Document resource
Background Accurate early risk stratification in non-variceal upper gastrointestinal bleeding (NVUGIB) is essential for guiding management and resource utilization. We aimed to compare the predictive performance of the ABC (Age, Blood tests, Comorbidities), AIMS65, Glasgow-Blatchford Score (GBS), and Rockall scores for mortality and other clinically relevant outcomes in a tertiary care setting.Methods This retrospective cohort study included 270 patients with confirmed NVUGIB. Patient flow and exclusions are shown in Figure 1 ( IDDF2026-ABS-0374 Figure 1. Flow diagram showing patient selection for the study). Risk scores were calculated at admission using clinical, biochemical, and endoscopic parameters. The primary outcome was 30-day mortality; secondary outcomes included in-hospital mortality, 30-day rebleeding, and transfusion requirements. Predictive performance was assessed using the area under the receiver operating characteristic curve (AUROC), with comparisons by the DeLong method.Results The mean age was 66.6 years, with a 30-day mortality of 5.9%. Elevated blood urea, hypoalbuminemia, renal dysfunction, higher ASA class, and altered mental status were significantly associated with mortality (p < 0.05). The ABC and AIMS65 scores showed excellent, comparable performance for 30-day mortality (AUROC 0.858 [95% CI 0.74–0.98] vs 0.850 [95% CI 0.67–1.00]; p = 0.878) ( IDDF2026-ABS-0374 Figure 2. ROC curve for 30-day mortality). DeLong comparisons (IDDF2026-ABS-0374 Table 1) confirmed similar performance between ABC and AIMS65, both significantly outperforming Rockall. Mortality increased progressively with higher scores across all systems, with the best risk separation seen in ABC and AIMS65 (IDDF2026-ABS-0374 Figure 3. Mortality risk comparison with score type). AIMS65 showed the highest AUROC for 30-day rebleeding (0.753), while GBS performed best for predicting transfusion requirements (AUROC 0.790).Conclusions ABC and AIMS65 demonstrate comparable and superior performance for mortality prediction in NVUGIB, both significantly outperforming Rockall. AIMS65 also shows good predictive ability for rebleeding, while GBS is most effective for predicting transfusion requirements. These findings support the use of ABC and AIMS65 for early prognostic assessment, with GBS complementing decision-making for therapeutic intervention.Abstract IDDF2026-ABS-0374 Figure 1Abstract IDDF2026-ABS-0374 Figure 2Abstract IDDF2026-ABS-0374 Figure 3Abstract IDDF2026-ABS-0374 Table 1DeLong comparison of AUROCs across clinical outcomesClinical OutcomeScoring SystemAUROCp-valueIn-hospital MortalityABC0.823AIMS650.8090.8436GBS0.7260.3497ROCKALL0.5090.0037*30-day MortalityABC0.858AIMS650.8500.8783GBS0.7390.1419ROCKALL0.4830.0001*30-day RebleedingAIMS650.753ABC0.7130.6120GBS0.7010.6816ROCKALL0.4510.0009*Blood TransfusionGBS0.790AIMS650.6700.0069*ROCKALL0.6490.0055*ABC0.6280.0003*