Clinical Performance of Prediction Rules and Nasogastric Lavage for the Evaluation of Upper Gastrointestinal Bleeding: A Retrospective Observational Study.

Introduction. The majority of patients with acute upper gastrointestinal bleeding (UGIB) are admitted for urgent endoscopy as it can be difficult to determine who can be safely managed as an outpatient. Our objective was to compare four clinical prediction scoring systems: Glasgow Blatchford Score (GBS) and Clinical Rockall, Adamopoulos, and Tammaro scores in a sample of patients presenting to the emergency department of a large US academic center. Methods. We performed a retrospective cohort study of patients during 2008-2010. Our outcome was significant UGIB defined as high-risk stigmata on endoscopy, or receipt of blood transfusion or surgery, or death. Results. A total of 393 patients met inclusion criteria. The GBS was the most sensitive for detecting significant UGIB at 98.30% and had the highest negative predictive value (90.00%). Adding nasogastric lavage data to the GBS increased the sensitivity to 99.57%. Conclusions. Of all four scoring systems compared, the GBS demonstrated the highest sensitivity and negative predictive value for identifying a patient with a significant UGIB. Therefore, patients with a 0 score can be safely managed as an outpatient. Our results also suggest that performing a nasogastric lavage adds little to the diagnosis UGIB.

Duke Scholars

Author Shein-Chung Chow Biostatistics & Bioinformatics, Division of Biostatistics