Insights From Inputs: Enhancing Revision Total Joint Arthroplasty Resource Allocation With Machine Learning Prediction

Comprehensive Summary

This study evaluates the accuracy of artificial neural networks (ANNs) in predicting duration of surgery (DOS), length of stay (LOS), and 30-day hospital readmission for patients undergoing revision total joint arthroplasty (rTJA). Researchers developed these models using preoperative factors from the American College of Surgeons National Surgical Quality Improvement Program (NSQIP) and a single-institution dataset, comparing their performance against traditional regression models and historical averages. The findings demonstrate that institutional models were superior for predicting DOS, achieving up to 76.2% accuracy for knee revisions, while national models better predicted LOS due to more comprehensive patient comorbidity data. For 30-day readmissions, the national models achieved modest area under the curve (AUC) scores of approximately 0.59. Importantly, all machine learning and regression models significantly outperformed the use of historical mean values, which is the current standard for many hospital scheduling practices. Local data is critical for timing surgical procedures, whereas national data offers better generalizability for identifying patient-related factors that influence hospital stays. Additionally, the use of Shapley Additive exPlanations (SHAP) values provides model interpretability, allowing clinicians to understand which specific patient features, such as surgeon rolling averages or comorbidities, drive the predictions.

Outcomes and Implications

This research is critical because revision arthroplasties are technically complex, resource-intensive procedures with a projected drastic increase over the coming decades. The study is clinically relevant for optimizing OR scheduling and managing hospital bed capacity, which helps prevent postoperative overcrowding and reduces financial burden. Because these predictive tools are already more accurate than the common practice of using historical averages, implementation by hospital administrators is highly achievable and beneficial for immediate use.