Novel indicator for colonoscopy insertion difficulty and its application in predictive modelling: a multicentre...

Area of Science:

• Gastroenterology and Endoscopy
• Medical Informatics and Machine Learning

Background:

• Colonoscopy is crucial for colorectal disease screening and diagnosis.
• Current measures of colonoscopy insertion difficulty, like cecal intubation time (CIT), are influenced by non-patient factors, limiting objectivity.
• Developing a standardized metric is essential for accurate assessment of patient-related procedural difficulty.

Purpose of the Study:

• To develop and validate a standardized colonoscopy insertion time (SCIT) metric.
• To identify factors associated with colonoscopy insertion difficulty.
• To develop machine learning models for predicting insertion difficulty.

Main Methods:

• A multicenter prospective observational study involving approximately 3000 adults undergoing sedated colonoscopy.
• Application of three standardization approaches (Z-score, median-based, min-max normalization) to derive SCIT.
• Use of mixed-effects models and machine learning techniques for analysis and prediction.

Main Results:

• The optimal SCIT standardization method will be selected based on its ability to minimize variability from operator, equipment, and center factors.
• Machine learning models will be developed to predict insertion difficulty using preprocedural clinical variables.
• Model performance will be rigorously evaluated using discrimination, calibration, and predictive accuracy measures.

Conclusions:

• The developed SCIT metric is expected to provide a more objective and patient-centered measure of colonoscopy insertion difficulty.
• Identification of predictive factors and development of machine learning models will aid in anticipating and managing procedural challenges.
• This research will enhance colonoscopy performance evaluation, potentially improving procedural efficiency and patient outcomes.