Establish and validate a clinical predictive model for pediatric gastrointestinal endoscopy.

AIM: Determining appropriate endoscopic referrals in children is challenging due to the lack of validated prediction tools. This study aimed to develop and validate a predictive model to support decision-making in pediatric endoscopy. METHODS: This single-center retrospective study included hospitalized children aged 2-14 years who underwent first-time gastrointestinal endoscopy at the First Affiliated Hospital of Kunming Medical University between January 2009 and June 2023. Demographic characteristics, clinical presentations, laboratory results, imaging examinations, and endoscopic findings were collected. Patients were randomly assigned to training (70%) and validation (30%) cohorts. Candidate variables were selected using univariate analysis followed by the least absolute shrinkage and selection operator (LASSO) regression. A multivariate logistic regression model was then constructed. Model performance was evaluated using receiver operating characteristic (ROC) curve analysis, accuracy, and decision curve analysis (DCA). RESULTS: A total of 1,425 children were included, of whom 64.9% had abnormal endoscopic findings. Three prediction models for gastrointestinal endoscopy (GIE), esophagogastroduodenoscopy (EGD), and colonoscopy achieved AUCs of 0.673, 0.686, and 0.701 in the training cohort, and 0.623-0.627 in the validation cohort. All models demonstrated modest predictive performance and clinical net benefit. CONCLUSIONS: We developed and internally validated three prediction models using routine clinical data to estimate the risk of positive endoscopic findings in children. These tools may support risk stratification and referral decisions, but prospective external validation is required before clinical use.