OBJECTIVES To assess the clinical impact of disagreement between two methods of measuring INR in our anticoagulation clinics. METHODS We assessed paired INR samples collected from the same patient at the same visit from 4 anticoagulation clinics associated with the Johns Hopkins Medical Institutions. One venous sample was analyzed in the Johns Hopkins Coagulation Laboratory and one fingerstick sample was analyzed by the Hemochron® Signature Plus POC device (International Technidyne Corporation, Edison, NJ, USA). A newly-developed analytic technique was applied that reports results in explicitly clinical terms. RESULTS Samples were collected from 1666 patients from January, 2006 through June, 2008. The lab was nearly twice as likely to report a value below the target INR range of 2-3 compared with the POC (36% vs. 19%). Furthermore, the lab was almost four times more likely to report that the INR was 1.0-1.49 – a range associated with a 13.5-fold increased risk of a thromboembolic event (18% vs. 5%). Overall, 29% of clinical decisions were projected to differ, depending on which measure was used. A majority (53%) of the predicted clinical differences occurred when the POC device reported a value within the target INR range. Based on epidemiologic evidence of risk as a function of the INR, we estimate that 67 thromboembolic events occurred over the 2.5 year period because of the discrepancy between the measures. CONCLUSIONS There are clinically relevant differences between INR values reported by the Hemochron POC device and our clinical lab that were not detected by conventional methods. The Hemochron device biases INR measures towrd the target range throughout the INR scale. This bias prevents many necessary warfarin dosing changes because the clinician does not understand the patient's true risk level. Therefore, patients spend more time at INR levels strongly associated with an increased risk of a negative outcome.