In the course of completing our study, "Stroke Prevention in Atrial Fibrillation: Impact of Mental Illness" (IIR 04-248), we learned that half of Veterans receiving warfarin in VA actually received INR testing in Medicare (MC) during Fiscal Year 2004 (FY04), suggesting that some anticoagulation care was occurring outside VA and that Veterans were engaging in Dual Use (DU) of VA and MC anticoagulation services. Moreover, patients with INR testing in VA+MC had lower unadjusted frequencies of anticoagulation process measures, suggesting that dual care was not supplementary but may be causing harm. Failure to coordinate VA AF patients' care could increase risk for strokes, bleeds and death. Our study investigates whether patients treated with warfarin who receive care in both VA and MC have less frequent anticoagulation services and poorer outcomes.
We propose the following objectives in studying Veterans with AF receiving warfarin in VA: 1.To characterize patterns of Dual Use (VA+MC); 2.To determine the effect of DU on processes of anticoagulation care; 3.To determine the effect of DU on outcomes of AF care (stroke, hemorrhage, death) and to determine moderation by comorbidities such as heart failure.
To achieve our objects of accessing quality of care for AF among Veterans who use both VA and Medicare, we merged the VA administrative data and Medicare claims data based on SSN, gender, and date of birth. Our cohort comprised of patients diagnosed with AF and treated with warfarin for AF management. Therefore, our definition of dual system use measured the necessary component of care of warfarin management, which was INR laboratory testing.
Since our last update from last year, we refined and validated our definition of dual system use. Our initial approach was to measure dual use as the fraction of MC INR test claims to the proportion of patient's total INR claims in VA and Medicare. This approach was proven to be difficult since the VA anticoagulation guideline (VHA Directive 2010-020) recommends that "INR be checked at no more than 4-week intervals" and a maximum of 6-week interval. Given that INR measured in every 4 to 6-week intervals, we would have underestimated the INR counts for patients who had INR 1 month before or after the 90-day dual use exposure period. Therefore, we decided to use days of INR monitoring for dual use measurement. Our updated definition of dual use was the proportion of INR monitoring eligible days that were covered by INR monitored days in Medicare. Because measurement of dual use showed a non-normal distribution, we defined dual use into 5 categories (0%, >0-25%, >25-50%, >50-75%, >75-100%), where 0% did not rely on Medicare (VA-only users) and 100% relied mostly Medicare for INR monitoring.
During the past year, we have established baseline characteristics and explored the distribution of the dual use (0%, >0-25%, > 25-50%, >50-75%, >75->99%) by the VA Medical Centers, VA's community-based outpatient clinics, driving distance to the VA facility from patient's residential zip code, VA priority status, and rurality of patient's residence. We used descriptive and multivariate analysis to assess the degree of dual use and its effect on health outcomes.
Because our inclusion criteria are based on warfarin and INR measurement and our baseline characteristics include cardiovascular medications, we must use pharmacy and laboratory data. Although we have these data in the VA, we do not have Medicare pharmacy (Part D) data. This Part D data recently became available at VIReC in September 2013, and we are in the process of obtaining the dataset.
We found that the level of dual use was substantially differed by the location of VA facilities, patient's residence, travel time to VA facilities or clinics, and VA priority status. The percent of co-payment increased with the increase reliance on Medicare. In addition, patients who lived in rural areas were more likely to seek Medicare for anticoagulation treatment. We also found that patients who relied more on Medicare for warfarin management had worse health profile at baseline. These characteristics will be used to adjust for confounders in our next phase of finding association between dual use and outcomes.
In observational studies, confounding can be controlled by matching the patients in different treatment groups by likelihood to receive treatment. In our case, we estimated the patients' likelihood to be dual users using propensity score matching. We identified 0% group as non-dual users and >0% as dual users. Using nearest neighbor matching (matching on closest propensity scores), we identified patients in both groups who have equal chance of being dual users.
Using only the propensity matched cohort, we found that statistical significance of multivariate regression improved with each addition of covariates for adjustment. This indicates that our estimate for outcome effect (stroke or death) improved with propensity matching.
Among 103,000 Medicare-eligible patients receiving warfarin from the VA for AF management between 2004-2008 (mean age 77.2 +/- 6.1, 1.6% female), 40% used both VA and Medicare for INR testing. Compared with VA-only users, dual users had significantly lower INR monitoring intensity and TTR (P<0.0001) and had higher unadjusted one-year incidence of stroke, mortality, and hemorrhage. After adjusting for patient and facility characteristics, dual use was associated with increased risk of incident intracranial hemorrhage. These findings were consistent in a propensity-matched cohort and results indicate that poor care coordination between health care systems could reduce warfarin benefits.
This study is well aligned with the objectives of VA's Stroke QUERI and CHF QUERI and will make important contributions to VA's stroke prevention quality improvement efforts. Identification of mediators of quality of AF care, such as Dual Use, will serve as a foundation to guide emerging systems improvement policies in VHA regarding selection of stroke prevention therapies in Veterans, particularly with the emergence and adoption of new safe and efficacious alternatives to warfarin for atrial fibrillation. Specifically, two new oral anticoagulants, dabigatran and rivaroxaban, have been approved by VA PBM and are on national formulary.
The impact of dual use on quality of care varied by the degree of reliance on VA and Medicare. While equally dual use and predominant Medicare use for warfarin management reduced the quality of care for warfarin management, predominant VA use had no significant difference on the quality of care compared to those who used VA only. Although possessing additional insurance coverage by Medicare can reduce challenges in geographical access to specialty care at the VA, we found evidence that dual system use can reduce the quality of warfarin management among AF patients. Disjointed care could compromise safety and effectiveness of anticoagulation management. Future research and policy should focus on ways to reduce fragmented care while ensuring patients' continued access to needed care.
- Turakhia MP, Santangeli P, Winkelmayer WC, Xu X, Ullal AJ, Than CT, Schmitt S, Holmes TH, Frayne SM, Phibbs CS, Yang F, Hoang DD, Ho PM, Heidenreich PA. Increased mortality associated with digoxin in contemporary patients with atrial fibrillation: findings from the TREAT-AF study. Journal of the American College of Cardiology. 2014 Aug 19; 64(7):660-8.
- Turakhia MP, Hoang DD, Xu X, Frayne S, Schmitt S, Yang F, Phibbs CS, Than CT, Wang PJ, Heidenreich PA. Differences and trends in stroke prevention anticoagulation in primary care vs cardiology specialty management of new atrial fibrillation: The Retrospective Evaluation and Assessment of Therapies in AF (TREAT-AF) study. American heart journal. 2013 Jan 1; 165(1):93-101.e1.
- Turakhia M, Than CT, Xu X, Schmitt SK, Yang F, Frayne SM, Phibbs CS, Holmes TH, Heidenreich PA. Impact of Baseline Stroke Risk and Bleeding Risk on INR Control among Patients with AF on Warfarin: Data from the TREAT-AF Study. Presented at: American Heart Association Annual Scientific Sessions; 2012 Nov 5; Los Angeles, CA.