Pathways to poor anticoagulation control

Summary: Background: While a considerable amount is known about which patient-level factors predict poor anticoagulation control with warfarin, measured by percent time in therapeutic range (TTR), less is known about predictors of time above or below target. Objective: To identify predictors of different patterns of international normalized ratio (INR) values that account for poor control, including 'erratic' patterns, where more time is spent both above and below INR target, and unidirectional patterns, where time out of range is predominantly in one direction (low or high). Methods: We studied 103 897 patients receiving warfarin with a target INR of 2-3 from 100 Veterans Health Administration sites between October 2006 and September 2008. Our outcomes were percent time above and below the target range. Predictors included patients' demographics, comorbidities, and other clinical data. Results: Predictors of erratic patterns included alcohol abuse (5.2% more time below and 3.7% more time above, P < 0.001 for all results), taking > 16 medications (4.6% more time below and 1.8% more time above compared to taking seven or fewer medications), and four or more hospitalizations during the study (6.6% more time below and 2% more time above compared to no hospitalization). In contrast, predictors like cancer, non-alcohol drug abuse, dementia, and bipolar disorder were associated with more time below the target range (3.4%, 5.2%, 2.6%, and 3.2%, respectively) and less (or similar) time above range. Conclusion: Different patient-level factors predicted unidirectional below-target and 'erratic' patterns of INR control. Distinct interventions are necessary to address these two separate pathways to poor anticoagulation.