Predictive validity of the Glasgow Blatchford Bleeding Score in an unselected emergency department population in continental Europe

Background and aim Glasgow Blatchford Bleeding Score stratifies patients presenting with acute upper gastrointestinal haemorrhage at the emergency department according to the likelihood of the need for treatment. The objective of this study was to validate the Glasgow Blatchford Bleeding Score for use in an emergency department in the Netherlands. Furthermore, we assessed its clinical usefulness for safe discharge of low-risk acute upper gastrointestinal haemorrhage patients and compared its test validity to that of other scoring systems. Methods This multicentre historic cohort study was conducted in two hospitals in the Netherlands. All 478 patients presenting with a suspicion of acute upper gastrointestinal haemorrhage at our emergency departments during a 1-year period were included. For each patient we calculated Glasgow Blatchford Bleeding Score and other commonly used scores. Test validity was assessed using the receiver operated characteristics curve analysis; calibration plots were used to assess the probability of the need for treatment of different levels of the scores. Results Glasgow Blatchford Bleeding Score had a good discriminative ability in predicting the need for treatment, receiver operated characteristics curve analysis showed an area under the curve of 0.879. Counting a score of 2 or less as low risk (negative), 104 patients (21.7%) were classified as low-risk, with a negative predictive value of 98.1%. These results were superior to those of the other scoring systems. Conclusion Patients presenting at an emergency department in continental Europe with acute upper gastrointestinal haemorrhage and a Glasgow Blatchford Bleeding Score of 2 or less can be safely discharged. The Glasgow Blatchford Bleeding Score performed better than the other commonly used scoring systems. Eur J Gastroenterol Hepatol 24: 382-387 (C) 2012 Wolters Kluwer Health vertical bar Lippincott Williams & Wilkins.