Impact of Delayed Adoption of Novel Atrial Fibrillation Treatments

ABSTRACT

Objective: To examine the relationship between adoption of direct oral anticoagulants (DOACs) and health and cost outcomes for patients with nonvalvular atrial fibrillation.

Study Design: Real-world cohort study.

Methods: US adults who newly initiated treatment for nonvalvular atrial fibrillation were identified from claims data. DOAC adoption and stroke rates were assessed at metropolitan statistical area (MSA) and individual levels. The MSA-level cross-sectional analysis examined the relationship between the adoption rate of a DOAC (vs warfarin) and an ischemic stroke. The individual-level instrumental variable analysis examined the impact of treatment choice predicted by regional adoption on stroke within 1 year after treatment initiation. Results were extrapolated to estimate the strokes and costs averted by patients moving from a slow-adopting (10th percentile) MSA to a rapid-adopting (90th percentile) MSA.

Results: DOAC uptake rates in MSAs at the 10th and 90th uptake percentile were 53.1% and 78.5%, respectively, in 2014. Overall DOAC uptake increased from 66.3% in 2014 to 91.4% in 2018. Increased DOAC adoption reduced average stroke rates by 1.41 percentage points or 63.2% (P = .002) using the MSA-level descriptive analysis and 1.08 percentage points or 71.2% (P = .002) using the individual-level instrumental variable analysis. Nationally, shifting DOAC rates from those seen in slow-adopting MSAs to those seen in rapid-adopting MSAs could avert up to 32,000 strokes and save up to $1.04 billion annually.

Conclusions: More rapid adoption of newly approved nonvalvular atrial fibrillation treatments was associated with reduced stroke rates and high cost savings. Managed care organizations should consider how delays in the uptake of innovative medications impact health and economic outcomes.

Am J Manag Care. 2024;30(12):In Press

Takeaway Points

• Real-world adoption of new pharmaceutical technologies can be slow.
• Using direct oral anticoagulant treatments for nonvalvular atrial fibrillation as a case study, the authors found that more rapid adoption of newly approved treatments was associated with improved health outcomes and cost savings.
• Managed care payers should consider a variety of mechanisms to increase speed of adoption for efficacious, cost-effective treatments.

The introduction of new pharmaceuticals has helped improve health outcomes for patients. Lung cancer mortality improved by 5.0% in 2018, largely due to new immuno-oncology and targeted treatments.¹ New cell and gene therapies have more than 4 times the health benefits compared with non–cell and gene therapies.² And statin use reduced all-cause and cardiovascular mortality in older adults by 57.1% and 55.6%, respectively (2002-2012).³

On average, it takes approximately 17 years from substance development and testing to market implementation.⁴ Then, despite the advances, significant delays in adoption exist for various reasons. From the insurer’s perspective, new treatment introduction requires a process of coverage decision-making, decisions on access restrictions such as prior authorization and step edits, decisions on co-pays and updates to coding/billing systems, and time needed for medical societies to update clinical guidelines.^5,6 Lack of physician awareness of new treatments also could contribute to adoption delays.^7,8 Finally, physicians’ persistence in treatment choice due to familiarity and experience could delay new treatments from becoming broadly adopted by patients.⁹ However, previous research has shown that treatment delays worsen health outcomes for certain diseases (eg, cystic fibrosis, cancer).^10,11

Against this backdrop, this study explored the clinical and economic impacts of adoption delays in novel medical treatment using direct oral anticoagulants (DOACs) for nonvalvular atrial fibrillation (NVAF) as a case study. NVAF, a cardiac arrhythmia, can lead to serious cardiovascular events, such as ischemic stroke. Individuals with AF are at approximately 5-fold increased risk of ischemic stroke,¹² and AF prevalence in the US population is expected to increase to more than 12 million by 2030.¹³ AF can be treated with warfarin¹⁴; however, DOACs have been shown to further reduce the relative risk of stroke compared with warfarin¹⁵ and improve health outcomes, as shown by results of real-world comparative effectiveness studies.^16-18

This study utilized US geographical variation in real-world DOAC adoption for NVAF to test the hypothesis that patients in regions with slower treatment adoption are less likely to receive newly approved treatments, thus potentially leading to worse health outcomes. This study (1) used a metropolitan statistical area (MSA)–level geographical variation of DOAC adoption rate for a descriptive analysis and (2) conducted an individual-level pseudoexperimental analysis by leveraging a likelihood of receiving DOAC treatment predicted by regional adoption.

METHODS

Data

This study utilized the Merative MarketScan Commercial Claims and Encounters and the Medicare Supplemental claims databases (2013-2019). The MarketScan databases, containing information for 39 million insured individuals nationwide, capture individual- and insurer-level health care utilization and expenditures across inpatient services, outpatient services, physician care, and prescription drugs.¹⁹ Per-person health care utilization and expenditures can be aggregated to the MSA level with this data set. Data from 2020 to 2022 were excluded due to potential anomalies in care during the COVID-19 pandemic.

Patient Population

The patient population consisted of new users of relevant medications that treat NVAF. Patients were required to be 18 years and older and have initiated taking either warfarin or at least 1 DOAC. DOACs, which are recommended for treatment of NVAF,²⁰ include oral, direct factor Xa inhibitors (apixaban [Eliquis], edoxaban [Lixiana], and rivaroxaban [Xarelto]) and a direct thrombin inhibitor (dabigatran [Pradaxa]). Patients were required to have an NVAF diagnosis, defined as at least 1 inpatient or at least 2 outpatient claims with an AF diagnosis (International Classification of Diseases, Ninth Revision [ICD-9] code 427.31 or International Statistical Classification of Diseases, Tenth Revision [ICD-10] codes I48.0, I48.1, I48.11, I48.19, I48.2, I48.20, I48.21, and I48.91),^21,22 and continuous insurance enrollment for at least 12 months prior to the date when treatment was first observed (hereafter, the index date).

Patients who were diagnosed with valvular AF (ICD-9 codes 394.0, 394.2, 396.0, 396.1, 396.8, and 424.0 or ICD-10 codes I05.0, I05.2, I08.0, I08.8, I34.2, I34.8, and I34.9)^21,23 or resided in an MSA with fewer than 11 individuals were excluded.

Study Design

This study was an observational cohort study. First, we analyzed trends in DOAC uptake over time and across different geographic regions. Second, the impact of DOAC adoption rates on ischemic stroke outcomes in MSA- and individual-level regression analyses were evaluated. Finally, we extrapolated regression results to identify the hypothetical impact of patients moving from a geographic region with slow (10th percentile) to rapid (90th percentile) DOAC adoption.

Variables

The key outcomes used to measure the impact of delayed DOAC adoption for patients with NVAF were the share of patients with NVAF in a given MSA who experienced ischemic stroke (ICD-9 codes 433.01, 433.11, 433.21, 433.31, 433.81, 433.91, 434.xx, and 436.xx or ICD-10 code I63) in the calendar year after initiating treatment and a binary variable for the individual-level analysis indicating whether the patient experienced a stroke within 1 year of treatment.^21,24

In the MSA-level analysis, the exposure variable was the share of individuals in a given MSA who used a DOAC to treat NVAF in the previous year (eAppendix Figure [eAppendix available at ajmc.com]) to proxy for DOAC adoption (hereafter, uptake rate). A binary indicator for whether there was use of DOAC treatment (ie, uptake) compared with warfarin was employed as the exposure variable in the individual-level analysis.

Control variables were incorporated in both MSA- and individual-level analyses. Demographic control variables included patient age, sex, and insurance type (commercial health maintenance organization, preferred provider organization, point-of-service, or fee-for-service vs Medicare Advantage). Comorbidities were measured using the Charlson Comorbidity Index (CCI).²⁵ Disease severity was measured using the CHA₂DS₂-VASc (congestive heart failure, hypertension, age ≥ 75 years [doubled], diabetes, prior stroke/transient ischemic attack/thromboembolism [doubled], vascular disease, age 65-74 years, sex category) score, which predicts risk of stroke among patients with NVAF²¹ and has previously been validated using claims data.²⁶ The number of hospitalizations patients had in the year prior to the index date was also included.

Statistical Analysis

The clinical effect of delay in adoption of DOACs was quantified using multiple statistical models. First, for the cross-sectional analysis at the MSA level, we leveraged the geographical variation in the adoption rate of DOACs to determine its effect on MSA-level stroke rates.²⁷ SEs were clustered at the MSA level. The advantage to this descriptive model includes explicitly identifying the exposure of the study—the rate of DOAC adoption. Additionally, this approach mitigates the potential concern of endogeneity in the individual-level ordinary least squares model, which can be created by selection bias (eg, patients with more severe disease may be more likely to begin a DOAC regimen) because average severity within an MSA would show less variation than severity across individuals and would be less likely to affect DOAC uptake rates for the MSA.^28,29 However, analyzing uptake by MSA is susceptible to aggregation bias.³⁰

To address both aggregation bias and endogeneity, we utilized an individual-level, 2-stage least squares (2SLS) regression with an instrumental variable. In clinical trials, randomization (eg, a random number generator) is used to proactively assign treatment in a way that is independent of expected outcomes; in the real world, the 2SLS approach aims to do similarly by identifying some third factor (ie, the “instrument”) that is correlated with the treatment that individuals receive but is not predictive of outcomes (except to the extent that the treatment itself has an impact). In our study, the exposure (ie, treatment) and outcome variables were individual DOAC uptake and stroke within 1 year of the index date, respectively; the instrument was the share of newly treated patients in a given patient’s own MSA-year who received a DOAC, without including the selected patient in this share calculation. MSA-level uptake of DOACs is a reasonable instrument because it is likely correlated with the patient’s own use of a DOAC (ie, satisfying the “relevance” assumption for instrumental variables) because health systems within the same geographic area are more likely to have similar prescribing patterns compared with neighboring health systems³¹; moreover, a patient’s own likelihood of stroke is likely largely uncorrelated with the outcomes of other patients in the same MSA (ie, satisfying the “exclusion restriction” assumption).³² The latter condition does not hold if the patient’s own treatment choice is considered in the uptake share calculation. The 2SLS approach has been used in prescription drug and other health care real-world data studies to address these sources of bias.^33-35 SEs were clustered at the individual level. All covariates included in the second-stage regression were used to predict the endogenous variable in the first-stage regression.

The individual-level 2SLS approach implicitly identifies the impact of the delayed adoption at the MSA level on stroke outcomes. Because the exposure of the model predicted by the instrumental variable in the first stage of 2SLS indicates the likelihood that a patient with NVAF residing in an MSA receives DOAC treatment, the second stage of the model estimates the marginal effect of the degree that DOACs are adopted within the MSA on the individual risk of stroke. In short, an individual’s MSA serves as a pseudoexperiment, where individuals in slow-adopting MSAs have a lower chance of being in the treatment arm (DOAC users) compared with individuals in a rapid-adopting MSA.

Using predictions from both MSA- and individual-level 2SLS regression analyses, we estimated the impact on stroke rate if a patient moved from a geographic region with slow adoption of DOACs (10th percentile of the DOAC adoption rate) to one with rapid adoption of DOACs (90th percentile). The number of averted strokes and corresponding reductions in health care costs from previously published literature were generated by extrapolating regression results.

Sensitivity Analysis

Additional regression analyses were evaluated as sensitivity analyses to demonstrate model robustness by (1) estimating the impact of DOAC adoption on time to first stroke using a Cox proportional hazard model after propensity score–matching individuals who used DOACs to similar individuals who used warfarin (eAppendix) and (2) conducting an individual-level, repeated cross-sectional analysis using a logit regression model.

RESULTS

Sample Construction and Descriptive Statistics

The data included 2.6 million adults treated for NVAF between 2014 and 2018. After applying the inclusion and exclusion criteria, 108,461 individuals from 200 MSAs were eligible for inclusion (Figure 1). Patients and MSA counts by year are found in eAppendix Table 1.

The DOAC group was younger and less likely to enroll in Medicare than the control group (mean [SD] age, 67.2 [12.9] years vs 70.9 [12.4] years, respectively; P < .001; Medicare patient share, 53.6% vs 67.8%; P < .001) (Table). The mean CCI scores, CHA₂DS₂-VASc scores, and hospital visits indicated that the treatment group was healthier relative to the control group (mean CCI score, 4.128 vs 5.288, respectively; P < .001; mean CHA₂DS₂-VASc score, 2.870 vs 3.494; P < .001; mean hospital visits, 0.711 vs 1.055; P < .001).

Impact of Increased DOAC Adoption Rate on Stroke Rate

DOAC adoption increased over the study period (Figure 2 and eAppendix Table 2). Use of DOACs for newly treated patients with NVAF was 66.3% in 2014 and increased to 91.4% by 2018. Despite increased DOAC use overall, there was significant variability across geographical areas (eAppendix Tables 3 and 4). In 2014, the gap in DOAC uptake rates between MSAs at the 10th and 90th percentiles was 25.4 percentage points (PP; 53.1% vs 78.5%, respectively). However, this gap shrunk to 14.0 PP by 2018 (83.3% vs 97.3%).

The unadjusted 1-year stroke rate among patients with NVAF during the study period was 0.87%—consistent with previous literature (Table and eAppendix Table 5).^36,37 Stroke rates were lower in the treatment group compared with the control group (0.76% vs 1.25%; P < .001). The unadjusted stroke rate decreased throughout the study period, from 0.96% in 2014 to 0.51% by 2018 (Figure 2).

The adjusted stroke rate in the first year from the index date predicted by the MSA-level descriptive model showed a decrease of 1.41 PP for every 1 PP increase in DOAC adoption (0.57 vs 1.98 PP for DOAC adoption rates of 100% and 0%, respectively; 95% CI, 1.39-1.42; P = .002) (Figure 3 and eAppendix Table 6). Uptake of DOACs reduced the 1-year risk of ischemic stroke by 1.08 PP compared with warfarin (0.63 vs 1.71 PP for the treatment group vs control group; 95% CI, 1.07-1.08; P = .002). First-stage regression results suggested that the instrumental variable was significant (P = .002 for the Anderson-Rubin Wald test; P = .001 for the Stock-Wright Lagrange multiplier test) (eAppendix Table 7). These results imply that a patient with NVAF residing in an MSA where uptake of DOACs was delayed was exposed to a 63.16% (1.08 PP/1.71 PP from the individual-level model) to 71.21% (1.41 PP/1.98 PP from the MSA-level model) higher risk of stroke. Sensitivity analyses found similar results when conducting an analysis of time to first stroke: The stroke risk of the treatment group relative to the control group was 0.63 to 0.86 (eAppendix Table 6).

Extrapolation to the US Population

Adoption of DOACs decreased stroke rates by 0.44 PP (1.09 vs 0.65 PP for MSAs at the 10th vs 90th percentiles of the DOAC adoption rate, respectively) and 0.34 PP (1.03 vs 0.69 PP) using the MSA-level descriptive and individual-level 2SLS models, respectively. This implies that if a patient moved from a geographic region with slow DOAC adoption (10th-percentile MSA) to one with rapid adoption (90th-percentile MSA), the annual likelihood of having a stroke would fall by 40.4% (–0.44%/1.09%) and 33.0% (– 0.34%/1.03%), respectively (Figure 4).

Extrapolating the results to the US population with NVAF (7.3 million^21,38), the MSA-level descriptive and individual-level 2SLS analyses predicted that moving uptake from 10th-percentile to 90th-percentile MSAs could have averted 32,954 and 24,932 strokes, respectively, per year during the study period. Using the 2023 inflation-adjusted annual cost of a stroke ($31,508),³⁹ increasing uptake rates from 10th-percentile to 90th-percentile MSAs would have saved the US between $786 million and $1.04 billion in 2023 US$ from averted strokes annually during the study period (eAppendix Table 8).

DISCUSSION

This study is the first to identify the clinical and economic impacts of delayed real-world adoption of new medical treatments across geographic areas. In this case study, patients taking a DOAC showed a lower observed 1-year stroke rate compared with patients taking warfarin (0.76% vs 1.25%, respectively) (Table). Regression results indicated that increasing uptake of DOAC use for newly treated patients with NVAF reduced ischemic stroke by approximately 1.08 to 1.41 PP, for a reduction in risk of 63.16% to 71.21%. Moreover, increasing NVAF novel treatment uptake in MSAs at the 10th percentile to the 90th percentile in terms of DOAC uptake rate could avert up to 32,954 strokes annually, saving the US up to $1.04 billion each year.

Results suggest that delayed adoption of a medicine after regulatory approval may unintentionally harm patients. A recent study showed that payers spend approximately $6.0 billion annually on utilization management and that physicians spend approximately $26.7 billion in time per year navigating utilization management.⁴⁰ A systematic literature review also demonstrated that prior authorization caused delays; moreover, higher levels of patient cost sharing increased medication abandonment while generally decreasing medication initiation and persistence.⁴¹ Health systems should consider the disadvantages of delayed adoption of a new medical treatment as one of the critical factors in optimal prescribing. At the same time, however, it is worth considering the real-world effectiveness of a new therapy, given that clinical practice settings differ from clinical trial settings.⁴²

Limitations and Strengths

This study has several limitations. First, because this was a retrospective analysis of real-world claims data, the risk of bias was higher compared with a randomized controlled trial. To address this issue, we used MSA-level uptake to pseudorandomly assign patients into a high vs low likelihood of receiving a DOAC based on the geographic region of physician practices. Second, the study included only individuals with commercial or Medicare insurance and underrepresented Medicaid and uninsured populations. However, the claims data set was large and geographically diverse, which allowed the capture of treatment patterns in the commercial and Medicare populations broadly across the US. Third, this study did not impose minimum exposure requirements for patients receiving treatment and did not require that patients remain on treatment during the year in which outcomes were assessed. Based on results from the Garfield-AF Registry (NCT01090362), a global prospective study of patients with AF, patients who discontinued oral anticoagulant therapy for 7 or more days were associated with 2 times higher risk of stroke than those who did not discontinue (HR, 2.21; 95% CI, 1.42-3.44).⁴³ A study done by Yao et al showed that patients treated with DOACs (rivaroxaban, apixaban, or dabigatran) had better adherence to the treatment than those treated with warfarin.⁴⁴ Increasing the time exposure could better estimate the long-run treatment impacts. Given the 12-month exposure of this study, the impact of delayed adoption of DOACs is likely underestimated. Fourth, this study did not evaluate bleeding incidence. However, according to a patient-level network meta-analysis of 4 pivotal randomized clinical trials of DOACs vs warfarin in AF, standard-dose DOACs were associated with a significantly lower hazard of intracranial bleeding and major bleeding compared with warfarin.⁴⁵ Thus, the overall benefits of uptake of DOACs could be even higher if the positive impact of reduced incidence of bleeding were included. Fifth, due to the primary outcome for the MSA analysis being ischemic stroke in the calendar year after initiating treatment, it is possible that individuals who initiated treatment later in a year (eg, December 2016) could have a stroke less than 1 year after treatment initiation (eg, January 2017) (eAppendix Figure). However, this issue would only occur if treatment initiation occurred disproportionally during the year. Sixth, incremental cost of DOACs was not incorporated; however, multiple studies have shown DOACs to be highly cost-effective.^46-48 Seventh, specific causes of slow uptake or how to alleviate these issues were not assessed. However, insurance coverage and utilization management policies likely play a role; those that require multiple visits to a physician, require patients and providers to navigate complex prior authorization rules, or increase patient cost sharing through highly restrictive formulary tiering or exclusion practices are clearly barriers to uptake. Last, the study excluded patients who could benefit from anticoagulation but do not receive it due to other medical reasons (eg, contraindications),⁴⁹ which may introduce some bias in the instrumental variable estimation because untreated individuals were excluded from our analysis.⁵⁰

Despite these limitations, this study has several strengths. First, stroke rates uncovered in this study were consistent with previous studies.^36,37 Second, study results were consistent between both statistical models. Lastly, regression results from sensitivity analyses conceptually demonstrate that increased DOAC uptake is associated with decreased stroke incidence.

CONCLUSIONS

Faster uptake of DOACs among patients with NVAF could reduce ischemic stroke rates and decrease medical cost. Specifically, a 1 PP increase in DOAC uptake was estimated to reduce stroke rates by up to 1.41 PP. Policies that slow uptake of novel therapies may have unintended consequences. Achieving more rapid access to new treatments, however, would require addressing the barriers to adoption (eg, utilization management, step edits, lack of physician awareness, lack of patient education) to ensure broader adoption of beneficial new health technologies. 

Acknowledgments

The authors would like to thank Moises Marin for his assistance with writing up portions of the manuscript. They also thank Danielle Rollmann, Zhongyun Zhao, Jennifer Malinowski, Silas Martin, and Brahim Bookhart for their assistance in providing initial comments on this manuscript.

Author Affiliations: Center for Healthcare Economics and Policy, FTI Consulting, Inc, Los Angeles, CA (JK, JS), and Washington, DC (JN); Janssen Scientific Affairs, LLC (AGD, JH), Titusville, NJ.

Source of Funding: This study was supported by Johnson & Johnson.

Author Disclosures: Dr Kim, Dr Nighohossian, and Dr Shafrin are employees of FTI Consulting, a consulting firm to health care and life sciences entities, and report receiving funding for work on this manuscript from Janssen. Dr Daifotis and Dr He are employed by Janssen, which is a company of Johnson & Johnson (J&J) that markets a direct oral anticoagulant. Dr Daifotis owns stock in J&J and options. Dr He owns stock in J&J and Merck & Co, Inc.

Authorship Information: Concept and design (JK, JN, AGD, JH, JS); acquisition of data (JN); analysis and interpretation of data (JK, JN, AGD, JH, JS); drafting of the manuscript (JK, JN, JS); critical revision of the manuscript for important intellectual content (JK, AGD, JH, JS); statistical analysis (JK, JS); administrative, technical, or logistic support (JK, JN, JS); and supervision (JN, AGD, JS).

Address Correspondence to: Jason Shafrin, PhD, Center for Healthcare Economics and Policy, FTI Consulting, 350 S Grande Ave, Los Angeles, CA 90071. Email: jason.shafrin@fticonsulting.com.

REFERENCES

1. Siegel RL, Miller KD, Fuchs HE, Jemal A. Cancer statistics, 2022. CA Cancer J Clin. 2022;72(1):7-33. doi:10.3322/caac.21708

2. Shafrin J, Quddus S, Marin M, Scanlon D. A decade of health innovation: the impact of new medicines on patient health and the implications for NICE’s size of benefit multiplier. Value Health. 2023;26(10):1435-1439. doi:10.1016/j.jval.2023.06.009

3. Orkaby AR, Driver JA, Ho YL, et al. Association of statin use with all-cause and cardiovascular mortality in US veterans 75 years and older. JAMA. 2020;324(1):68-78. doi:10.1001/jama.2020.7848

4. Balas EA, Boren SA. Managing clinical knowledge for health care improvement. Yearb Med Inform. 2000:(1):65-70.

5. Anderson KE, Darden M, Jain A. Improving prior authorization in Medicare Advantage. JAMA. 2022;328(15):1497-1498. doi:10.1001/jama.2022.17732

6. Gustavsen G, Phillips K, Pothier K. The Reimbursement Landscape for Novel Diagnostics: Current Limitations, Real-World Impact, and Proposed Solutions. Health Advances; 2010. Accessed June 15, 2022. https://web.archive.org/web/20231015154023/https://www.bio.org/sites/default/files/legacy/bioorg/docs/Health_Advances&BIO_Novel_Diagnostics_Reimburs_20110103.pdf

7. Lin EH, Simon GE, Katzelnick DJ, Pearson SD. Does physician education on depression management improve treatment in primary care? J Gen Intern Med. 2001;16(9):614-619. doi:10.1046/j.1525-1497.2001.016009614.x

8. Nagtegaal R, Tummers L, Noordegraaf M, Bekkers V. Nudging healthcare professionals towards evidence-based medicine: a systematic scoping review. J Behav Public Adm. 2019;2(2). Accessed June 15, 2022. https://journal-bpa.org/index.php/jbpa/article/view/71

9. Janakiraman R, Dutta S, Sismeiro C, Stern P. Physicians’ persistence and its implications for their response to promotion of prescription drugs. Manage Sci. 2008;54(6):1080-1093. doi:10.1287/mnsc.1070.0799

10. Stanojevic S, Vukovojac K, Sykes J, Ratjen F, Tullis E, Stephenson AL. Projecting the impact of delayed access to elexacaftor/tezacaftor/ivacaftor for people with cystic fibrosis. J Cyst Fibros. 2021;20(2):243-249. doi:10.1016/j.jcf.2020.07.017

11. Hanna TP, King WD, Thibodeau S, et al. Mortality due to cancer treatment delay: systematic review and meta-analysis. BMJ. 2020;371:m4087. doi:10.1136/bmj.m4087

12. Tsao CW, Aday AW, Almarzooq ZI, et al. Heart disease and stroke statistics—2022 update: a report from the American Heart Association. Circulation. 2022;145(8):e153-e639. doi:10.1161/CIR.0000000000001052

13. Colilla S, Crow A, Petkun W, Singer DE, Simon T, Liu X. Estimates of current and future incidence and prevalence of atrial fibrillation in the U.S. adult population. Am J Cardiol. 2013;112(8):1142-1147. doi:10.1016/j.amjcard.2013.05.063

14. Hart RG, Pearce LA, Aguilar MI. Meta-analysis: antithrombotic therapy to prevent stroke in patients who have nonvalvular atrial fibrillation. Ann Intern Med. 2007;146(12):857-867. doi:10.7326/0003-4819-146-12-200706190-00007

15. Ntaios G, Papavasileiou V, Diener HC, Makaritsis K, Michel P. Nonvitamin-K-antagonist oral anticoagulants versus warfarin in patients with atrial fibrillation and previous stroke or transient ischemic attack: an updated systematic review and meta-analysis of randomized controlled trials. Int J Stroke. 2017;12(6):589-596. doi:10.1177/1747493017700663

16. Escobar C, Martí-Almor J, Cabeza AP, Martínez-Zapata MJ. Direct oral anticoagulants versus vitamin K antagonists in real-life patients with atrial fibrillation. a systematic review and meta-analysis. Rev Esp Cardiol (Engl Ed). 2019;72(4):305-316. doi:10.1016/j.rec.2018.03.009

17. Hua Y, Sun JY, Su Y, et al. The safety and efficacy of rivaroxaban compared with warfarin in patients with atrial fibrillation and diabetes: a systematic review and meta-analysis. Am J Cardiovasc Drugs. 2021;21(1):51-61. doi:10.1007/s40256-020-00407-z

18. Baker WL, Roberts MS, Bessada Y, et al. Comparative outcomes associated with rivaroxaban versus warfarin use in elderly patients with atrial fibrillation or acute venous thromboembolism managed in the United States: a systematic review of observational studies. Curr Med Res Opin. 2023;39(9):1183-1194. doi:10.1080/03007995.2023.2247988

19. Merative MarketScan Research Databases. Merative. 2022. Accessed June 15, 2022. https://www.merative.com/content/dam/merative/documents/brief/marketscan-explainer-general.pdf

20. January CT, Wann LS, Calkins H, et al. 2019 AHA/ACC/HRS focused update of the 2014 AHA/ACC/HRS guideline for the management of patients with atrial fibrillation: a report of the American College of Cardiology/American Heart Association Task Force on Clinical Practice Guidelines and the Heart Rhythm Society in collaboration with the Society of Thoracic Surgeons. Circulation. 2019;140(2):e125-e151. doi:10.1161/CIR.0000000000000665

21. Turakhia MP, Shafrin J, Bognar K, et al. Estimated prevalence of undiagnosed atrial fibrillation in the United States. PLoS One. 2018;13(4):e0195088. doi:10.1371/journal.pone.0195088

22. Osman M, Deshmukh AJ, Holmes DR, Alkhouli M. Racial differences in the prevalence of diagnosed atrial fibrillation among hospitalized patients. Mayo Clin Proc. 2021;96(9):2495-2497. doi:10.1016/j.mayocp.2021.07.006

23. Rome BN, Gagne JJ, Avorn J, Kesselheim AS. Non-warfarin oral anticoagulant copayments and adherence in atrial fibrillation: a population-based cohort study. Am Heart J. 2021;233:109-121. doi:10.1016/j.ahj.2020.12.010

24. Tirschwell DL, Longstreth WT Jr. Validating administrative data in stroke research. Stroke. 2002;33(10):2465-2470. doi:10.1161/01.str.0000032240.28636.bd

25. Quan H, Sundararajan V, Halfon P, et al. Coding algorithms for defining comorbidities in ICD-9-CM and ICD-10 administrative data. Med Care. 2005;43(11):1130-1139. doi:10.1097/01.mlr.0000182534.19832.83

26. Reinhardt SW, Chouairi F, Miller PE, et al. National trends in the burden of atrial fibrillation during hospital admissions for heart failure. J Am Heart Assoc. 2021;10(11):e019412. doi:10.1161/JAHA.120.019412

27. Cameron AC, Miller DL. A practitioner’s guide to cluster-robust inference. J Hum Resour. 2015;50(2):317-372. doi:10.3368/jhr.50.2.317

28. Shafrin J, Ganguli A, Gonzalez YS, Shim JJ, Seabury SA. Geographic variation in the quality and cost of care for patients with rheumatoid arthritis. J Manag Care Spec Pharm. 2016;22(12):1472-1481. doi:10.18553/jmcp.2016.22.12.1472

29. Shafrin J, Griffith J, Shim JJ, Huber C, Ganguli A, Aubry W. Geographic variation in diagnostic ability and quality of care metrics: a case study of ankylosing spondylitis and low back pain. Inquiry. 2017;54:46958017707873. doi:10.1177/0046958017707873

30. Luloff AE, Greenwood PH; New Hampshire Agricultural Experiment Station. Definitions of community: an illustration of aggregation bias, Station Bulletin, no.516. NHAES Bulletin. 1980;477. Accessed June 15, 2022. https://scholars.unh.edu/agbulletin/477

31. Zeliadt SB, Potosky AL, Penson DF, Etzioni R. Survival benefit associated with adjuvant androgen deprivation therapy combined with radiotherapy for high- and low-risk patients with nonmetastatic prostate cancer. Int J Radiat Oncol Biol Phys. 2006;66(2):395-402. doi:10.1016/j.ijrobp.2006.04.048

32. Labrecque J, Swanson SA. Understanding the assumptions underlying instrumental variable analyses: a brief review of falsification strategies and related tools. Curr Epidemiol Rep. 2018;5(3):214-220. doi:10.1007/s40471-018-0152-1

33. Chen Y, Briesacher BA. Use of instrumental variable in prescription drug research with observational data: a systematic review. J Clin Epidemiol. 2011;64(6):687-700. doi:10.1016/j.jclinepi.2010.09.006

34. Newhouse JP, McClellan M. Econometrics in outcomes research: the use of instrumental variables. Annu Rev Public Health. 1998;19:17-34. doi:10.1146/annurev.publhealth.19.1.17

35. Greenland S. An introduction to instrumental variables for epidemiologists. Int J Epidemiol. 2000;29(4):722-729. doi:10.1093/ije/29.4.722

36. Bengtson LGS, Lutsey PL, Chen LY, MacLehose RF, Alonso A. Comparative effectiveness of dabigatran and rivaroxaban versus warfarin for the treatment of non-valvular atrial fibrillation. J Cardiol. 2017;69(6):868-876. doi:10.1016/j.jjcc.2016.08.010

37. Go AS, Singer DE, Toh S, et al. Outcomes of dabigatran and warfarin for atrial fibrillation in contemporary practice: a retrospective cohort study. Ann Intern Med. 2017;167(12):845-854. doi:10.7326/M16-1157

38. QuickFacts: United States. US Census Bureau. Accessed April 22, 2022. https://www.census.gov/quickfacts/fact/table/US/AGE295221#AGE295220

39. Wang G, Joo H, Tong X, George MG. Hospital costs associated with atrial fibrillation for patients with ischemic stroke aged 18-64 years in the United States. Stroke. 2015;46(5):1314-1320. doi:10.1161/STROKEAHA.114.008563

40. Howell S, Yin PT, Robinson JC. Quantifying the economic burden of drug utilization management on payers, manufacturers, physicians, and patients. Health Aff (Millwood). 2021;40(8):1206-1214. doi:10.1377/hlthaff.2021.00036

41. Ismail WW, Witry MJ, Urmie JM. The association between cost sharing, prior authorization, and specialty drug utilization: a systematic review. J Manag Care Spec Pharm. 2023;29(5):449-463. doi:10.18553/jmcp.2023.29.5.449

42. Prasad V, Cifu A, Ioannidis JP. Reversals of established medical practices: evidence to abandon ship. JAMA. 2012;307(1):37-38. doi:10.1001/jama.2011.1960

43. Cools F, Johnson D, Camm AJ, et al; GARFIELD-AF Investigators. Risks associated with discontinuation of oral anticoagulation in newly diagnosed patients with atrial fibrillation: results from the GARFIELD-AF Registry. J Thromb Haemost. 2021;19(9):2322-2334. doi:10.1111/jth.15415

44. Yao X, Abraham NS, Alexander GC, et al. Effect of adherence to oral anticoagulants on risk of stroke and major bleeding among patients with atrial fibrillation. J Am Heart Assoc. 2016;5(2):e003074. doi:10.1161/JAHA.115.003074

45. Carnicelli AP, Hong H, Connolly SJ, et al; COMBINE AF (A Collaboration Between Multiple Institutions to Better Investigate Non-Vitamin K Antagonist Oral Anticoagulant Use in Atrial Fibrillation) Investigators. Direct oral anticoagulants versus warfarin in patients with atrial fibrillation: patient-level network meta-analyses of randomized clinical trials with interaction testing by age and sex. Circulation. 2022;145(4):242-255. doi:10.1161/CIRCULATIONAHA.121.056355

46. Lip GY, Kongnakorn T, Phatak H, et al. Cost-effectiveness of apixaban versus other new oral anticoagulants for stroke prevention in atrial fibrillation. Clin Ther. 2014;36(2):192-210.e20. doi:10.1016/j.clinthera.2013.12.011

47. Pinyol C, Cepeda JM, Roldan I, et al. A systematic literature review on the cost-effectiveness of apixaban for stroke prevention in non-valvular atrial fibrillation. Cardiol Ther. 2016;5(2):171-186. doi:10.1007/s40119-016-0066-2

48. Lee S, Anglade MW, Pham D, Pisacane R, Kluger J, Coleman CI. Cost-effectiveness of rivaroxaban compared to warfarin for stroke prevention in atrial fibrillation. Am J Cardiol. 2012;110(6):845-851. doi:10.1016/j.amjcard.2012.05.011

49. Steinberg BA, Ballew NG, Greiner MA, et al. Ischemic and bleeding outcomes in patients with atrial fibrillation and contraindications to oral anticoagulation. JACC Clin Electrophysiol. 2019;5(12):1384-1392. doi:10.1016/j.jacep.2019.07.011

50. Swanson SA, Robins JM, Miller M, Hernán MA. Selecting on treatment: a pervasive form of bias in instrumental variable analyses. Am J Epidemiol. 2015;181(3):191-197. doi:10.1093/aje/kwu284