Early indicators of prognosis in upper gastrointestinal hemorrhage.

Abstract

Endoscopy allows accurate risk stratification of patients presenting with gastrointestinal bleeding; frequently, however, it is not immediately available. Initial management and triage of patients thus depends on nonendoscopic information. We sought to risk stratify patients with upper gastrointestinal bleeding using variables available on initial presentation (ie., before endoscopy). A retrospective observational study was performed using data from 335 admissions with an initial diagnosis of upper gastrointestinal hemorrhage. All patients underwent endoscopy and were evaluated for an adverse outcome during their hospitalization. An adverse outcome was defined as death, the need for any operation, recurrent hematemesis, recurrent melena after initial clearing, or a hematocrit falling despite transfusion. Univariate analysis identified 17 distinct variables associated (p < 0.05) with an adverse outcome. A stepwise logistic regression identified five variables as independent predictors (p < 0.05) of an adverse outcome: an initial hematocrit <30%, initial systolic blood pressure < 100 mm Hg, red blood in the nasogastric lavage, history of cirrhosis or ascites on exam, and a history of vomiting red blood. We derived a decision rule based on patients having 0-5 of these independent predictors. This decision rule allowed identification of a large patient population with a <10% chance of an adverse outcome. Risk stratification is possible from information available at the time of initial presentation. If confirmed in other populations, these predictors can be used to identify patients who require a less intensive level of care.