Background: Randomized control trials and observational studies show high-quality warfarin therapy leads to safe and effective stroke prophylaxis. In usual community practice, patient, physician and health care system factors are barriers to optimal anticoagulation. We examined the predictive relationship between inpatient and outpatient INR values in chronic non-valvular atrial fibrillation (AF) patients hospitalized for ischemic stroke (S), bleed (B) and control events (C) in usual community practice.
Methods: This nested case-control analysis identified AF patients hospitalized for S, B and C using medical and pharmacy claims spanning 4.5 years ('98-'03) and validating diagnosis with chart abstraction. AF was defined as 2 medical claims for AF >or= 42 days apart with a related prescription claim for warfarin. INRs from both outpatient and inpatient settings were used to yield a continuous history of coagulation status. Time-in-therapeutic-range (TTR) was calculated by Rosendaal's linear interpolation method. Correlation of inpatient and prognostic utility of last outpatient INRs was tested with S or B hospitalizations using univariate and multivariate logistic regression.
Results: Overall, 614 hospitalizations (means: age 73.9, CHADS(2) = 3.24; 52% male) included S (n = 98), B (n = 101) and C (n = 415) events. Average TTR was 28.6% (49.4% at INR <2.0, 21.9% at INR >3.0). First INR on admission (INR <2.0 or >3.0) was associated with S and B hospitalizations (OR-adjusted [95%CI], 1.68 [1.04-2.73] and 1.72 [1.02-2.90]), respectively. Last outpatient INR <2.0 was not associated with S (OR-adjusted [95%CI], 1.12 [0.77-1.81]), and INR >3.0 was not associated with B (OR-adjusted [95%CI], 1.25 [0.67-2.32]). Last outpatient INR measurement occurred at 28, 22 and 24 days (median; S, B & C, respectively) before hospitalization.
Conclusion: Patients were observed within therapeutic range less than 30% of their time on warfarin. While inpatient INRs were clearly associated with both ischemic stroke and bleed events, last outpatient INR before event was not predictive.