Outcome-Specific Performance of HALP, AIMS65, Glasgow–Blatchford, and Rockall Scores in Emergency Department Patients with Acute Upper Gastrointestinal Bleeding: A Retrospective Cohort Study

Background/Objectives: The Hemoglobin–Albumin–Lymphocyte–Platelet (HALP) score integrates hematologic, nutritional, immune, and platelet parameters, but its prognostic value in acute upper gastrointestinal bleeding (UGIB) remains uncertain. We compared the performance of HALP with AIMS65, Glasgow–Blatchford Score (GBS), and Rockall scores. Methods: We retrospectively studied consecutive adults with endoscopy-confirmed UGIB presenting to the emergency department from January 2021 through December 2024. Outcomes were in-hospital mortality, packed red blood cell (PRBC) transfusion within 24 h, and composite severe UGIB, defined as ≥2 PRBC units transfused within 24 h or UGIB-related intensive care unit admission. Discrimination was assessed using receiver operating characteristic curves and areas under the curve (AUCs); pairwise comparisons used the correlated-sample DeLong test. Firth-penalized logistic regression assessed the adjusted association between HALP and in-hospital mortality. Results: Of 209 patients, 24 (11.5%) died and 36 (17.2%) had variceal bleeding. For mortality, AUCs were 0.580 for inverse HALP, 0.802 for AIMS65, 0.733 for GBS, and 0.731 for Rockall. For 24-h PRBC transfusion, AUCs were 0.629, 0.642, 0.814, and 0.511, respectively; for severe UGIB, they were 0.622, 0.594, 0.781, and 0.501. No statistically significant adjusted association between HALP and mortality was observed (aOR, 0.91 per 10-point increase; p = 0.571). Conclusions: HALP showed limited discrimination across the evaluated UGIB outcomes, and no statistically significant adjusted association with mortality was observed. AIMS65 had the highest observed AUC point estimate for mortality, although its AUC did not differ significantly from those of GBS or Rockall. GBS had the highest AUCs for PRBC transfusion and severe UGIB. These findings support outcome-specific selection of UGIB risk scores and do not support the use of HALP as a stand-alone risk-stratification tool without external validation.