Predicting short-term major postoperative complications in intestinal resection for Crohn's disease: A machine...

Area of Science:

• Gastroenterology
• Surgical Oncology
• Medical Informatics

Background:

• Crohn's disease (CD) patients undergoing intestinal resection (IR) face high rates of postoperative complications.
• Predicting these complications is crucial for surgical decision-making and patient outcomes.

Purpose of the Study:

• To develop and validate machine learning (ML) models for predicting short-term major postoperative complications in CD patients post-IR.
• To identify key predictive factors for these complications.

Main Methods:

• Retrospective analysis of 259 CD patients who underwent IR (2017-2022).
• Development of logistic regression and random forest (RF) models using a training cohort.
• Validation of model performance using AUC in a separate cohort.

Main Results:

• 5.0% of patients experienced major postoperative complications (Clavien-Dindo ≥ III) within 30 days.
• The RF model achieved a higher AUC (0.965) compared to the logistic model (0.916).
• Key predictors included preoperative CD activity index (CDAI) ≥ 220, low serum albumin, and extended operation time.

Conclusions:

• Both logistic regression (via nomogram) and RF models effectively predict short-term major complications after IR in CD.
• The RF model demonstrated superior predictive performance.
• Preoperative CDAI, serum albumin, and operation time are critical factors for risk stratification and personalized perioperative care.