Introduction Acute upper gastrointestinal bleeding (UGIB) is a common medical emergency with significant morbidity and mortality. It is essential for clinicians to effectively triage patients into risk groups allowing appropriate allocation of resources. Several scoring systems have been devised following admission with suspected UGIB however the most useful score is debated. The most useful score should rapidly stratify risk and not require endoscopy. The aim of this study was to investigate the performance of 3 pre-endoscopic predictive scoring tools in a single, large cohort of patients admitted with suspected UGIB.

Method The Aberdeen bleeding unit records demographics, presenting symptoms, co-morbidities, and outcomes on all admissions. Patients were included if they presented with features suggestive of UGIB and all variables required to calculate the 3 scores (admission-Rockall, Glasgow-Blatchford Score (GBS) and pre-endoscopic Baylor score). Receiver operating characteristic (ROC) curves were drawn for each score and the area under the curve (AUROC) calculated for 3 outcomes; requirement for blood transfusion, rebleeding and 30 day mortality. Sensitivity and negative predictive value (NPV) analysis was performed to assess very low scores (Rockall <1, GBS <2, Baylor <6) for potential early discharge.

Results 6,788 patients (M:F=1.3:1, median age 65) were included in the study. For requirement for blood transfusion, AUROC was significantly higher for GBS (0.92, 95% CI 0.91–0.93, p < 0.001) compared to Rockall (0.68, 95% CI 0.67–0.70) and Baylor scores (0.65, 95% CI 0.64–0.67) respectively. For rebleeding, AUROC was significantly higher for GBS (0.76, 95% CI 0.75–0.78, p < 0.001) compared to Rockall (0.66, 95% CI 0.64–0.68) and Baylor scores (0.61, 95% CI 0.59–0.63). For 30 day mortality, GBS and Rockall were not significantly different (0.75, 95% CI 0.74–0.77 and 0.76, 95% CI 0.74–0.78 respectively) but were both significantly higher than the Baylor score (0.71, 95% CI 0.69–0.73). Table 1shows the sensitivity and NPV analysis.

Conclusion GBS was superior to both Rockall and Baylor score for predicting transfusion and rebleeding. GBS and Rockall scores were superior to the Baylor score in predicting mortality. Using GBS <2 less than 1 in 300 patients would have an adverse outcome.

Disclosure of interest None Declared.