SPEED‐EMR Score: A Simple Predictive Model for Estimating Colorectal Endoscopic Mucosal Resection Duration

ABSTRACT Introduction EMR is widely used for large and complex nonpedunculated superficial colorectal lesions, but objective tools to predict procedure duration are lacking. We aimed to develop a simple clinical model to estimate EMR duration: the SPEED‐EMR Score. Methods Retrospective cohort study of consecutive conventional EMRs of nonpedunculated colorectal lesions ≥ 15 mm performed at a tertiary referral center between 2020 and 2025. EMR duration was defined from the first lesion image to final post‐resection documentation. Predictors of duration were identified using multivariable linear regression and converted into an integer‐based score. Model performance and discrimination for predefined time‐based complexity categories were evaluated. Results A total of 1077 EMRs were analyzed. Median lesion size was 20 mm (IQR 15–25), and median EMR duration was 13 min (IQR 7–24). Independent predictors of longer duration were lesion size, flat/depressed morphology, NICE type 2 pattern, difficult anatomical location, and recurrent lesion at a previously treated site (all p ≤ 0.002). These variables composed the SPEED‐EMR Score (0–15 points). Predicted duration was calculated as 3.23 + 3.94 × SPEED‐EMR score. The final model showed R 2 = 0.445. Discrimination was good for low‐complexity (≤ 20 min; AUC: 0.77) and high‐complexity (≥ 40 min; AUC: 0.75) procedures. Conclusions SPEED‐EMR is a novel, bootstrap internally validated tool for objective prediction of colorectal EMR duration. It may support optimized scheduling, workflow efficiency, and resource allocation in therapeutic endoscopy services.