Introduction Revision total hip arthroplasty (rTHA) for acetabular component failure is complicated by variable acetabular bone loss. The Paprosky classification system guides reconstruction based on defect severity and location, but the decision to use acetabular augments remains largely subjective. Machine learning (ML) offers a data-driven approach to predict surgical decision-making by identifying nonlinear relationships among patient and defect characteristics. This proof-of-concept study validates a novel ML modeling framework designed to predict augment usage in rTHA from patient demographics and defect characteristics. Materials and Methods Following IRB approval (IRB#: 853927), a retrospective review of 427 consecutive aseptic acetabular revision cases (2012–2025) performed by a single arthroplasty surgeon was conducted. Model features included patient demographics, Paprosky defect classification, and presence of a chronic pelvic discontinuity. Ten ML algorithms were trained using stratified 5-fold cross-validation and grid-search optimization. Model performance was assessed by area under the receiver operating characteristic curve (AUC-ROC), accuracy, F1-score, and Brier score. Calibration and decision curve analyses were performed. Feature importance and directionality analyses were conducted for the top-performing model. Results Of 427 cases, 94 (22.0%) required augments. Augment recipients were older (median 64.9 vs. 61.7 years, p = 0.004). Paprosky II/III defects and presence of a chronic pelvic discontinuity were significantly more prevalent in augment cases (all p < 0.05). The bagging ensemble classifier achieved the best discrimination (AUC-ROC = 0.82, accuracy = 87.2%, F1 = 0.73, Brier score = 0.12). Decision curve analysis demonstrated greater net benefit compared to “Treat All” and “Treat None” strategies. Paprosky defect type and a chronic pelvic discontinuity were the strongest predictors, while demographic variables had minimal influence. Conclusion This study demonstrates the feasibility and accuracy of an ML framework for predicting acetabular augment use in rTHA, establishing a foundation for future models that support pre-operative decision-making in complex revision surgery.
Turlip et al. (Thu,) studied this question.