Machine Learning Predicts the Need for Surgical Intervention in Adhesive Small Bowel Obstruction

Area of Science:

• Gastroenterology
• Surgical Oncology
• Medical Informatics

Background:

• Adhesive small bowel obstruction (ASBO) is a common surgical emergency.
• Predicting the need for surgical intervention in ASBO is crucial for timely management.
• Current non-operative management (NOM) strategies have limitations in predicting failure.

Purpose of the Study:

• To evaluate the predictive performance of machine learning (ML) algorithms for surgical intervention in ASBO.
• To identify optimal timing for transitioning ASBO patients to surgery.
• To compare ML algorithms with traditional logistic regression for predicting surgical need.

Main Methods:

• Retrospective study of 106 ASBO patients treated with long transnasal intestinal tube (LT) decompression.
• Application of traditional logistic regression and five ML algorithms (including random forest) for risk prediction.
• Analysis of clinical variables such as drainage volume, diagnosis-to-intubation interval, and small bowel dilatation.

Main Results:

• Non-operative management failed in 26% of patients.
• Key predictors for surgery included high day 1 drainage volume, short diagnosis-to-intubation interval, and small bowel dilatation at 48 hours.
• The random forest algorithm achieved the highest predictive performance (AUC 0.889, accuracy 0.864).

Conclusions:

• Machine learning algorithms demonstrate significant potential in predicting surgical intervention for ASBO.
• Findings support the development of a framework for earlier clinical decision-making in ASBO management.
• The study highlights the utility of ML in optimizing surgical timing for ASBO patients.