Comparative performance of pre- and post-endoscopic risk scores for outcome prediction in endoscopy-confirmed upper gastrointestinal bleeding in the emergency department

Upper gastrointestinal bleeding (UGIB) requires early risk stratification, but commonly used scores were developed for different clinical moments and endpoints. We compared albumin, international normalized ratio, mental status, systolic blood pressure, age ≥65 years (AIMS65), Glasgow-Blatchford score (GBS), pre-Baylor, full Baylor, pre-Rockall, and full Rockall for the prediction of 30-day mortality, red-cell transfusion, endoscopic therapy, and rebleeding in an endoscopy-confirmed emergency department cohort. We retrospectively evaluated consecutive adults with suspected UGIB who underwent endoscopy and had an upper gastrointestinal bleeding source confirmed in the emergency department pathway between January 1, 2016 and 1 June 1, 2019. Thirty-day mortality was ascertained from hospital records and national death records. Rebleeding was defined as recurrent hematemesis and/or melena, or hemodynamic instability with rebleeding confirmed on repeat endoscopy, as documented in the available follow-up records. Score discrimination was assessed with receiver operating characteristic analysis. The analytic cohort comprised 347 patients (mean age 65.54 ± 15.82 years; 62.6% male). Index-hospital disposition included direct discharge after endoscopy in 27 patients (7.8%), discharge after hospitalization in 282 (81.3%), and in-hospital death in 38 (11.0%); the separate 30-day mortality endpoint used for receiver operating characteristic analysis comprised 46 deaths (13.3%). Blood transfusion occurred in 205 patients (59.1%), endoscopic therapy in 29 (8.4%), and rebleeding in 71 (20.5%). AIMS65 showed the highest area under the curve (AUC) for 30-day mortality (0.724, 95% confidence interval [CI] 0.647–0.802). GBS showed the highest AUC for transfusion (0.784, 95% CI 0.734–0.834) and rebleeding (0.703, 95% CI 0.639–0.767). For endoscopic therapy, only GBS reached nominal statistical significance, but discrimination was limited (AUC 0.604, 95% CI 0.551–0.656). In this selected cohort of endoscopy-confirmed ED patients with UGIB, score performance was outcome-specific rather than uniform across all endpoints. AIMS65 showed the highest mortality discrimination, whereas GBS showed the highest discrimination for transfusion and rebleeding. Endoscopic therapy prediction remained weak. These scores should be interpreted as structured adjuncts to clinical and endoscopic assessment, not as standalone decision rules.