Abstract
Introduction
Bleeding-specific risk scores are widely used in upper gastrointestinal bleeding (UGIB) but were primarily developed in non-critically ill populations. We compared the prognostic performance of the Sequential Organ Failure Assessment (SOFA) score with bleeding-specific scores in patients with UGIB requiring intensive care.Methods
We conducted a retrospective cohort study using MIMIC-III, including adults presenting with UGIB and requiring direct ICU admission. The primary outcome was 30-day mortality. Discrimination of SOFA, Glasgow-Blatchford Score (GBS), AIMS65, pre- and post-endoscopic Rockall scores, serum lactate, and blood urea nitrogen-to-creatinine ratio was assessed using areas under the receiver operating characteristic curves.Results
Among 979 patients, 73 (7.5%) died within 30 days. SOFA showed the highest discrimination (AUC 0.818), followed by lactate (0.774), GBS and AIMS65 (0.767 each), pre-endoscopic Rockall (0.649), post-endoscopic Rockall (0.632), and blood urea nitrogen-to-creatinine ratio (0.604). A SOFA threshold of 5.5 provided 67.1% sensitivity and 84.0% specificity. Higher SOFA was associated with significantly lower 30-day survival (P Conclusion
In critically ill patients with UGIB, SOFA showed better discrimination for 30-day mortality than bleeding-specific scores, supporting the prognostic relevance of early organ dysfunction assessment.