A Discussion on The Utility of Discharge Location Prediction Models for Total Joint Arthroplasty Surgery

Abstract

Objective. Total Joint Arthroplasty (TJA) remains one of the highest frequency elective surgical procedures. Medicare is expected to spend close to $ 50 Billion on TJAs by 2030. Predicting discharge location could allow for cost mitigation and the ability to set appropriate expectations for patients preoperatively. Our aim is to determine the validity of one predictive model. We hypothesize that this tool will demonstrate comparable predictive value as in the pilot study. Methods. We conducted a cross-sectional study of unilateral, primary, total joint replacements from January 2020 through February 2021. Nine variables were input into a predictive model at https://dukeriskcalculators.shinyapps.io/Dispo/ and percent likelihood of discharge to SNF/rehabilitation facility was recorded and analyzed. Receiver operating characteristics curve (ROC) analysis was utilized to eval-uate the model’s predictive capability. Results. Our cohort consisted of 264 patients. 9.1% of patients were discharged to an SNF/rehabilitation facility. ROC analysis demonstrated an area under the curve (AUC) of 0.72 indicating good predictive value. The mean percent likelihood of discharge to SNF/rehab was 31% ± 9% (mean ± 95% confidence interval) for patients whose final discharge location was an acute rehabilitation facility; The mean percent likelihood of discharge to SNF/rehab was 15% ± 2% for patients who were discharged home. Conclusions. The predictive model analyzed uses nine easily accessible variables and demonstrates good predictive capability. Discharge to acute rehabilitation is a costly and often unnecessary intervention. Predictive models can reduce discharge to these facilities, reduce healthcare costs, and improve patient outcomes.