A modified risk model for upper gastrointestinal bleeding: comparative evaluation of the H3B2A1 score

Abstract

Background: Acute upper gastrointestinal bleeding (UGIB) remains a critical emergency with significant morbidity and mortality despite advances in diagnostic and therapeutic approaches. Risk stratification is essential for the early identification of high-risk patients, guiding clinical decisions, and optimizing resource allocation. The H3B2A1 score is a modified version of the H3B2 score that incorporates albumin as an additional parameter. This study aimed to evaluate the predictive performance of the H3B2A1 score in comparison to established risk models, including AIMS65, GlasgowBlatchford Score (GBS), and modified GBS (mGBS), in assessing mortality and intensive care unit (ICU) admission in UGIB patients. Methods: In this retrospective study, 233 patients with UGIB confirmed by endoscopic evaluation at a tertiary care center were included. The predictive accuracy of the H3B2A1 score for mortality and ICU admission was assessed and compared with AIMS65, GBS, mGBS, and H3B2 scores. Receiver operating characteristic (ROC) curve analysis was performed to determine the area under the curve (AUC), cutoff values, sensitivity and specificity for each scoring system. Results: The H3B2A1 score demonstrated comparable accuracy in predicting mortality (AUC: 0.750, sensitivity 88%) to the AIMS65 score (AUC: 0.754, sensitivity 88%), with both scores exhibiting moderate predictive power (AUC = 0.70–0.90). In predicting ICU admission, the AIMS65 score had the highest predictive accuracy (AUC: 0.844, sensitivity 96.1%), followed by H3B2A1 (AUC: 0.645; sensitivity 64.7%) and H3B2 (AUC: 0.624; sensitivity 82.4%). Incorporating albumin in the H3B2A1 score improved its prognostic performance compared to the original H3B2 model. Conclusions: The H3B2A1 score is a practical and effective tool for risk stratification in UGIB patients, demonstrating moderate predictive ability for mortality and ICU admission. Its simplicity and clinical applicability make it valuable to existing risk assessment models. Further prospective studies are warranted to validate its utility in diverse patient populations.