Construction of a prediction model for rebleeding in patients with acute upper gastrointestinal bleeding

Abstract

Abstract Background The incidence of rebleeding in patients with upper gastrointestinal bleeding (UGIB) remains despite advances in intervention approaches. Therefore, early prediction of the risk of rebleeding could help to greatly reduce the mortality rate in these patients. We aim to develop and validate a new prediction model to predict the probability of rebleeding in patients with AUGIB. Methods A total of 1170 AUGIB patients who completed the procedure of emergency gastroscopy within 48 h of admission were included. Logistic regression analyses were performed to construct a new prediction model. A receiver operating characteristic curve, a line graph, and a calibration and decision curve were used to assess the predictive performance of our new prediction model and compare its performance with that of the AIMS65 scoring system to determine the predictive value of our prediction model. Results A new prediction model was constructed based on Lactic acid (LAC), neutrophil percentage (NEUTP), platelet (PLT), albumin (ALB), and D-DIMER. The AUC values and their 95% confidence interval (CI) for the new prediction model and the AIMS65 score were 0.746 and 0.619, respectively, and 0.697–0.795 and 0.567–0.670, respectively. In the training group, the C index values based on the prediction model and the AIMS65 scoring system were 0.720 and 0.610, respectively. In the validation group, the C index values based on the prediction model and the AIMS65 scoring system were 0.828 and 0.667, respectively. The decision and calibration curve analysis also showed that the prediction model was superior to the AIMS65 scoring system in terms of accuracy of prediction, consistency, and net clinical benefit. Conclusion The prediction model can predict the probability of rebleeding in AUGIB patients after endoscopic hemostasis therapy.

Keywords

• Acute upper gastrointestinal bleeding
• Rebleeding
• AIMS65 score
• Nomogram
• Predictive model