Superiority of a preparation‐related model for predicting inadequate bowel preparation in patients undergoing colonoscopy: A multicenter prospective study
Long Chen, Gui Ren, Hui Luo, Linhui Zhang, Limei Wang, Jianghai Zhao, Rongchun Zhang, Xiaoying Zhang, Xiaoyu Kang, Yanglin Pan
Abstract
Abstract Background and Aim Three models based on patient‐related factors have been developed to predict inadequate bowel preparation (BP). However, the performance of the models seems suboptimal. This study aimed to develop a novel preparation‐related model and compare it with the available patient‐related models. Methods Patients receiving standard BP were prospectively enrolled from five endoscopic centers. Patient‐related and preparation‐related factors for inadequate BP (defined by segmental Boston Bowel Preparation Scale score < 2) were identified by logistic regression. A preparation‐related model was derived and internally validated in 906 patients. The comparisons of models were assessed by discrimination and calibration. The preparation‐related model was also externally validated. Results Several patient‐related factors (male and American Society of Anesthesiologists Physical Status Classification System score ≥ 3) and preparation‐related factors (drinking‐to‐stool interval ≥ 3 h, preparation‐to‐colonoscopy interval ≥ 6 h, and poor rectal effluent) were found to be independently associated with inadequate BP (all P < 0.05). C ‐statistics was 0.81 for the preparation‐related model in the training cohort ( n = 604), significantly higher than three available patient‐based models (0.58–0.61). Similar results were observed in the validation cohort ( n = 302). Calibration curves showed close agreement in the preparation‐related model ( R 2 = 0.315 in the training cohort and 0.279 in the validation cohort). The preparation‐related model was externally validated in another 606 patients with C ‐index of 0.80. Conclusions A new preparation‐related model (consisting of drinking‐to‐stool interval ≥ 3 h, preparation‐to‐colonoscopy interval ≥ 6 h, and poor last rectal effluent) was developed and performed better than three available patient‐related models. This easy‐to‐use model may be a useful decision‐support tool on individualized plans in patients undergoing BP.