Background & aims: Risk scoring systems are used increasingly to assess patients with upper gastrointestinal hemorrhage (UGIH). There have been comparative studies to identify the best system, but most have been retrospective and included small sample sizes, few patients with severe bleeding and with low mortality. We aimed to identify the optimal scoring system.
Methods: We performed a prospective study to compare the accuracy of the Glasgow Blatchford score (GBS), an age-extended GBS (EGBS), the Rockall score, the Baylor bleeding score, and the Cedars-Sinai Medical Center predictive index in predicting patients' (1) need for hospital-based intervention or 30-day mortality, (2) suitability for early discharge, (3) likelihood of rebleeding, and (4) mortality. We analyzed the area under receiver operating characteristic (AUROC) curve, sensitivity, specificity, and positive and negative predictive values for each system. The study included 831 consecutive patients admitted with UGIH during a 2-year period.
Results: The GBS and EGBS better predicted patients' need for hospital-based intervention or 30-day mortality than the other systems (AUROC, 0.93; P < .001) and were also better in identifying low-risk patients (sensitivity values, 0.27-0.38; specificity values, 0.099-1). The EGBS identified a significantly higher proportion of low-risk patients than the GBS (P = .006). None of the systems accurately predicted which patients would have rebleeding or patients' 30-day mortality, on the basis of low AUROC and specificity values.
Conclusions: The GBS accurately identifies patients with UGIH most likely to need hospital-based intervention and also those best suited for outpatient care. The EGBS seems promising but must be validated externally. No scoring system seems to accurately predict patients' 30-day mortality or rebleeding. ClinicalTrials.gov number, NCT01589250.
Copyright © 2012 AGA Institute. Published by Elsevier Inc. All rights reserved.