Development of machine learning prediction models for postoperative outcomes in adult male circumcision

Comprehensive Summary

Shpaner et al. developed machine learning models to predict short-term postoperative complications after adult male circumcision. The analysis included 194 adult patients treated at a single center in Milan between 2023 and 2024. Bleeding, edema, pain, or infection occurring within seven days of surgery were considered as complications; these complications occurred in about 30% of patients. Three models were tested: logistic regression, random forest, and support vector machine (SVM). Among them, the SVM performed best, achieving an AUC of 0.907 and the lowest Brier score, indicating strong discrimination and calibration. Most complications occurred in the traditional circumcision group (57 of 58 cases), with only one event in the laser group. SHAP analysis showed that intraoperative blood loss and surgical technique were the strongest predictors of complications, followed by age and diabetes status.

Outcomes and Implications

Adult male circumcision is widely performed for various reasons, including infection control and HIV prevention. Although generally considered a low-risk procedure, early complications such as bleeding and edema can still affect recovery and patient satisfaction. This study shows that complication risk is strongly influenced by surgical technique and intraoperative blood loss, with traditional methods accounting for nearly all early adverse events in this cohort. If validated further, a predictive model like the SVM could help surgeons identify higher-risk patients before discharge and tailor postoperative monitoring or counseling accordingly. This may be especially relevant in outpatient or resource-limited settings where follow-up can be inconsistent. Regardless, because the laser group had very few complications and the data came from a single center, broader validation is necessary before applying the model more widely.