Sex Differences in Oral Anticoagulation and Outcomes of Stroke and Intracranial Bleeding in Newly Diagnosed Atrial Fibrillation.

Background Female sex is an independent predictor of stroke in patients with atrial fibrillation (AF). Older data suggest undertreatment with anticoagulation among women compared with men. However, it is unknown if novel therapies and updated guidelines have impacted sex differences in AF treatment and outcomes. Methods and Results We performed a retrospective cohort study of 2.3 million women and men with a new diagnosis of AF and CHA2DS2-VASc ≥2 from Marketscan US commercial claims data from 2008 to 2015 to determine whether women with AF remain undertreated and whether this difference mediates observed differences in outcomes. There were 358 649 patients with newly diagnosed AF (43% women). Compared with men, women were older, with higher CHA2DS2-VASc scores, and higher comorbidity burden (P<0.0001 for all). Oral anticoagulation-eligible women with CHA2DS2-VASc scores ≥2 were more likely to not receive anticoagulation (50.0% women versus 43.9% men). Women, compared with men, had a higher risk of ischemic stroke (adjusted hazard ratio [aHR], 1.27; 95% CI, 1.21-1.32; P<0.0001) and hospitalization (aHR, 1.06; 95% CI, 1.05-1.07, P<0.0001) but had a lower risk of intracranial bleeding (aHR, 0.91; 95% CI, 0.83-0.99, P=0.03). In mediation analysis, nonreceipt of oral anticoagulation partially mediated the observed increased risk of stroke and decreased risk of intracranial bleeding in women. Conclusions In the care of newly diagnosed AF in the United States, women, compared with men, are less likely to receive oral anticoagulation. This appears to mediate the increased risk of both stroke and hospitalization but also appears to mediate lower observed intracranial bleeding risk.

atrial fibrillation

direct oral anticoagulant

oral anticoagulant

intracranial hemorrhage

Clinical Perspective

What Is New?

Compared with men, women with newly diagnosed atrial fibrillation were older, with higher CHA2DS2‐VASc scores and higher comorbidity burden.

Despite this, women were less likely to receive oral anticoagulation to reduce the risk of stroke, including direct oral anticoagulants.

Women, compared with men, had a higher risk of ischemic stroke and hospitalization but lower risk of intracranial bleeding.

What Are the Clinical Implications?

Oral anticoagulation among women partially mediated the observed risk differences by sex in ischemic stroke and hospitalization, suggesting an important target for improving outcomes in women with new atrial fibrillation.

Female sex is an independent predictor of stroke in patients with atrial fibrillation (AF), even among anticoagulated patients.1, 2, 3, 4 Consequently, female sex has been incorporated into risk stratification schemes and clinical guidelines for anticoagulation in AF.5, 6, 7 When vitamin K antagonists were the only oral anticoagulation option for stroke prevention, data suggested undertreatment in women compared with men,8 even though harms of therapy, such as major bleeding, were considered comparable.9 The direct oral anticoagulants (DOACs) have been shown in randomized trials to be at least as effective as warfarin for reduction of stroke but safer than warfarin regarding risk of intracranial hemorrhage (ICH) and, in most cases, all major bleeding.10, 11, 12 Despite relatively rapid diffusion of practice, as well as endorsement of DOACs as first‐line therapy in some professional society guidelines,13 it is not known whether sex differences persisted after the introduction of DOACs. We therefore sought to determine whether such differences were present and, if so, what factors may be associated with residual sex differences in therapy and outcomes.

Methods and Results

We analyzed data from the Truven Health MarketScan Commercials Claims and Encounters and Medicare Supplemental Databases (Truven Health Analytics Inc., Cambridge, MA). The databases capture person‐specific clinical use, expenditures, and enrollment across inpatient, outpatient, and prescription drug services. Data are generated from a selection of large employers, health plans, and government and public organizations. Linked and merged data sets that we used in the study include the Inpatient Admissions file, which contains records that summarize information about a hospital admission; the Outpatient Services file, which contains encounters and claims for services from a doctor's office, hospital outpatient facility, emergency room, or other outpatient facility; and the Outpatient Pharmaceutical Claims File and Enrollment Detail File. This data set has been extensively used for health services and outcomes research in AF and has been used in our prior work.14, 15 This study was approved by the local institutional review board (Stanford, CA) and the Veterans Affairs Research and Development Committee (Palo Alto, CA). Requirement for informed consent was waived.

We included all patients with a primary or secondary inpatient or outpatient diagnosis of AF (International Classification of Diseases, Ninth Revision [ICD‐9] code 427.31 or 427.32) from 2008 to 2014 (Figure 1). We selected this study period because it captured the period that DOACs were approved for AF and introduced to the market, specifically dabigatran (2010) and rivaroxaban (2011). In addition, patients were required to have no prior AF diagnosis in the previous year, continuous insurance enrollment in the MarketScan databases for at least 6 months before and 1 month after the index AF diagnosis date, a second confirmatory AF diagnosis between 30 and 365 days of the new AF diagnosis, and any outpatient medication within 90 days of the first AF diagnosis. The rationale is to increase specificity for AF in the cohort, minimize “rule‐out” diagnoses, and identify patients who continued to use the Veterans Affairs system for subsequent care. We have previously used this approach for prior claims‐cohort studies.16, 17

Figure 1

Cohort inclusion criteria.

Final analysis cohort included 358 649 patients.

Primary Predictor and Outcomes

The primary predictor was patient sex, obtained from claims data enrollment records. The primary outcome was outpatient drug receipt of any oral anticoagulant (OAC) and any DOAC (dabigatran, rivaroxaban, or apixaban; we did not evaluate edoxaban because the drug was not approved or available in the United States during the observation period). We also assessed time to clinical outcomes of ischemic stroke, intracerebral hemorrhage, and all‐cause hospitalization.

Clinical Covariates

Baseline comorbidities (cardiovascular and noncardiovascular) were determined using comorbidity‐specific ICD‐9 codes from up to 1 year before the new AF date based on our previous work.18 We assessed comorbidities using the Charlson and Selim Comorbidity Indices and assessed stroke risk using the CHA2DS2 and CHA2DS2‐VASc scores. Modified HAS‐BLED bleeding risk score, which takes into account hypertension, abnormal renal/liver function, stroke, bleeding, elderly age, and drugs/alcohol, was calculated as a measure of baseline bleeding risk.

Statistical Analysis

We compared sex differences in baseline characteristics using a chi‐square test for categorical variables and a Student t test for continuous variables. Univariable and multivariable Cox proportional hazards regression were used to examine the association between sex and OAC drug receipt, type of OAC, and each clinical outcome. In the multivariable model, we adjusted for age, region, insurance plan,19 Charlson Comorbidity Index score, congestive heart failure, hypertension, diabetes mellitus, and baseline medications. Logistic regression was conducted to test if anticoagulation drug receipt within 30 days before and up to 90 days following a new AF diagnosis mediated the effect of sex differences in outcomes. Given that indirect effects can work through a mediator of interest, we performed mediation analysis, using multivariable logistic regression, to determine whether the association of sex on clinical outcomes was mediated by anticoagulation. Mediation was assessed for in a stepwise fashion using the Baron and Kenny approach.20, 21 All analyses were performed using SAS, version 9.1 (Cary, NC) and STATA, version 11.0 (College Station, TX).

We identified 358 649 patients meeting our cohort inclusion and exclusion criteria (age 66.2±13.4; 43% women; CHA2DS2‐VASc score 2.7±1.8). Analysis of baseline characteristics (Table 1) demonstrated that, compared with men, women were older and had higher CHA2DS2VASc scores and higher Selim comorbidity indices. More women had a history of hypertension, stroke, and anemia, but fewer women had a history of diabetes mellitus and myocardial infarction compared with men. We found baseline sex differences in prescription of antiarrhythmic drugs, with more women prescribed antiarrhythmics than men (9.0% women versus 7.7% men, P<0.0001), and more women also on rate‐controlling medications than men (40.5% women versus 38.7% men, P<0.0001).

Table 1

Baseline Characteristics

	All Patients (N=358 649)	Male (N=205 756)	Female (N=152 893)	P Value
Female, %	43
Male, %	57
Age, y	66.2±13.4	64.2±13.2	68.9±13.3	<0.0001
Charlson Comorbidity Index	1.4±1.5	1.4±1.5	1.4±1.4	0.49
Selim Comorbidity Index	3.6±2.7	3.5±2.7	3.8±2.7	<0.0001
CHADS2 score	1.4±1.2	1.3±1.2	1.5±1.2	<0.0001
CHA2DS2‐VASc score	2.7±1.8	2.1±1.7	3.4±1.6	<0.0001
HAS‐BLED score	1.7±1.2	1.6±1.2	1.8±1.2	<0.0001
Disease, n (%)
Congestive heart failure	79 966 (22.3)	46 164 (22.4)	33 802 (22.1)	0.02
Hypertension	225 298 (62.8)	125 118 (60.8)	100 180 (65.5)	<0.0001
Diabetes mellitus	103 167 (28.8)	62 465 (30.4)	40 702 (26.6)	<0.0001
Prior stroke/transient ischemic attack	23 766 (6.6)	11 837 (5.8)	11 929 (7.8)	<0.0001
Prior myocardial infarction	21 239 (5.9)	13 992 (6.8)	7247 (4.7)	<0.0001
Anemia	50 944 (14.2)	25 841 (12.6)	25 103 (16.4)	<0.0001
Prior bleeding	38 273 (10.7)	22 380 (10.9)	15 893 (10.4)	<0.0001
Peripheral artery disease	28 378 (7.9)	16 015 (7.9)	12 363 (8.1)	0.0009
Chronic kidney disease	44 036 (12.3)	26 010 (12.6)	18 026 (11.8)	<0.0001
Region, n (%)				<0.0001
Northeast	67 105 (18.7)	38 312 (18.6)	28 793 (18.8)
North Central	113 898 (31.8)	64 019 (31.1)	49 879 (32.6)
South	114 281 (31.9)	66 328 (32.2)	47 953 (31.4)
West	60 692 (16.9)	35 444 (17.2)	25 248 (16.5)
Unknown	2673 (0.8)	1653 (0.8)	1020 (0.7)
Insurance plan type, n (%)				<0.0001
Comprehensive	110 584 (30.8)	56 998 (27.7)	53 586 (35.1)
EPO	1534 (0.4)	1084 (0.5)	450 (0.3)
HMO	44 510 (12.4)	25 692 (12.5)	18 818 (12.3)
POS	18 463 (5.2)	11 122 (5.4)	7341 (4.8)
PPO	163 913 (45.7)	98 237 (47.7)	65 676 (43.0)
POS with capitation	887 (0.3)	610 (0.3)	277 (0.2)
CDHP	7212 (2.0)	4708 (2.3)	2504 (1.6)
HDHP	2897 (0.8)	2025 (1.0)	872 (0.6)
Missing	8649 (2.4)	5280 (2.6)	3369 (2.2)
Baseline medications
Cardiovascular medications, n (%)
Aspirin	3608 (1.0)	1835 (0.9)	1773 (1.2)	<0.0001
Warfarin	49 927 (13.9)	29 061 (14.1)	20 866 (13.7)	<0.0001
Dabigatran	1817 (0.5)	1153 (0.6)	664 (0.4)	<0.0001
Rivaroxaban	1182 (0.3)	663 (0.3)	519 (0.3)	0.37
Clopidogrel	34 498 (9.6)	21 707 (10.6)	12 791 (8.4)	<0.0001
Apixaban	160	97	63	<0.0001
	(0.04)	(0.05)	(0.04)
Angiotensin‐converting enzyme inhibitors or angiotensin receptor blockers	181 621 (50.6)	102 646 (49.9)	78 975 (51.7)	<0.0001
Diuretics	142 564 (35.5)	72 997 (35.5)	69 567 (45.5)	<0.0001
Niacin or fibrates	14 101 (3.9)	10 576 (5.1)	3525 (2.3)	<0.0001
Statins	148 140 (43.5)	89 477 (43.5)	58 663 (38.4)	<0.0001
Antiarrhythmic drugs, n (%)
All Class I	14 759 (4.1)	6555 (3.2)	8204 (5.4)	<0.0001
Class III (sotalol/dofetilide)	5757 (1.6)	3457 (1.7)	2300 (1.5)	<0.0001
Amiodarone	9186 (2.6)	5934 (2.9)	3252 (2.1)	<0.0001
Rate‐controlling drugs, n (%)
Metoprolol	80 532 (22.5)	44 469 (21.6)	36 063 (23.6)	<0.0001
Carvedilol	27 928 (7.8)	18 250 (8.9)	9678 (6.3)	<0.0001
Atenolol	33 071 (9.2)	16 954 (8.2%)	16 117 (10.5%)	<0.0001
		(8.2)	(10.5)
Calcium channel blockers, n (%)
Diltiazem	20 404 (5.7)	9736 (4.7)	10 668 (7.0)	<0.0001
Verapamil	7829 (2.2)	3622 (1.8)	4207 (2.8)	<0.0001

CDHP indicates consumer‐directed health plan; EPO, exclusive provider organization; HDHP, high‐deductible health plan; HMO, health maintenance organization; POS, point of service; and PPO, preferred provider organization.

Overall, women had a lower prevalence of OAC drug receipt in the 90 days following a new AF diagnosis (warfarin [38.1% women versus 41.1% men; P<0.0001] and any DOAC [11.9% women versus 14.4% men; P<0.0001]) compared with men (Table 2). When we restricted the population to anticoagulation‐eligible patients with low bleeding risk (CHA2DS2VASc score of ≥2 and HAS‐BLED score of ≤3), these differences persisted for warfarin receipt (40.4% women versus 45.7% men; P<0.0001) and DOAC receipt (13.0% women versus 14.5% men; P<0.0001). When restricting even further to only high‐risk patients with a CHA2DS2VASc score of ≥4, we found that women were still less likely than men to receive any form of anticoagulation (49.0% women versus 53.0% men; P<0.0001). The odds ratio for the association between female sex with no anticoagulation was 1.24 (95% CI, 1.22–1.26; P<0.0001) unadjusted and 1.20 (95% CI, 1.18–1.22; P<0.0001) adjusted, with similar significant findings among the subgroup of anticoagulation‐eligible patients.

Table 2

Anticoagulation by Sex and CHA2DS2‐VASC and HAS‐BLED Scores

Medication(s)	All Patients (N=358 649) , n (%)	Male (N=205 756) , n (%)	Female (N=87 581) , n (%)	P Value	Anticoagulation‐Eligible Patientsa (N=226 94)9, n (%)	Male (N=107 439) , n (%)	Female (N =119 510) , n (%)	P Value
No anticoagulation	176 239 (49.1)	96 424 (46.9)	79 815 (52.2)	<0.0001	105 255 (46.4)	46 121 (42.9)	59 134 (49.5)	<0.0001
Any anticoagulation	182 410 (50.9)	109 332 (53.1)	73 078 (47.8)	<0.0001	121 694 (53.6)	61 318 (57.1)	60 376 (50.5)	<0.0001
Warfarin	142 868 (39.8)	84 637 (41.1)	58 231 (38.1)	<0.0001	97 374 (42.9)	49 130 (45.7)	48 244 (40.4)	<0.0001
Direct oral anticoagulants	49 193 (13.7)	30 318 (14.8)	18 875 (12.4)	<0.0001	31 034 (13.9)	15 545 (14.5)	15 489 (13)	<0.0001
Dabigatran	22 057 (6.2)	14 017 (6.8)	8040 (5.3)	<0.0001	13 775 (6.1)	7163 (6.7)	6612 (5.5)	<0.0001
Rivaroxaban	20 587 (5.7)	12 466 (6.1)	8121 (5.3)	<0.0001	12 946 (5.9)	6301 (5.9)	6645 (5.6)	0.0018
Apixaban	6549 (1.8)	3835 (1.9)	2714 (1.8)	0.05	4313 (1.9)	2081 (1.9)	2232 (1.9)	0.23

Anticoagulation eligible defined as CHA2DS2Vasc Score≥2 and HAS‐BLED score ≤3.

Women, compared with men, experienced higher risk of ischemic stroke and all‐cause hospitalization (Table 3). After adjustment for age, Charlson Comorbidity Index score, heart failure, hypertension, diabetes mellitus, geographic region, insurance plan, and receipt of concomitant drug therapies, female sex remained associated with a higher risk of stroke (hazard ratio [HR], 1.27; 95% CI, 1.21–1.32; P<0.0001) and all‐cause hospitalization (HR, 1.06; 95% CI, 1.05–1.07; P<0.0001) (Table 3). There was a modest decreased association of ICH after adjustment (HR, 0.91; 95% CI, 0.83–0.99; P=0.03). Kaplan–Meier curves for ischemic stroke demonstrate the lower survival curve for women compared with men, shown in Figure S1.

Table 3

Primary Outcomes (N=358 649)

Outcomes	Sex	Patients, N	Events, N (%)	Unadjusted Incidence Rate (per 1000 person‐years)	Unadjusted Hazard Ratioa (95% CI)	P Value	Adjusted Hazard Ratioa , b (95% CI)	P Value
All‐cause hospitalization	Female	152 893	93 068 (60.9)	344.7 (342.5–346.9)	1.14 (1.13–1.15)	<0.001	1.06 (1.05–1.07)	<0.0001
	Male	205 756	115 558 (56.2)	297.9 (296.2–299.6)
Stroke	Female	152 893	5114 (3.3)	10.9 (10.6–11.2)	1.52 (1.46–1.59)	<0.0001	1.27 (1.21–1.32)	<0.0001
	Male	205 756	4574 (2.2)	7.2 (7.0–7.4)
ICH	Female	152 893	921 (0.6)	1.9 (1.8–2.1)	1.04 (0.95–1.13)	0.39	0.91 (0.83–0.99)	0.03
	Male	205 756	1189 (0.6)	1.8 (1.7–2.0)

ICH indicates intracranial hemorrhage.

Reference group is male.

Adjusted for age, Charlson Comorbidity Index score, congestive heart failure, hypertension, diabetes mellitus, region, insurance plan, and receipt of concomitant drug therapies (antiplatelet agent, angiotensin‐converting enzyme/angiotensin receptor blocker, stain, niacin/fibrate).

For our mediation analysis limited to those with CHA2DS2VASc score of ≥2, there was evidence of partial mediation by anticoagulant drug receipt for ischemic stroke (indirect effect, 0.089; 95% CI, 0.86–0.93; P<0.0001) and all‐cause hospitalization (indirect effect, 1.021; 95% CI, 1.004–1.037; P=0.014) (Figure 2). Anticoagulation also partially mediated ICH (indirect effect, 1.41; 95% CI, 1.28–1.55; P<0.0001) (Figure 2), although the absolute risk was small and comparable with men.

Figure 2

Mediation of sex differences in outcomes by OAC use.

OAC use partially mediates gender differences in all‐cause hospitalization, stroke, and ICH for the cohort with CHA2DS2VASc score ≥2. ICH indicates intracerebral hemorrhage; and OAC, oral anticoagulant.

Discussion

In summary, we found that in a contemporary cohort of US patients with commercial health insurance, women are less likely to be prescribed oral anticoagulation, especially a DOAC. Women with AF also experience higher risk of ischemic stroke and all‐cause hospitalization, yet lower risk of ICH compared with men. Importantly, anticoagulation differences in men and women statistically mediate the observed differences in stroke and all‐cause hospitalization but do not explain the differences completely. Not surprisingly, OAC also mediates the risk of ICH in women.

There are a variety of reasons that may explain why women are less likely to be prescribed oral anticoagulation. First, despite recent evidence suggesting an overall higher risk profile of women for stroke in AF and updated guidelines with integration of sex into the CHA2DS2‐VASc score, clinicians may subscribe to more recent registry data demonstrating that sex may be more of a risk modifier than a risk factor, especially at lower risk scores.22 However, in our subanalysis limited to only high‐stroke‐risk patients, in which sex would be expected to play a more important role as a modifier, we found that the same sex differences persisted regardless of overall number of risk factors, with women consistently receiving less anticoagulation than men regardless of risk. Alternatively, clinicians may be downplaying the risk in women for similar reasons that risk for heart disease among women is not fully appreciated, resulting in lower anticoagulation prescriptions for women.23 Additionally, women may decide against anticoagulation therapy on the basis of the kind of shared decision‐making support and risk framing experienced by them. Gender concordance between patient and providers has shown to result in improved patient survival in cardiac patients,24 but 88% of cardiologists and 96% of electrophysiologists are men.25 Patient nonmedical factors, such as time and cost, may also impact ability to treat AF effectively among women.26 For example, women are more likely than men to delay care because of logistical barriers,26 which may translate into lower compliance with time‐intensive International Normalized Ratio monitoring required when taking warfarin. A better understanding of these issues may help us identify missed opportunities to close gender gaps in anticoagulation, such as through the increased prescription of DOACs, which have no laboratory monitoring requirements.

Our findings demonstrating lower anticoagulation with DOACs among women support older findings from North America (US Medicare population and a Canadian study).27, 28, 29 The more recent PINNACLE (https://cvquality.acc.org/NCDR-home/registries/outpatient-registries/pinnacle-registry) Registry of US ambulatory encounters also demonstrated lower DOAC use among women.30 Our findings importantly expand on these results by evaluating outcomes by sex and mediation of those outcomes by anticoagulation. The reasons for lower anticoagulation with DOACs, relative to warfarin, among women in the United States is still unclear, especially since there are data specifically supporting the advantages of DOACs versus warfarin in women.31 The seminal DOAC trials enrolled 35% to 40% women and did not show any treatment heterogeneity in men versus women. Still, women might be perceived in some situations to be more frail and have lower body mass and, as such, may be prescribed lower doses of the same medications, sacrificing efficacy.29

Along with worse stroke outcomes, women were more likely to experience hospitalization after a new diagnosis AF compared with men. Reasons for this are likely multifactorial, though prior data showing delays to referral for catheter ablation among women with AF32 may be a marker of the overall delays to all forms of appropriate treatment for AF among women.33 Worse outcomes coupled with the lower oral anticoagulation found in our study suggest that the high risk among women is not mirrored by appropriate clinical responses to mitigate this increased risk.

However, we also found that women were less likely to suffer ICH, in contrast to their higher bleeding risk when receiving other cardiovascular treatments.34 While a number of risk factors for ICH have been established, such as hypertension35 and older age,36 our data actually showed a lower prevalence of many of these risk factors among women compared with men. The implication is that ICH without OAC is less common in women, compared with men, at least among those with AF. The mediation of these offsetting risks (ischemic stroke, hospitalization versus ICH) by OAC in women suggests a nuanced influence of anticoagulation for both benefit and harm among women.

There important limitations to this study. Our study focuses on receipt of OAC drugs but does not ascertain actual pill consumption or adherence. While we could not exclude patients with all possible contraindications to anticoagulation, we did perform a subanalysis of patients categorized as “anticoagulation eligible.” We also were unable to capture to what degree differences in anticoagulation may be attributable to differences in patient preferences by sex. Given that the CHA2DS2‐VASc score did not become formally incorporated into US guidelines until 2014, it is possible that our cohort through 2015 only started to capture the response to guideline changes in the form of increased oral anticoagulation among women. Furthermore, the recent American Heart Association/American College of Cardiology/Heart Rhythm Society focused update to the guidelines in 20197 downgraded the class of recommendation for oral anticoagulation among women with CHA2DS2‐Vasc scores of 2 from I to IIb; however, our cohort would not capture these recent changes. Because we used ICD‐9 codes for outcomes, we are dependent on the reliability of these selected codes, although high‐specificity algorithms were used for outcomes. Because of limitations of the data source, we were not able to evaluate all‐cause mortality. Finally, while we performed extensive adjustment for comorbidities and other demographics, confounding from unadjusted covariates may persist.

Conclusions

In patients with newly diagnosed of AF, receipt of oral anticoagulation, including DOACs, is substantially lower among women. Women, compared with men, had a high risk of ischemic stroke and hospitalization, and oral anticoagulation partially mediated the observed risk difference. OAC use also partially mediated ICH, although the absolute risk was small and comparable with men.

Sources of Funding

This work was supported by a grant from the Stanford Women & Sex Differences in Medicine. Dr Yong is supported by an American Heart Association Mentored Clinical and Population Research Award and a Veterans Affairs HSR&D Career Development Award.

Disclosures

Dr Laasberg is a consultant for Genentech, Biogen, Moleac, NuvOx Pharma and Nektar therapeutics. Dr Turakhia reports research support from Janssen Inc; research support and consultant fees/honoraria from Medtronic Inc; research support from AstraZeneca; research support from Veterans Health Administration; other from AliveCor; consultant fees/honoraria from St Jude Medical; research support and consultant fees/honoraria from Boehringer Ingelheim; consultant fees/honoraria from Precision Health Economics; other from Zipline Medical; consultant fees/honoraria and other from iBeat Inc; consultant fees/honoraria from Akebia; research support and consultant fees/honoraria from Cardiva Medical; consultant fees/honoraria from Medscape/theheart.org; non‐financial research support from Amazon; other from iRhythm; and outside the submitted work. Dr Turakhia is an editor for JAMA Cardiology. The remaining authors have no disclosures to report.

Figure S1

Click here for additional data file.