Predicting Non-Home Discharge Following Primary Total Hip Arthroplasty Using an Artificial Neural Network to Identify the Most Important Variables.

Background: Total hip arthroplasty (THA) has continued to increase in incidence. With expanding focus on value-based care and adoption of bundled payments at some facilities, cost containment measures, including discharging patients to home as soon as possible postoperatively, have gained attention. Therefore, it is important to determine patient characteristics associated with discharge disposition. The purpose of this study was to develop and utilize an artificial neural network (ANN) model to determine the most important non-modifiable and modifiable factors that can predispose patients to be discharged to a non-home destination following primary THA..

Methods: The National Surgical Quality Improvement Program database was used to identify patients who underwent primary THA from 20162019. Demographic, comorbidity, preoperative, and intraoperative variables were analyzed in this study. Statistically significant variables, with a p-value < 0.05, were inputted into the ANN model.

Results: In total, 124,691 patients were analyzed in the study, of which 19,275 (15.5%) were discharged to a non-home destination. The ANN reached a Receiver Operating Characteristic (ROC) Area-Under-the-Curve (AUC) of 0.793. The five most important variables which helped to predict non-home discharge following primary THA were age, operative time, preoperative hematocrit, functional status, and preoperative international normalized ratio (INR).

Conclusion: The present study used an ANN model and identified several significant factors which can help predict patients being discharged to a non-home location following primary THA.

Clinical relevance: Clinicians should be aware of these variables and explore reductions to nonhome discharges through preoperative patient optimization, where possible.