Establishment and Validation of Risk Prediction Models for Postoperative Pain After Endoscopic Submucosal Dissection: A Retrospective Clinical Study

Shanshan Wu,Shuren Wang,Yonghong Ding,Zongwang Zhang
DOI: https://doi.org/10.2147/JMDH.S470204
2024-08-12
Abstract:Objective: Postoperative pain is a common complication in endoscopic submucosal dissection (ESD) patients. This study aimed to develop and validate predictive models for postoperative pain associated ESD. Methods: We retrospectively constructed a development cohort comprising 2162 patients who underwent ESD at our hospital between January 2015 and April 2022. The dataset was randomly divided into a training set (n = 1541) and a validation set (n = 621) in a 7:3 ratio. The bidirectional stepwise regression with Akaike's information criterion (AIC) and multivariate logistic regression analysis were used to screen the predictors of post-ESD pain and construct three nomograms. We evaluated the model's discrimination, precision and clinical benefit through receiver operating characteristic (ROC) curves, calibration plots, Hosmer-Lemeshow (HL) goodness-of-fit test and decision curve analysis (DCA) in internal validation. Results: The proportion of patients developing postoperative pain in the training and testing data set was 25.6% and 28.5%, respectively. Three nomograms were constructed according to the final logistic regression models. The clinical prediction models for preoperative risks, preoperative and intraoperative risks, and perioperative risks consisted of seven, nine and six independent predictors, respectively, after bidirectional stepwise elimination. The models demonstrated the AUC of 0.794 (95% CI 0.768-0.820), 0.823 (95% CI 0.799-0.847) and 0.817 (95% CI 0.792-0.842) in the training cohort and 0.702 (95% CI 0.655-0.748), 0.705 (95% CI 0.659-0.752) and 0.747 (95% CI 0.703-0.790) in the validation cohort. The calibration plot, HL and DCA demonstrated the model's favorable clinical applicability. Conclusion: We developed and validated three robust nomogram models, which might identify patients at risk of post-ESD pain and promising for clinical applications.
What problem does this paper attempt to address?