Time Prediction Accuracy and Influencing Factors in Colorectal ESD by Trainees: A Mixed-Model Analysis

Abstract Purpose: Accurate prediction of procedure time is important for endoscopy unit efficiency and may support procedural planning in colorectal endoscopic submucosal dissection (ESD), especially when performed by trainees. However, factors influencing prediction accuracy are not well understood. Methods: This retrospective study analyzed 275 colorectal ESD cases performed by three trainees between May 2022 and January 2025. Trainees completed a preoperative “Strategy Note” that included predicted procedure time based on lesion assessment. The primary outcome was prediction error, defined as the difference between predicted and actual procedure time. A linear mixed model (LMM) was used to identify factors associated with prediction error. Case number, defined as the sequence of cases from each trainee’s first use of the Strategy Note, was included as a fixed effect along with lesion size, morphology, location, fold-crossing, and flexure involvement, with a random effect for each trainee. Results: The median predicted time was similar among trainees (30 min), while actual procedure time differed significantly (p = 0.008). LMM analysis identified tumor size > 3 cm (estimate: +15.43 min, p = 0.01), protruded morphology (+ 10.47 min, p = 0.03), and flexure involvement (+ 5.93 min, p = 0.04) as significant predictors of increased prediction error. Subgroup analysis revealed varying trends among trainees, even for lesions ≤ 3 cm. Conclusion Prediction accuracy in colorectal ESD varies among trainees and is significantly affected by lesion complexity. Tumor size, morphology, and flexure involvement should be considered in structured training programs to enhance procedural planning and operator preparedness.