A risk prediction model based on machine learning algorithm for parastomal hernia after permanent colostomy.

Area of Science:

• Colorectal cancer surgery
• Surgical oncology
• Machine learning in medicine

Background:

• Postoperative parastomal hernia (PSH) is a common complication after permanent colostomy in colorectal cancer patients.
• Early identification of patients at high risk for PSH is crucial for implementing preventive measures and improving patient outcomes.

Purpose of the Study:

• To develop and validate a machine learning-based risk prediction model for PSH in colorectal cancer patients undergoing permanent colostomy.
• To assist healthcare professionals, particularly nurses, in identifying high-risk patient groups for proactive PSH prevention strategies.

Main Methods:

• A case-control study involving 495 colorectal cancer patients with permanent colostomy and a 1-year follow-up.
• Development of multiple binary classification models (LR, SVC, KNN, RF, LGBM, XgBoost) using selected variables via LASSO regression.
• Model performance evaluation using AUC, specificity, sensitivity, accuracy, PPV, NPV, and F1-score, with clinical utility assessed by DCA and model interpretation via SHAP and nomogram.

Main Results:

• The Random Forest (RF) model achieved the highest predictive performance, with an AUC of 0.888, demonstrating superior discrimination.
• Key predictors identified by SHAP analysis include BMI, operation duration, COPD, prealbumin, TNM staging, stoma site, TRAM, CRP, ASA classification, and stoma diameter.
• The RF model showed the highest clinical net benefit on decision curve analysis, indicating its practical utility.

Conclusions:

• The developed Random Forest model provides a robust and clinically relevant tool for predicting PSH risk in patients with colostomies.
• This model facilitates the early identification of high-risk individuals, enabling timely implementation of preventive care and potentially reducing PSH incidence.