·A total of 686 patients with ANVUGIB admitted to the First Affiliated Hospital of Soochow University from January 1, 2016 to December 31, 2019 were collected. The general conditions, clinical characteristics, laboratory results, endoscopic findings and prognosis were recorded. The Baylor score, Rockall score, Glasgow Blatchford Score (GBS), Cedars-Sinai Medical Center Predictive Index(CSMCPI), AIMS65 and MAP (ASH) for each patient were calculated. The collected data were randomly divided into a training set (n=481) and a validation set (n=205). A new prediction model for rebleeding was established by binary Logistic regression based on the training set data. The discrimination and calibration of the training and validation set were evaluated respectively, and then the model was transformed into a risk scoring system. The comparisons between the receiver operating characteristics (ROC) curves were based on the Delong test.

·Binary Logistic regression analysis showed that low systolic blood pressure, low hemoglobin level, American Society of Anesthesiologists (ASA) grade>3, and endoscopic findings of clots, visible vessels and active bleeding were independent risk factors for rebleeding. The area under the curve (AUC) of the prediction model was 0.892 (95% CI 0.838?0.946, P=0.001), Hosmer-Lemeshow test P=0.934. The validation set AUC=0.915 (95% CI 0.851?0.980, P=0.001), Hosmer-Lemeshow test P=0.871. The scoring system (AUC=0.882, 95% CI 0.823?0.942) was better at predicting rebleeding than all other scores mentioned except MAP (ASH) score (P<0.05).

·The scoring system could predict rebleeding after ANVUGIB, which could be an option in clinical practice.