Improving quality measurement for anticoagulation: adding international normalized ratio variability to percent time in therapeutic range.

BACKGROUND Among patients receiving warfarin, percent time in therapeutic range (TTR) and international normalized ratio (INR) variability predict adverse events individually. Here, we examined what is added to the prediction of adverse events by using both measures together. METHODS AND RESULTS We included 40 404 patients anticoagulated for atrial fibrillation, aged 65+, within the Veterans Health Administration. TTR and log-transformed INR variability were calculated for each patient. Our study outcomes were ischemic stroke and major bleeding, defined using International Classification of Diseases-9 codes. We estimated the hazard ratios (HRs) for the study outcomes using 3 nested Cox regression models, including (1) TTR or log INR variability separately; (2) TTR and log INR variability together; and (3) both predictors together plus an interaction term. We divided TTR into 3 categories (high, >70%; moderate, 50% to 70%; low, <50%) and log INR variability into 2 categories (stable and unstable). The reference groups high TTR and stable anticoagulation each denote good control. Higher log INR variability (ie, unstable control) predicted ischemic stroke (HR=1.45, P<0.001) and major bleeding (HR=1.57, P<0.001) independently, regardless of TTR levels. Our model with interaction terms showed that High log INR variability predicted a significantly higher risk for ischemic stroke and major bleeding compared with low log INR variability, at moderate TTR levels (HR= 1.27 and HR=1.29, respectively) and at high TTR levels (HR=1.55 and HR=1.56, respectively), but not at low TTR levels. CONCLUSIONS Unstable anticoagulation predicts warfarin adverse effects independent of TTR. Moreover, knowledge about anticoagulation stability further stratifies the risk for adverse events at given levels of TTR.