Supplementary file 1_Postoperative recurrence prediction model for perianal abscess using machine learning algorithms.docx

ObjectiveThis study aimed to develop a machine learning–based model to predict recurrence risk after perianal abscess surgery, thereby supporting personalized follow-up and intervention strategies.
MethodsClinical data were collected from patients with perianal abscess who underwent surgery at Nanjing Hospital of Chinese Medicine, Affiliated to Nanjing University of Chinese Medicine between January 2022 and June 2023. Significant predictors were identified using the least absolute shrinkage and selection operator (LASSO) algorithm combined with multivariate logistic regression. The Synthetic Minority Over-sampling Technique (SMOTE) was applied to balance class distribution, and several machine learning (ML) algorithms were employed for model construction. Model performance was evaluated by the area under the receiver operating characteristic curve (AUC), sensitivity, specificity, and accuracy. Model calibration was assessed using calibration curves. The effectiveness was evaluated through Decision Curve Analysis (DCA). Finally, the SHapley Additive ExPlanations (SHAP) were used to interpret the best-performing model and quantify the contribution of each predictor to its predictions.
ResultsA total of 737 patients with perianal abscess were included in the study. A history of diabetes, abscess space, and the aggregate index of systemic inflammation (AISI) were identified as the three strongest predictors of recurrence. Among all evaluated models, the CatBoost model showed the highest discriminatory power in the training set (AUC = 0.821, 95% CI: 0.777–0.864), validation set (AUC = 0.744, 95% CI: 0.616–0.872), and temporal validation set (AUC = 0.735, 95% CI: 0.649–0.821).
ConclusionThe machine learning–based model effectively identifies patients at high risk of recurrence after perianal abscess surgery. The CatBoost model achieved the best predictive performance, while SHAP analysis enhanced interpretability, supporting individualized patient management.