Machine learning-based prediction of meniscal tears in ACL reconstruction using BMI, time to surgery, injury mechanism, and Tegner activity score: A temporally validated decision tool.

Purpose: Concomitant arthroscopically confirmed meniscal tears are common in patients undergoing anterior cruciate ligament (ACL) reconstruction and can influence intraoperative planning and postoperative rehabilitation. Robust tools for preoperative, individualized risk stratification remain limited. The objective of this study was to develop, temporally validate, and implement a clinically interpretable preoperative prediction tool for concomitant meniscal tears in ACL reconstruction. Methods: A retrospective analysis was conducted on 649 consecutive patients undergoing primary arthroscopic ACL reconstruction. Ten candidate machine-learning algorithms were developed using routinely available preoperative variables. Model selection was performed via five-fold cross-validation in the development cohort. The selected model was evaluated in an internal validation set and an independent temporal validation cohort (comprising patients treated in a subsequent period to assess model stability). Discrimination (area under the receiver operating characteristic curve, AUC), calibration, and clinical utility (decision curve analysis) were assessed. Model interpretability was examined using SHapley Additive exPlanations (SHAP). An open-access web calculator was created for point-of-care use. Results: Logistic regression using four routinely available preoperative predictors (body mass index, time from injury to surgery, injury mechanism and preoperative Tegner activity score) provided the most reliable performance. Discrimination remained consistent across cohorts (AUC 0.845 in training, 0.850 in internal validation, and 0.840 in temporal validation), with acceptable calibration. Decision curve analysis demonstrated a favourable net benefit across clinically relevant threshold probabilities. SHAP analyses supported the relative contribution and direction of effects of the four predictors. The final model was deployed as a web-based calculator. Conclusions: An accurate, interpretable, and temporally validated preoperative prediction model for concomitant meniscal tears in ACL reconstruction was developed. By integrating four routine clinical variables into an online calculator, this tool may enhance surgical planning and inform shared decision-making prior to ACL reconstruction. Level of Evidence: Level IV, retrospective cohort study.