Potentially Inappropriate Prescribing of Direct-acting Oral Anticoagulants in the Veterans Health Administration
This study examines variations in overall use, and potentially inappropriate use, of non-warfarin direct-acting oral anticoagulants across sites within the Veterans Health Administration.
Objectives: Little is known about variation in rate of uptake or potentially inappropriate prescribing of non-warfarin direct-acting oral anticoagulants (DOACs). Our objective was to examine variations in overall use, and potentially inappropriate use, of DOACs across sites within the Veterans Health Administration (VA).
Study Design: Longitudinal database study.
Methods: We examined 32,719 unique patients who received a DOAC from 1 of the 130 VA sites of care between October 1, 2010, and September 30, 2014 (fiscal year [FY] 2011-FY2014). We examined the growth of DOAC use over time, and how it varied by site. We examined the proportion of patients receiving DOACs who were potentially inappropriate recipients, because of the presence of a prosthetic heart valve or insufficient renal function.
Results: Use of DOACs within VA increased, with more initial prescriptions in FY2014 (18,369 prescriptions) than all previous years combined (14,350). This use was uneven by site; the top 10 sites collectively accounted for 26% of DOAC prescriptions in FY2014, while the bottom 10 sites accounted for 0.6%. Overall, 4.58% of DOAC prescriptions were potentially inappropriate: 1.59% because of a prosthetic heart valve, and 3.04% because of insufficient renal function. The rate of inappropriate prescribing varied widely among sites, from a low of 0% (3 sites) to a high of 12.6%.
Conclusions: Uptake of DOACs within VA is accelerating but is uneven across sites. While potentially inappropriate prescribing of DOACs is relatively uncommon, it varies among sites from completely absent to as high as 12.6% of prescriptions. Understanding this variation will have implications for pharmacy budgets and for patient safety.
Am J Pharm Benefits. 2016;8(4):e75-e80
The introduction of non-warfarin direct-acting oral anticoagulants (DOACs) has generated considerable enthusiasm. Clinicians and patients, alike, have been eager to adopt effective long-term oral anticoagulants that do not require the close monitoring or frequent dose titrations that are needed to effectively use warfarin. However, DOACs are not appropriate for all patients and in all situations.
Specifically, their use is proscribed when the patient has insufficient renal function, and their use is not recommended for patients with prosthetic heart valves or who have other, less-common indications for anticoagulation.1 Relatively few studies have been published about how these new agents are being adopted into clinical practice, although one such study suggested that they are frequently used for off-label indications, such as hypertensive heart disease and coronary artery disease.2
In this study, we examined the increasing use of DOACs within the Veterans Health Administration (VA). We also examined the rates of potentially inappropriate use of DOACs within VA, due to inappropriate indications or insufficient renal function. Finally, we examined differences among VA sites, in terms of overall utilizations of these agents, as well as potentially inappropriate utilization.
The increasing uptake of DOACs by health systems, over time, will have potentially important implications for resource utilization, while ensuring that appropriate prescribing of these agents will have important implications for patient safety. The role of the present study is to provide a barometer of how these trends are progressing in the nation’s largest integrated healthcare system, the VA.
We identified all patients who received any of the DOACs from a VA pharmacy from October 1, 2010, through September 30, 2014, which corresponds to VA fiscal years (FYs) 2011-2014. DOACs were first included in the VA formulary during FY2011. During our study period, there were 3 DOACs in use: dabigatran, rivaroxaban, and apixaban. The study was approved by the Institutional Review Board of the Bedford [Massachusetts] VA Medical Center.
Sites of Care
Part of our analysis was to examine variation in rates of use of DOACs, and of potentially inappropriate prescribing, by VA site. We included 130 VA sites of care, each of which includes a hospital, an outpatient care center, and several outlying community-based clinics. If a patient visited more than 1 site (2.8% of patients), we assigned the site at which the medication was initiated.
Inappropriate Use of DOACs: Prosthetic Heart Valves
One major way that a prescription for a DOAC might be inappropriate would be to use it in a patient who has a prosthetic heart valve. For the present analysis, short-term primary prophylaxis of venous thromboembolism (VTE) was not considered; we focused on longer-term indications for anticoagulation therapy. All 3 DOACs eventually received FDA indications for 2 such indications: stroke prevention in non-valvular atrial fibrillation and treatment of VTE (including prevention of recurrent VTE).3-5
In contrast, the use of DOACs for some indications not only lacks an FDA indication, but also is contraindicated or discouraged. A prominent example would be using a DOAC in a patient with a prosthetic heart valve. The RE-ALIGN study found that dabigatran is inferior to warfarin in patients with a mechanical prosthetic heart valve, resulting in, among other problems, higher rates of both bleeding and thromboembolism.6
Other DOACs have not been studied with prosthetic heart valves, but in the absence of direct evidence we should also assume that warfarin would be preferred over them as well. Therefore, the package insert for dabigatran states that its use is contraindicated in patients with mechanical heart valves and is “not recommended” for those with bioprosthetic valves3; the inserts for the other DOACs state that use with any prosthetic heart valve is “not recommended.”4,5 VA guidance documents reiterate these recommendations.1
We used a well-validated list of International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes to identify indications for therapy, then established a hierarchy of indications.7 Prosthetic valve was considered the most important indication when present, followed by VTE (any further treatment was assumed to be for VTE treatment and prevention of recurrence), followed by atrial fibrillation/atrial flutter.
Other, less-common indications for anticoagulation included systemic embolus, mural thrombus, and others. A list of ICD-9-CM codes used to define indications for therapy can be found in the eAppendix (available at www.ajmc.com). Of note, ICD-9-CM codes do not reliably differentiate between a mechanical and a bioprosthetic heart valve, although neither is considered appropriate for DOAC use in VA.1
Inappropriate Use of DOACs: Kidney Function
VA guidelines do not recommend the use of dabigatran or rivaroxaban if the patient has an estimated glomerular filtration rate (eGFR) below 30 mL/min/1.73m2, or of apixaban if the eGFR is below 25 mL/min/1.73m2.1 FDA-mandated package inserts for all 3 DOACs do allow for administration of a reduced dose in severe renal failure, despite an absence of direct evidence for such use.
The package insert for dabigatran allows a reduced dose of 75 mg twice daily for eGFR 15-30 mL/min/1.73m2,3 despite the fact that no clinical trial studied this dose or this level of renal function.8 The package inserts for rivaroxaban and apixaban similarly allow reduced doses in severe renal failure,4,5 despite the absence of direct trial data.9,10 VA guidance documents conservatively recommend that only doses which have directly been studied should be given, and only at levels of renal function which were studied.1
Therefore, within VA, it is recommended that dabigatran and rivaroxaban should not be used below an eGFR of 30 mL/min/1.73m2, or apixaban below an eGFR of 25 mL/min/1.73m2.1 Anytime a patient whose kidney function fell below these cutoffs received a DOAC, we considered it to be potentially inappropriate. If kidney function improved upon re-measurement, receipt of a DOAC was considered appropriate thereafter. Of note, reduced doses of DOACs (such as dabigatran 75 mg) are not approved for use within VA, and are therefore not an option, as stated in VA Criteria for Use guidance.1
We recalculated the eGFR for a patient whenever a new creatinine value was recorded, using the 4-variable version of the Modification of Diet in Renal Disease (MDRD) equation.11 Several methods can be used to calculate kidney function based on creatinine values, including the Chronic Kidney Disease Epidemiology Collaboration equation, the Cockcroft-Gault equation, and MDRD.12
We chose MDRD for this study because it was the most appropriate match for the creatinine assay used by VA during our study period,13 and because it does not require the patient’s weight, which can be missing. In addition, the major difference among the prevalent methods lies in their ability to detect Stage II chronic kidney disease—that is, slight decrements in eGFR which have prognostic significance, but no immediate clinical significance.14 Because we only considered an eGFR below 30 mL/min/1.73m2 (or 25 mL/min/1.73m2) to be potentially problematic in combination with DOAC use, this distinction is not relevant to our study.
We examined the characteristics of patients in our source population as of the date of each patient’s first prescription for a DOAC. We examined the rates of use for each DOAC and for warfarin by year: overall, by indication, and by medical center. We then examined the rates of potentially inappropriate use for all DOACs and for each DOAC individually: due to inappropriate indications, due to insufficient renal function, and overall.
For the cross-section of patients who used DOACs in FY2014, whether as new or continuing users, we examined the variation in potentially inappropriate use by site. We examined the relationship between site volume of DOAC prescriptions and the rate of potentially inappropriate prescribing. Finally, we examined changes in the rate of potentially inappropriate prescribing over time. A P value of <.05 was considered statistically significant. All analyses were conducted using SAS, version 9.4 (SAS Corporation, Cary, NC).
We studied 32,719 unique patients who received DOACs from VA from October 1, 2010, through September 30, 2014. Patient characteristics as of each patient’s first DOAC “fill” from a VA pharmacy are described in Table. Most patients receiving DOACs (75%) were aged 65 years or older. Reflecting the VA population, 97% were male and 77% were white, non-Hispanic.
Dabigatran accounted for the majority of DOAC prescriptions (56%), with rivaroxaban accounting for much of the remainder (38%) and relatively few patients receiving apixaban (5%). A considerable proportion of DOAC recipients (14.8% in total) either had an unusual indication (5.1%) for receiving a DOAC (see eAppendix for a list of such indications) or no recorded indication at all (9.7%). Atrial fibrillation was by far the most common reason to receive DOAC therapy (66% of all prescriptions).
Patients receiving DOAC therapy for atrial fibrillation were somewhat older than other patients, with 82% of them being aged 65 years or older. In keeping with the timing of FDA indications present during most of the study period, dabigatran was the predominant DOAC for atrial fibrillation (74%), while rivaroxaban was the predominant DOAC for venous thromboembolism (75%).
Increased Use of DOACs Over Time
As would be expected, the use of DOACs increased during the study period, as clinicians gained familiarity with these agents and as new indications received FDA approval (Figure 1). There was evidence of acceleration in this trend, with more new starts for DOACs in FY2014 (18,369) than all previous years combined (14,350).
Use varied widely by site. For example, during FY2014, out of 130 VA sites of care, the 10 highest-utilizing sites collectively accounted for 26% of DOAC prescriptions, while the 10 lowest-utilizing sites collectively accounted for 0.6%. At 5 of the 10 highest-utilizing sites, there were more new starts for a DOAC in FY2014 than for warfarin (data not shown). For example, at the highest-utilizing site, there were 709 DOAC starts in FY2014 and 420 warfarin starts. We also characterized the duration for all 225,238 DOAC prescriptions written during the study period. Most of them were for 30 days’ worth of medication (84%), followed by 90 days (10%), 60 days (2%), and other durations (4%).
Potentially Inappropriate Use of DOACs
A small proportion of DOAC prescribing over the entire study period was for patients who had prosthetic heart valves (519 patients; 1.59% of all DOAC patients). Of those patients, 470 (1.44%) were inappropriate at the start of therapy, and 49 (0.15%) later had such a code, yet continued DOAC therapy thereafter. Similarly, a small number of patients received DOACs despite having eGFR below 30 mL/min/1.73m2 for dabigatran or rivaroxaban, or 25 mL/min/1.73m2 for apixaban (995 patients; 3.04% of all DOAC patients).
Of those patients, 371 (1.13%) had an inappropriately low eGFR at the time of their first prescription, and 624 (1.91%) developed inappropriately low eGFR at a later date, but received at least 1 additional fill for a DOAC after that date. Of the patients who developed inappropriately low eGFR after their first fill and remained on DOAC, 269 (43%) received only 1 additional fill, while 355 (57%) received at least 2.
Overall, 4.58% of patients received a potentially inappropriate DOAC fill; 2.54% were not appropriate candidates at the time of inception, while 2.04% initially appropriate candidates became inappropriate but continued to receive DOAC therapy.
The phenomenon of potentially inappropriate prescribing was also highly variable among sites (Figure 2). Of 130 VA sites, we focused on 115 sites with at least 50 patients with DOAC fills, to avoid problems with small cell sizes and unstable estimates. At 3 of these 115 sites, none of the DOAC fills was potentially inappropriate. However, 7 sites had 9% or more of patients with a DOAC fill flagged as potentially inappropriate. At 1 site, 12.5% of 104 DOAC patients had a potentially inappropriate fill.
We examined whether sites with a higher overall volume of DOAC prescriptions would have a higher (or lower) rate of potentially inappropriate prescribing, again limiting our analysis to sites with at least 50 total patients receiving a DOAC. There was no correlation between volume of prescriptions and the rate of potentially inappropriate prescriptions (r = 0.06, P = .56). We did find that the rate of potentially inappropriate prescribing of DOACs decreased over time across VA: 7.9% in FY2011, 6.6% in FY2012, 5.2% in FY2013, and 3.5% in FY2014.
In this study, we examined the growing use of DOACs over time in VA, the largest integrated healthcare system in the United States. We found that DOAC use is growing at an increasing rate, with more new starts in FY2014 than all previous years combined. We also found that the expanding use of DOACs is highly variable across sites, with some sites continuing to use almost no DOACs, while other sites adopting them at very high rates.
At a few sites, more patients began DOACs in FY2014 than warfarin. We also found that a small but appreciable proportion of DOAC use in VA appeared to be potentially inappropriate, either because of insufficient renal function or the presence of a prosthetic heart valve. Our best estimate of the overall rate of potentially inappropriate prescribing was 4.58%; however, this rate was also highly variable by site.
Of sites with at least 50 new starts for DOACs in FY2014, 3 sites had no potentially inappropriate prescriptions, while 5 sites had a rate of potentially inappropriate prescribing of more than 9%. It is encouraging that we found that the rate of potentially inappropriate prescribing has been decreasing over time in VA.
Understanding patterns of prescribing of DOACs has implications in terms of cost and patient safety. DOACs have important implications for pharmacy budgets, for 4 main reasons: (a) their use is growing, as we have shown; (b) their cost is much greater than that of warfarin, even considering the labor of managing warfarin15; (c) anticoagulants are indicated for a large and ever-growing number of patients16; and (d) patients often continue to use anticoagulants for many years. In a forthcoming study, we will examine the reasons why some VA sites have embraced DOACs much more avidly than others, using qualitative methods to examine key informant interviews.
We do know that VA guidance changed early in FY2014 from an implicit “fail first” policy,17 where patients were expected to try warfarin first, to a policy that explicitly allowed using DOACs without a trial of warfarin first.1 We will be asking about how that change in policy impacted local pharmacy policy. We also know that while VA Pharmacy Benefits Management provides uniform guidance in the form of Criteria for Use documents, each site still may differ somewhat in how they choose to implement those recommendations.
Qualitative methods will be ideal to examine these site-level differences in the implementation of nominally uniform guidelines. Our findings will have implications not only for DOACs, but for guiding a value-based approach to evaluating many new agents that are developed, almost all of which are much more expensive than previous medications.
In addition, our findings also have implications for patient safety. Overall, we found a relatively low rate of potentially inappropriate prescriptions for DOACs across VA, which is, in part, a testament to vigorous oversight by VA pharmacists, who are in turn supported by resources from the national VA pharmacy service, including data dashboards, medication use evaluations, and internally produced criteria-for-use documents.1
Together with other strengths of the VA system (highly functional electronic medical record [EMR] system, well-educated clinicians, integrated healthcare system, etc), this may explain why we saw a relatively low overall rate of potentially inappropriate prescribing of DOACs. However, we did find opportunities for improvement, including an appreciable number of patients who received DOACs despite the presence of a prosthetic heart valve or insufficient renal function.
In addition, we saw that the rate of potentially inappropriate prescribing at some sites was twice or even 3 times the VA average. Over the coming months and years, VA should work to ensure that potentially inappropriate prescribing of DOACs, which runs contrary to official VA guidelines1 should occur at a rate approximating zero, as is already the case at a few VA sites.
There have been few previous studies about potentially inappropriate prescribing of DOACs, possibly because few databases contain detailed enough information to explore the issue. One early report examined the uptake of dabigatran in 2011, when other DOACs were not yet in use.2 Like our study, that study suggested that uptake of dabigatran was increasing at an accelerating rate.
The previous study also found that off-label use of dabigatran was considerable, although much of the off-label use in that study was for uncommon indications, and relatively little was for patients with prosthetic heart valves.2 In addition, that study did not examine renal function as a contraindication to DOAC use.
Another recent study examined potentially inappropriate prescribing practices with dabigatran, but that paper focused on co-administration with interacting medications (such as nonsteroidal anti-inflammatory medications).18 Thus, to our knowledge, our report is largely new information, and points to an important quality issue that needs to be addressed.
Our study has several strengths. We used a large and powerful database with many observations, many sites of care, and considerable clinical detail. However, we also acknowledge certain limitations. Our method for identifying prosthetic heart valves relied largely on a “V” code (“V43.3”); this code accounted for 98% of the patients whom we identified as having a prosthetic heart valve.
V codes have often been questioned as a source of data for being under-coded, in part because they are not billable. In the VA context, no code is billable, and this V code is commonly used to record the indication for anticoagulation, although any code can be subject to variations in coding practices.
We have had success using this code in the past, and have generally identified patients with prosthetic valves at a rate which concords well with other studies.7 However, it is possible that our algorithm did not identify 100% of the patients who had a prosthetic heart valve. In any event, we are confident that the use of DOACs for prosthetic heart valves in VA is at least as prevalent as our study found.
A second limitation is that VA has a highly functional EMR, centralized formulary management, and a proactive pharmacy service, all of which would tend to reduce the amount of inappropriate prescribing. Therefore, the issues that we observed in this study are unlikely to be unique to VA; indeed, we suspect it is likely that the rate of inappropriate prescribing that we described here would be considerably higher under conditions that prevail outside VA. Third, VA patients are mostly male and have a high burden of comorbidity. However, it is unclear how this fact would have altered the present study.
Fourth, we did not use formal methods to explicitly characterize the uptake of DOACs over time, such as a spatiotemporal model. Such methods could refine the assertion that adoption of DOACs accelerated during the study period. However, we did tabulate the number of DOAC prescriptions by site, and overall, year by year. This is a somewhat less sophisticated, but nevertheless serviceable, way to at least qualitatively assess the rate of spread of a new medication.
Finally, this study relied upon automated data, such as ICD-9-CM codes, as it would have been impractical to access and review this many charts. Automated data do have well-recognized limitations. For this reason, we used a definition of potentially inappropriate prescribing, which could be determined with confidence from automated data, such as the presence of a prosthetic heart valve or insufficient kidney function.
However, we generally assumed that other issues, such as the absence of a creatinine test or the absence of a recognized indication for DOAC therapy, were merely indicative of incomplete documentation, and did not include them in our definition of potentially inappropriate prescribing.
In summary, we found that uptake of DOACs is increasing at an accelerating rate in the VA system, but that utilization remains highly variable by site. We also found an approximately 4.58% rate of potentially inappropriate prescribing of DOACs across VA, with some sites having a rate that was 2 or even 3 times as high, and other sites recording no inappropriate prescriptions. Our study has important implications for pharmacy budgets in an era when new medications challenge our attempts to seek value for our pharmacy dollar, and also has important implications for efforts to ensure safe and appropriate prescribing of anticoagulants.
Author Affiliations: Center for Healthcare Organization and Implementation Research, Bedford VA Medical Center (AJR, JIR, DRM), MA; Department of Medicine, Section of General Internal Medicine, Boston University School of Medicine (AJR), MA; VA Salt Lake City Healthcare System (ALA), UT; Department of Health Policy and Management, Boston University School of Public Health (DRM), MA. Integrated Benefits Institute (BG), San Francisco, CA.
Source of Funding: Supported by a locally initiated project from the VA Health Services Research & Development Center for Healthcare Organization and Implementation Research. The sponsor had no role in the design and conduct of the study; the collection, management, analysis, and interpretation of the data; or the preparation, review, and approval of the manuscript. The opinions expressed in this manuscript do not necessarily represent the official views of the Department of Veterans Affairs.
Author Disclosures: Dr Allen is on the speakers’ bureau for Janssen Pharmaceuticals and is a consultant for Boehringer Ingelheim Pharmaceuticals. The other authors report no relationship or financial interest with any entity that would pose a conflict of interest with the subject matter of this article.
Authorship Information: [add] Dr Rose had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Address correspondence to: Adam J. Rose, MD, MSc, Center for Healthcare Organization and Implementation Research, Bedford VA Medical Center, 200 Springs Rd, Bldg 70, Bedford, MA 01730. E-mail: email@example.com.
1. VA Pharmacy Benefits Management Service. Target Specific Oral Anticoagulants: Criteria for Use for Stroke Prevention in Nonvalvular Atrial Fibrillation. January, 2014. www.pbm.va.gov. Accessed December 16, 2015.
2. Kirley K, Qato DM, Kornfield R, Stafford RS, Alexander GC. National trends in oral anticoagulant use in the United States, 2007 to 2011. Circ Cardiovasc Qual Outcomes. 2012;5(5):615-621.
3. Boehringer Ingelheim: Dabigatran Prescribing Information. Available at: http://bidocs.boehringer-ingelheim.com/BIWebAccess/ViewServlet.ser?docBase=renetnt&folderPath=/Prescribing%20Information/PIs/Pradaxa/Pradaxa.pdf. Accessed April 27, 2015.
4. Janssen Pharmaceuticals: Rivaroxaban Prescribing Information. Available at: http://www.xareltohcp.com/sites/default/files/pdf/xarelto_0.pdf. Accessed April 27, 2015.
5. Bristol-Myers Squibb. Apixaban prescribing information. Available at: http://packageinserts.bms.com/pi/pi_eliquis.pdf. Accessed April 27, 2015.
6. Eikelboom JW, Connolly SJ, Brueckmann M, et al; RE-ALIGN Investigators. Dabigatran versus warfarin in patients with mechanical heart valves. N Engl J Med. 2013;369(13):1206-1214. doi: 10.1056/NEJMoa1300615.
7. Rose AJ, Hylek EM, Ozonoff A, Ash AS, Reisman JI, Berlowitz DR. Patient characteristics associated with oral anticoagulation control: results of the Veterans AffaiRs Study to Improve Anticoagulation (VARIA). J Thromb Haemost. 2010;8(10):2182-2191. doi: 10.1111/j.1538-7836.2010.03996.x.
8. Connolly SJ, Ezekowitz MD, Yusuf S, et al; RE-LY Steering Committee and Investigators. Dabigatran versus warfarin in patients with atrial fibrillation. N Engl J Med. 2009;361(12):1139-1151. doi: 10.1056/NEJMoa0905561.
9. Granger CB, Armaganijan LV. Newer oral anticoagulants should be used as first-line agents to prevent thromboembolism in patients with atrial fibrillation and risk factors for stroke or thromboembolism. Circulation. 2012;125(1):159-164; discussion 164. doi: 10.1161/CIRCULATIONAHA.111.031146.
10. Patel MR, Mahaffey KW, Garg J, et al; ROCKET AF Investigators. Rivaroxaban versus warfarin in nonvalvular atrial fibrillation. N Engl J Med. 2011;365(10):883-891. doi: 10.1056/NEJMoa1009638.
11. Levey AS, Bosch JP, Lewis JB, Greene T, Rogers N, Roth D. A more accurate method to estimate glomerular filtration rate from serum creatinine: a new prediction equation. Modification of Diet in Renal Disease Study Group. Ann Intern Med. 1999;130(6):461-470.
12. Doogue MP, Polasek TM. Drug dosing in renal disease. Clin Biochem Rev. 2011;32(2):69-73.
13. National Institute of Diabetes and Digestive and Kidney Diseases: National Kidney Disease Education Program. Laboratory Evaluation: Estimating GFR. http://nkdep.nih.gov/lab-evaluation/gfr/estimating.shtml. Accessed December 16, 2015.
14. Earley A, Miskulin D, Lamb EJ, Levey AS, Uhlig K. Estimating equations for glomerular filtration rate in the era of creatinine standardization: a systematic review. Ann Intern Med. 2012;156(11):785-795. W-270, W-271, W-272, W-273, W-274, W-275, W-276, W-277, W-278. doi: 10.7326/0003-4819-156-6-201203200-00391.
15. Freeman JV, Zhu RP, Owens DK, et al. Cost-effectiveness of dabigatran compared with warfarin for stroke prevention in atrial fibrillation. Ann Intern Med. 2011(1);154:1-11. doi: 10.7326/0003-4819-154-1-201101040-00289.
16. Miyasaka Y, Barnes ME, Gersh BJ, et al. Secular trends in incidence of atrial fibrillation in Olmsted County, Minnesota, 1980 to 2000, and implications on the projections for future prevalence. Circulation. 2006;114(2):119-125.
17. VA Pharmacy Benefits Management. Dabigatran: Criteria for Use. Version: December 2011. www.pbm.va.gov. Accessed May 3, 2012.
18. McDonald CJ, Kalisch Ellett LM, Barratt JD, Caughey GE. An international comparison of spontaneous adverse event reports and potentially inappropriate medicine use associated with dabigatran. Pharmacoepidemiol Drug Saf. 2015;24(4):399-405. doi: 10.1002/pds.3648.