and multivariate logistic regression analyses. 153,053 cases were included. 68,283 patients had long LOS, 32,190 had short LOS, and 2,621 patients had same-day discharge. The model for long LOS had an area under the ROC curve (AUC) of 0.868 and an accuracy of 80.7%. The model for short LOS had an AUC of 0.757 and an accuracy of 70.6%. The model for same-day discharge had an AUC of 0.814 and an accuracy of 78.8%. This is the first study to develop machine learning models that may aid in patient selection for outpatient THA by predicting whether a patient has a long LOS, short LOS, or same-day discharge. As outpatient THA becomes more common, preoperative patient selection will be important for optimizing THA.