Validation of a decision rule to predict patients at low risk of variceal upper gastrointestinal hemorrhage.

Abstract

Determining the likelihood of a variceal versus nonvariceal source of upper gastrointestinal bleeding (UGIB) guides ED therapy, but can be difficult to determine on clinical grounds. A simple decision rule, using only platelet and international normalized ratio (INR) values, was previously derived in a single center and had high sensitivity (97%). We sought to validate this decision rule using multi-center data. We performed this decision rule validation using data collected from 21 Canadian hospitals, comprising 2020 patients. The parent study enrolled patients aged ≥18years at participating hospitals with nonvariceal or variceal UGIB from January 2004 through May 2005. To validate the existing decision rule, we computed the test characteristics of the rule on this study population. The existing decision rule, in order to predict patients at low risk for variceal hemorrhage, is designed to be highly sensitive for variceal UGIB. In the previously derived rule, patients are not low risk if either is present: INR ≥1.3 or platelet count ≤200×109/L. We additionally added a third common-sense criterion to the decision rule in a separate analysis: whether the patient has previously had variceal hemorrhage. 2001 patients were eligible for analysis, including 214 (10.7%) with a variceal source of gastrointestinal hemorrhage. Median age was 69 (IQR 55-79), and 764 (38%) were women. The two-step rule correctly identified 204 of the 214 (95.3%) patients with variceal hemorrhage; adding prior variceal hemorrhage as a variable identified 5 more patients (209/214 [97.7%]). Of the 2001 patients, 953 (47%) would have been classified as low risk for variceal hemorrhage; of these patients, 5 (0.5%) experienced variceal hemorrhage. The sensitivity of the rule in this validation cohort was 95.3% (95% CI 91.6-97.7%), with a negative likelihood ratio of 0.09 (95% CI 0.05-0.16). Adding prior variceal hemorrhage increased sensitivity to 97.7% (95% CI 94.6-99.2%), with a negative likelihood ratio of 0.04 (95% CI 0.02-0.11). We have validated a simple decision rule to identify patients at low risk for variceal UGIB. This two-step (three-step if prior history of variceal hemorrhage is known) rule is simple to use, and may enable safe deferment of unnecessary or harmful therapies.