Prediction Model for Significant Bleeding in Patients with Supratherapeutic International Normalized Ratio After Oral Administration of Warfarin

Background and Objective The use and range of indications for anticoagulation therapy are steadily growing. The objective of this study was to develop a scoring model to predict the occurrence of significant bleeding in patients taking warfarin with a supra-therapeutic international normalized ratio. Methods Data were collected from the medical records of patients taking warfarin with an international normalized ratio > 3.5. The characteristics of bleeding episodes and the need for transfusion of blood products were recorded. Regression models were constructed to predict the occurrence of significant bleeding (requiring a transfusion of more than 2 units of packed red blood cells, intrapericardial or intracranial hemorrhage). The predictive values of previously published scores (ATRIA: anemia, hypertension, severe renal disease, age ≥ 75 years, or prior bleeding history; and ORBIT: old, reduced hemoglobin, bleeding history, kidney insufficiency or antiplatelet treatment) were compared with our New Bleeding Score (NBLDSCOR); the areas under the curve for the receiver-operating characteristic plots were compared using a non-parametric DeLong test. Results Significant bleeding was reported in 87 out of 389 admitted patients. With an area under the curve of 0.736 ± 0.032, NBLDSCOR was the best predictor of significant bleeding in this population. Neither ATRIA nor ORBIT was a good predictor of significant bleeding, where the area under the curve for the receiver-operating characteristic plot for ATRIA was 0.654 ± 0.034 and for ORBIT was 0.604 ± 0.033. The predictive power of NBLDSCOR was superior to ATRIA and ORBIT ( p