Machine Learning to Predict Discharge Destination After Total Knee Arthroplasty and Total Hip Arthroplasty

Discharge destination impacts costs and perioperative planning for primary total knee (TKA) or hip arthroplasty (THA). The purpose of this study was to create a tool to predict discharge destination in contemporary patients. Models were developed using more than 400,000 patients from the National Surgical Quality Improvement Program database. Models were compared with a previously published model using area under the receiver operating characteristic curve (AUC) and decision curve analysis (DCA). AUC on patients with TKA was 0.729 (95% confidence interval [CI]: 0.719 to 0.738) and 0.688 (95% CI: 0.678 to 0.697) using the new and previous models, respectively. AUC on patients with THA was 0.768 (95% CI: 0.758 to 0.778) and 0.726 (95% CI: 0.714 to 0.737) using the new and previous models, respectively. DCA showed substantially improved net clinical benefit. The new models were integrated into a web-based application. This tool enhances clinical decision making for predicting discharge destination following primary TKA and THA. (Journal of Surgical Orthopaedic Advances 32(4):252- 258, 2023)

Key words: total knee arthroplasty, total hip arthroplasty, machine learning, artificial intelligence, clinical risk prediction