Discrepancies in Identification of Major Bleeding Events in Patients Taking Warfarin


University of Washington Medical Center, Department of Pharmacy, Box 356015, 1959 Northeast Pacific Street, Seattle WA 98195; e-mail: akwitt@u.washington.edu.


Study Objective. To assess the level of agreement between one subjective definition of major bleeding (criteria A) and two objective definitions: the International Society on Thrombosis and Haemostasis ([ISTH] criteria B) and the Italian Study of Complications of Anticoagulant Therapy ([ISCOAT] criteria C).

Design. Retrospective cohort study.

Setting. Four anticoagulation clinics at a university-affiliated medical center.

Patients. One hundred twenty patients aged 21–96 years who were taking long-term warfarin therapy and experienced a nonminor bleeding event between July 1, 2001, and June 30, 2006.

Measurements and Main Results. Bleeding events were evaluated using three definitions of major bleeding: criteria A—event was fatal or required hospitalization; criteria B—event was fatal, was symptomatic in a critical area or organ, caused a decrease in hemoglobin level of 2 g/dl or more, and/or led to transfusion of 2 or more units of whole blood or red blood cells; and criteria C—event met criteria B definition (but less inclusively identifies critical areas or organs) and/or necessitated surgical or angiographic intervention. One hundred thirty-eight bleeding events in the 120 patients met at least one of the definitions of major bleeding and thus were included in the analysis. Level of agreement among the definitions was determined by the κ statistic. The level of agreement between criteria B and C was excellent; no discrepancies were found; however, the level of agreement between criteria A and criteria B or C was poor, with a κ statistic value of −0.134 (95% confidence interval −0.196 to −0.071). Use of criteria A resulted in underreporting of 31 major bleeding events by excluding events that were managed on an outpatient basis.

Conclusion. The discrepant event rates determined by using three different definitions of major bleeding underscore the need for implementation of a standardized objective definition. Use of a standardized definition in clinical trials to accurately predict and compare patient outcomes and in clinical practice as a marker for patient safety will increase quality of care and decrease total costs of care in patients treated with warfarin. We suggest implementation of the ISTH criteria in clinical practice as a method to standardize the reporting of major bleeding events.