Abstract
BACKGROUND: Acute non-variceal upper gastrointestinal bleeding (ANVUGIB) is a common life-threatening emergency. Despite advances in endoscopic hemostasis, the 7-day rebleeding rate remains as high as 15-30%. Existing risk assessment tools show limited performance in Chinese populations, underscoring the need for a high-precision model tailored to local patients. OBJECTIVE: To develop, validate, and deploy an individualized model for predicting 7-day rebleeding risk in Chinese patients with ANVUGIB using early clinical information. METHODS: We retrospectively included 818 patients with ANVUGIB treated at the General Hospital of Central Theater Command between January 2020 and December 2023, randomly divided into a training cohort (n = 572) and an internal validation cohort (n = 246) at a 7:3 ratio. An additional 147 patients admitted between January and August 2024 were used as a temporally independent external validation cohort. Predictor variables were selected using least absolute shrinkage and selection operator (LASSO) regression, followed by multivariable logistic regression modeling. Model performance was evaluated using receiver operating characteristic (ROC) curves, calibration plots, and decision curve analysis (DCA), and generalizability was tested in the external validation cohort. RESULTS: Five independent predictors were identified: syncope, pulse rate, red cell distribution width, serum albumin, and bowel sounds. The prediction model incorporating these variables achieved areas under the curve (AUCs) of 0.843 (95% CI 0.784-0.903), 0.833 (95% CI 0.742-0.924), and 0.825 (95% CI 0.700-0.950) in the training, internal validation, and external validation cohorts, respectively. Calibration plots and decision curve analysis confirmed good consistency and clinical utility. CONCLUSION: We developed and validated a 7-day rebleeding risk prediction model for ANVUGIB in a Chinese emergency department population. The model outperformed existing scoring systems, and deployment as a Shiny-based web tool enables early risk identification and individualized decision-making in emergency care.