Comparison and Validation of Prognostic Scoring Systems in Risk Stratification of Patients with Acute Non-variceal Upper Gastrointestinal Bleeding

Abstract Multiple prognostic scoring systems exist for risk stratification in non-variceal upper gastrointestinal bleeding (NVUGIB) but the optimal tool remains unclear. This study compared six scoring systems for predicting clinical outcomes in NVUGIB. This single-center retrospective analysis of 303 acute NVUGIB patients compared Glasgow Blatchford score (GBS), AIMS65, MAP(ASH), age, blood tests and comorbidities (ABC), H3B2, and Nagoya university score (N-score) at admission. Discriminative performance was assessed using area under the receiver operating characteristic curve (AUROC) analysis against clinical outcomes. Binary logistic regression was performed to identify predictors of endoscopic intervention and blood transfusion. MAP(ASH) achieved the highest accuracy for mortality (AUROC 0.917). For rebleeding, MAP(ASH) was also superior (AUROC 0.795) followed by GBS (AUROC 0.773). However, for endoscopic intervention, H3B2 demonstrated the best performance (AUROC 0.736), followed by N-score (AUROC 0.731). For blood transfusion, GBS (AUROC 0.877) and MAP(ASH) (AUROC 0.864) were superior. Multivariate analysis identified hematemesis as an independent predictor of endoscopic intervention and pulse rate as one of the strong predictors for the need of blood transfusion. MAP(ASH), ABC, and AIMS65 perform well for mortality risk assessment, while GBS and MAP(ASH) are best for predicting transfusion requirements. All pre-endoscopy scores have only moderate accuracy for predicting rebleeding and need for endoscopic intervention.