Constructing and validating a risk prediction model for postoperative bleeding after colorectal EMR in the Chinese
Bingfeng He1,2, Jiawei Zhang2,3, Mingli Su2,3
1Wuzhou Medical College, Xuzhou, China.
Surgical Endoscopy
|November 17, 2025
Summary
A new machine learning model accurately predicts post-endoscopic mucosal resection (EMR) bleeding risk in Chinese patients. This tool aids clinical decisions, improving outcomes after colorectal EMR procedures.
Area of Science:
- Gastroenterology
- Medical Informatics
- Machine Learning
Background:
- Endoscopic mucosal resection (EMR) is a key treatment for early colorectal lesions.
- Clinically significant post-EMR delayed bleeding (CSPEB) is a frequent complication.
- Accurate prediction of CSPEB is crucial for patient management.
Purpose of the Study:
- To develop and validate a machine learning model for predicting CSPEB after colorectal EMR.
- To identify key risk factors for post-EMR bleeding.
- To create a clinically applicable tool for risk assessment.
Main Methods:
- Retrospective analysis of 3888 patients undergoing colorectal EMR (Jan 2018-Sep 2024).
- External validation using 1000 patients from a separate institution (2022-2024).
- Development of a random forest model with weighted sampling, evaluated using ROC, calibration plots, and DCA. SHAP values used for interpretability.
Main Results:
- CSPEB incidence was 1.4%.
- Identified risk factors include sigmoid location, lesion size/number, APTT, fibrinogen, and hemoclip count.
- The model achieved AUC of 0.87 (internal) and 0.80 (external validation), demonstrating strong predictive performance and clinical utility.
Conclusions:
- The first machine learning model for predicting post-EMR bleeding in the Chinese population was developed and validated.
- The model offers accurate and interpretable risk predictions.
- A web-based application integrates the model to support clinical decision-making and enhance patient outcomes.
