Disparities and Temporal Trends in the Use of Anticoagulation in Patients with Ischemic Stroke and Atrial Fibrillation

Abstract

Background and Purpose

Ischemic stroke (IS) secondary to atrial fibrillation (AF) is largely preventable with the use of anticoagulation. We sought to identify race-ethnic and sex disparities with the use of direct oral anticoagulants (DOACs), aspirin, and warfarin in IS patients with AF, and to identify temporal trends in the utilization of these medications.

Methods:

The FLorida PuErto Rico Atrial Fibrillation (FLiPER-AF) Stroke Study included 24,040 IS cases enrolled in the Florida-Puerto Rico Collaboration to Reduce Stroke Registry (CReSD) from 2010 to 2016. Multivariable logistic regression models were performed to evaluate the effect of race-ethnicity and sex on utilization of DOACs, aspirin and warfarin for stroke prevention in AF after adjustment for sociodemographic, hospital and clinical factors.

Results:

Among 24,040 IS cases, 54% were women and 10% Black, 12% FL-Hispanics, 4% PR-Hispanic and 74% Whites. From 2010 to 2016, DOAC use increased from 0% to 36%, warfarin use decreased from 51% to 17% and aspirin use remained relatively stable (42% to 40%). After adjustment, Blacks had higher odds of warfarin (OR=1.22, 95% CI 1.07–1.40) prescription at discharge compared to Whites. Men had higher rates of aspirin (42.1% vs 38.8%), warfarin (33.6% vs. 28.9%) and DOAC (21.3% vs. 19.3%) use compared to women. After adjustment, women had lower odds of being discharged on aspirin (OR=0.92, 95% CI 0.86–0.98) or warfarin (OR=0.91, 95% CI 0.84–0.99). There was no sex difference in use of DOACs.

Conclusions:

Our study confirmed the increasing use of DOACs, down-trending use of warfarin, while aspirin use remained similar over the years. There are gender and race-ethnic disparities in anticoagulation use in IS patients with AF. It is critical to understand underlying drivers of these disparities in order to develop better practice strategies for stroke prevention in patients with AF.

Clinical Trial Registration:

URL: https://www.clinicaltrials.gov/ct2/show/NCT03627806. Unique identifier: NCT03627806.

Introduction

Atrial fibrillation (AF) is the most common sustained cardiac arrhythmia.¹ AF is an independent risk factor for stroke, accounting for 15% of all ischemic strokes.^{1, 2} Patients with AF and ischemic stroke have higher mortality, stroke recurrence rate, stroke-related complications and greater disability at time of stroke discharge compared to stroke patients without AF.^3–5

Anticoagulation with warfarin has long been the mainstay of treatment for stroke prevention in patients with AF. Nevertheless, warfarin has been underutilized in these patients.^6–9 Recently, the FDA has approved four direct oral anticoagulants (DOACs) for the treatment of AF: dabigatran, rivaroxaban, apixaban and edoxaban. DOACs have been shown to be at least as effective as warfarin in the prevention of stroke in patients with non-valvular AF and may even carry a lower risk of bleeding.^10–12 Additionally, DOACs introduction into routine practice has been associated with improved rates of overall anticoagulant use for patients with AF.¹³

Previous population based studies have shown that the use of dabigatran and rivaroxaban was on the rise from 2010 to 2012, after FDA approval; however, this was prior to the release of apixaban.¹⁴ Furthermore, sex disparities have been identified in existing population based studies in the treatment of AF, with some variability in the anticoagulant utilization patterns for women.^15–19 Less is known about the effect of race-ethnicity on the utilization of DOACs. In our study we seek to identify utilization patterns of aspirin, warfarin and DOACs in hospitalized patients with ischemic stroke and AF over a longer time period and inclusive of dabigatran, rivaroxaban and apixaban. Additionally, we seek to identify race-ethnic and sex disparities in the treatment of AF in patients hospitalized for ischemic stroke.

Methods

The Florida Puerto Rico Atrial Fibrillation Stroke Study (FLiPER–AF) is aimed to identify new data on disparities in stroke outcomes for patients with AF and evaluate the clinical practice for AF detection and treatment in stroke patients in the large stroke hospital systems of care that is representative of stroke practice and diverse patient populations in Florida and Puerto Rico. It utilizes the stroke registry data of the Florida-Puerto Rico Collaboration to Reduce Stroke Registry (FL-PR CReSD) collected from 86 FL-PR CReSD hospitals (75 in Florida and 11 in Puerto Rico) from January 2010 to December 2016. A non-identified dataset was used and approval from the international review board was not required. The data that support the findings of this study are available from the corresponding author upon reasonable request.

The registry includes patients with the primary diagnosis of ischemic stroke, transient ischemic attack (TIA), subarachnoid hemorrhage, intracerebral hemorrhage, and stroke not otherwise specified. Data were collected using the American Heart Association (AHA) Get With The Guidelines–Stroke (GWTG-S), a voluntary, national hospital program established with the goal of improving quality of stroke care. In addition, new data collection elements were added for those hospitals participating in FL-PR CReSD and FLiPER–AF, including questions on ethnicity, language, education, and AF diagnosis, use of a prolonged cardiac monitoring and management.

A total of 147,842 cases over age 18 years of age were enrolled in FL-PR CReSD. Of these, 104,308 cases with primary diagnosis of ischemic stroke and data on AF status were included in the FLiPER–AF Stroke Study. Cases with the primary diagnosis of intracerebral hemorrhage (12%), subarachnoid hemorrhage (5%), TIA (1%), and stroke not otherwise specified (1%) were excluded from the FLiPER–AF Stroke Study.

Trained personnel at participating hospitals used GWTG-S data-collection tools to collect information on patients presenting to the hospitals with stroke symptoms. Data were collected using AHA’s Patient Management Tool, an online, interactive assessment and reporting system. Information collected for each hospitalization included patient demographics (age, sex and race/ethnicity- non-Hispanic white, non-Hispanic black, or Hispanic), health insurance status (private, Medicare, Medicaid/no insurance or unknown), medical history (current smoker, hypertension, diabetes, dyslipidemia, coronary artery disease, peripheral vascular disease, previous stroke/TIA, heart failure and renal insufficiency), serum creatinine at admission, mode of hospital arrival (via emergency medical services-EMS from home/scene, private transport, or unknown), stroke severity at presentation assessed by the NIH Stroke Scale (NIHSS), modified Rankin Score (mRS) at discharge, in-hospital mortality and discharge disposition. High data quality of the GWTG database is maintained through careful training of chart abstractors, standardized coding instructions, limitations of data fields to realistic entries, audit trails, and required site data quality reports. Data on hospital-level characteristics (academic status, number of beds, and number of years in GWTG-S) were obtained from the AHA database in addition to a self-reported hospital characteristics survey distributed to all hospitals participating in FL-PR CReSD.²⁰

A combination of persistent or paroxysmal AF and medical history of AF/Flutter was used to define AF in the FLiPER–AF Stroke Study. The use of aspirin, warfarin and DOACs prescribed at discharge from stroke hospitalization were analyzed overall, by race-ethnicity and sex. Temporal trends in the utilization of aspirin, warfarin and DOACs from 2010 to 2016 were investigated.

Statistical Analysis

Univariate analyses were used to compare the frequencies of the pre-specified patient-level characteristics of ischemic stroke patients with AF by race-ethnicity and by sex. Continuous variables were summarized as means with standard deviation (SD) or medians with interquartile range (IQR), and categorical variables were presented as frequencies with percentages. For continuous variables, differences between race-ethnicity and sex were assessed using the ANOVA test. For categorical variables, the Pearson chi-square test was used. Additionally, univariate analyses comparing prescription of anticoagulant medication (aspirin, warfarin and DOAC) at discharge across race-ethnicity and sex were presented. To further analyze the race-ethnicity and sex differences on prescription of anticoagulant at discharge, multilevel logistic regressions with generalized estimating equations (GEE) were conducted to account for within-hospital and between-hospital variability. Multilevel logistic regression models with outcomes of each of the three anticoagulants were built to adjust for pre-specified patient-level characteristics, age, sex, race/ethnicity, insurance status, history of chronic renal insufficiency (CRI), serum creatinine at admission, the NIHSS score, and hospital-level characteristics of academic status. Model results were presented as odds ratios (OR) with 95% confidence intervals (95% CI) and p values. P values less than 0.05 were considered statistically significant. A second model including CHA₂DS₂ VASc scores was performed and did not affect the results. To further investigate the temporal trends of use of anticoagulation, histogram of temporal trends of overall use of the three anticoagulants and line chart was presented to compare temporal trends stratified by race-ethnicity and sex. All statistical analyses were performed using SAS version 9.3 software.

Results

Sample characteristics

In a sample of 24,040 patients with IS and AF the mean age was 79±11 years. The majority were women (53.8%) and had hypertension (71 %), severe stroke (NIHSS ≥ 6, 58%) and greater stroke risk (CHA₂DS₂ VASc score 6+, 61.1%) (Table 1).

Table 1.

Characteristics of ischemic stroke patients with AF by race/ethnicity.

	All		FL-White		FL-Black		FL-Hispanic		PR-Hispanic
	N	%	N	%	N	%	N	%	N	%
All	24040	100	17732	73.8	2354	9.8	3034	12.6	920	3.8
Age (mean ± SD), yrs*	79±11		80±10		72±13		78±11		77±10
Women*	12923	53.8	9440	53.2	1296	55.1	1693	55.8	494	53.7
Insurance*
Private	8744	36.4	7252	40.9	573	24.3	576	19	343	37.3
Medicare	9612	40	6785	38.3	954	40.5	1552	51.2	321	34.9
No Insurance/Medicaid	917	3.8	434	2.4	234	9.9	241	7.9	8	0.9
Unknown	4767	19.8	3261	18.4	593	25.2	665	21.9	248	27
Medical History
Smoker*	1793	7.5	1295	7.3	244	10.4	213	7	41	4.5
Hypertension*	17181	71.5	12599	71.1	1711	72.7	2053	67.7	818	88.9
Diabetes Mellitus*	6268	26.1	4048	22.8	859	36.5	919	30.3	442	48
Dyslipidemia*	10055	41.8	7896	44.5	859	36.5	1031	34	269	29.2
CAD/prior MI*	7277	30.3	5561	31.4	589	25	819	27	308	33.5
Prior Stroke/TIA*	6896	28.7	5089	28.7	726	30.8	818	27	263	28.6
Chronic renal insufficiency*	1323	5.5	968	5.5	195	8.3	136	4.5	24	2.6
Serum Creatinine, (median (IQR))*‡	15903	1.0 (0.8–1.3)	11873	1.0 (0.8– 1.3)	1394	1.1 (0.9– 1.5)	1959	1.0 (0.8– 1.3)	677	0.9 (0.7– 1.2)
NIHSS*
0 to 5	7657	42	5955	44.1	687	38.5	810	37.1	205	28.3
6+	10555	58	7565	56	1098	61.5	1372	62.9	520	71.7
CHA2DS2 VASc*†
2–3	1625	6.8	1087	6.1	274	11.6	235	7.7	29	3.2
4–5	7737	32.2	5682	32	777	33	1044	34.4	234	25.4
6+	14678	61.1	10963	61.8	1303	55.4	1755	57.8	657	71.4
Length of Stay*
0–6 days	14822	64.1	11687	68.4	1180	51.9	1634	55.9	321	39.3
>6 days	8286	35.9	5402	31.6	1096	48.2	1292	44.2	496	60.7
mRS at discharge*
0–2	5991	68.2	4266	67.2	600	69	922	70.5	203	79.9
3–5	2793	31.8	2087	32.9	270	31	385	29.5	51	20.1

AF indicates atrial fibrillation; CAD indicates coronary artery disease; MI indicates myocardial infarction; TIA indicates transient ischemic attack; IQR indicates interquartile range; NIHSS indicates National Institutes of Health Stroke Scale.

^*P<0.05

^†CHA₂DS₂ VASc score at time of discharge.

^‡Serum creatinine at time of admission.

Whites made up the greatest proportion of patients (74%) vs. 10% Black, 12% FL-Hispanic, and 4% PR-Hispanic. Whites were older (mean age 80±10 years) compared to Blacks (72±13), FL-Hispanics (78±11), or PR-Hispanics (77±10), and more likely to have private insurance (41% vs. Black 24%, FL-Hispanic 19%, PR-Hispanic 37%), and dyslipidemia (44.5% vs. Black 36.5%, FL-Hispanic 30%, and PR-Hispanic 29%). White patients had shorter hospital length of stay (LOS ≤6 days 68% vs. Black 52%, FL-Hispanic 56%, PR-Hispanic 39%) and the greatest degree of disability at discharge (mRS 3–5 64% vs Black 63%, FL-Hispanic 29.5%, PR-Hispanic 20%).

Black patients were the youngest (mean age 72±13 years) and more likely to have Medicaid or no insurance (10% vs. White 2%, FL-Hispanic 8%, PR-Hispanic 1%), a history of smoking (10% vs. White 7%, FL-Hispanic 7%, and PR-Hispanic 4.5%), a history of CRI (8% vs. 5.5% White, 4.5% FL-Hispanic and 2.6% PR-Hispanic) and prior stroke/TIA (31% vs. White 29%, FL-Hispanic 27%, PR-Hispanic 29%) compared to other race-ethnic groups.

FL-Hispanics comprised 12.6% of the patients with a mean age of 78±11years and had the highest rate of Medicare (51% vs. White 34%, Black 40.5%, PR-Hispanic 35%). FL-Hispanics had the highest proportion of patients with CHA₂DS₂ VASc of 4–5 compared to other race-ethnic groups; however, the majority of FL-Hispanics had CHA₂DS₂ VASc of 6+. PR-Hispanics made up 4% of the study population with a mean age of 77±10 years. Compared to the other groups, PR-Hispanics had the highest rate of hypertension (89% vs. White 71%, Black 73%, FL-Hispanic 68%), diabetes mellitus (48% vs. White 23%, Black 36.5%, FL-Hispanic 30%), and coronary artery disease (CAD) or prior myocardial infarction (MI) (33.5% vs. White 31%, Black 25%, Fl-Hispanic 27%). Additionally, PR-Hispanics had a greater proportion of patients with length of stay >6 days (61% vs. White 32%, Black 48%, Fl-Hispanic 44%) and the greatest rate of functional independence at discharge (mRS 0–2 80%, White 67%, Black 69%, FL-Hispanic 70.5%).

In regards to sex differences, women were older (mean age 81±10) than men (76±11), and more likely had Medicare (42.6% vs. 37%), hypertension (73.1% vs. 69.5%), prior stroke/TIA (29.3% vs. 27.9%), severe stroke (NIHSS 6+, 47.6% vs. 39.6%) and greater stroke risk (CHA₂DS₂ VASc of 6+, 73.9% vs. 46.1%) compared to men (Supplemental Material, Table I). Men were younger and more likely to have private insurance (38% vs. 35%), history of smoking (10% v.s 5.3%), diabetes (28.5% vs. 24%), dyslipidemia (43.9% vs. 40.1%), CAD/prior MI (35.8% vs. 25.6%), history of CRI (6.4% vs. 4.7%), milder stroke (NIHSS 0–5, 47.4% vs. 37.5%), and lower stroke risk (CHA₂DS₂ VASc of 2–5, 53.9% vs. 26.1%) compared to women.

Effect of race-ethnicity and sex on anticoagulation at discharge

Blacks had the highest rate of being discharged on aspirin (43.6% vs. White 42.2%, FL-Hispanic 35.5%, PR-Hispanic 11.1%) and warfarin (35.2% vs. White 31%, FL-Hispanic 28.5%, PR-Hispanic 29.8%), and the lowest rate of DOACs at discharge (18.4% vs. White 19.5%, FL-Hispanic 24.6%, PR-Hispanic 25.7%) (Table 2). PR-Hispanics had the highest rate of DOAC prescription (25.7%) and the lowest rate of aspirin prescription at discharge (11.1 %). Men had a higher likelihood of being discharged on aspirin (42.1% vs 38.8%), warfarin (33.6% vs. 28.9%) and DOACs (21.3% vs, 19.3%) compared to women.

Table 2.

Utilization of aspirin, warfarin, DOAC by race-ethnicity in patients with ischemic stroke and AF.

Use of Anticoagulation	All		FL-White		FL-Black		FL-Hispanic		PR-Hispanic
	N	%	N	%	N	%	N	%	N	%
Aspirin*	9684	40.3	7480	42.2	1026	43.6	1076	35.5	102	11.1
Warfarin*	7466	31.1	5499	31	829	35.2	864	28.5	274	29.8
DOAC*	4866	20.2	3450	19.5	434	18.4	746	24.6	236	25.7

AF indicates atrial fibrillation; DOAC indicates direct oral anticoagulant; FL indicates Florida; PR indicates Puerto Rico.

^*P<0.05

In the multivariate analysis using Whites as a reference, PR-Hispanics were less likely to be discharged on aspirin (OR=0.23, 95% CI: 0.14–0.38, P<0.0001) and more likely to be discharged on a DOAC (OR=2.5, 95% CI:1.54–4.07, P=0.0002). Blacks were more likely to be discharged on warfarin (OR=1.22, 95% CI: 1.07–1.40, P=0.004). (Table 3)

Table 3.

Effect of race-ethnicity and sex on utilization of aspirin, warfarin and DOAC in patients with IS and AF.

	FL White	FL Black	FL Hispanic	PR Hispanic	Men	Women
Aspirin
N (%)	7480 (42.2)	1026 (43.6)	1076 (35.5)	102 (11.1)	4676 (42.1)	5008 (38.8)
AOR (95% CI)	Ref	1.05 (0.93, 1.18)	0.91 (0.80, 1.03)	0.23 (0.14, 0.38)*	Ref	0.92 (0.86, 0.98)*
Warfarin
N (%)	5499 (31)	829 (35.2)	864 (28.5)	274 (29.8)	3730 (33.6)	3736 (28.9)
AOR (95% CI)	Ref	1.22 (1.07, 1.40)*	1.02 (0.89, 1.18)	0.75 (0.35, 1.62)	Ref	0.91 (0.84, 0.99)*
DOAC
N (%)	3450 (19.5)	434 (18.4)	746 (24.6)	236 (25.7)	2372 (21.3)	2494(19.3)
AOR (95% CI)	Ref	0.84 (0.63, 1.11)	0.98 (0.72, 1.35)	2.50 (1.54, 4.07) *	Ref	0.96 (0.86, 1.07)

Adjusted for age, sex, race-ethnicity, insurance status, academic hospitals, NIHSS, medical history of CRI, length of stay and serum creatinine.

AF indicates atrial fibrillation; DOAC indicates direct oral anticoagulant; FL indicates Florida; PR indicates Puerto Rico; AOR indicates adjusted odds ratio.

Compared to men in adjusted models, women had lower odds of being discharged on aspirin (OR=0.92, 95% CI: 0.86–0.98, P=0.007) and warfarin (OR=0.91, 95% CI: 0.84–0.99, P=0.03) (Table 3); however, there was no difference in DOAC prescription at discharge between women and men (OR=0.96, 95% CI: 0.86–1.07, P=0.45).

Length of hospital stay and antithrombotic at discharge

Overall, the majority of Florida subjects had LOS ≤6 days (65.1%), whereas the majority of Puerto Rico patients had LOS >6 days (60.7%). (Supplemental Material, Table II) Patients discharged on aspirin and warfarin were more likely to have shorter LOS in Florida and more prolonged LOS in Puerto Rico. Patients discharged on a DOAC were more likely to have shorter LOS in both Florida and Puerto Rico.

Temporal trend in the utilization of aspirin, warfarin and DOAC at discharge

From 2010 – 2016, the overall rate of aspirin use remained stable (41.7% in 2010 and 40.2% in 2016). In contrast, the overall rate of warfarin use steadily declined from 2010 to 2016 (51.5% vs. 17.3%) while the rate of DOAC use steadily increased (0% vs. 36.2%) (Figure 1). These overall trends were observed across all race-ethnic groups (Figure 2) and in both men and women (Figure 3) with similar slopes over time. The overall proportion of patients with ischemic stroke and AF on anticoagulation remained stable, 51.5% in 2010 and 53.5% in 2016. In both Florida and Puerto Rico, “risk for bleeding” was the most commonly listed contraindication to anticoagulation (Supplemental Material, Table III and IV). In Florida, the second most common contraindication listed was “terminal illness/comfort measures only,” whereas in Puerto Rico “risk for falls” was the second most common contraindication.

Figure 1.

Temporal trend in the use of aspirin, warfarin and DOACs among ischemic stroke with AF (N=24,040) from 2010–2016. DOAC indicates direct oral anticoagulant; AF indicates atrial fibrillation.

Figure 2.

Temporal trend in the use of A) warfarin, B) DOACs and C) aspirin in ischemic stroke patients with AF from 2010–2016 by race-ethnicity. DOAC indicates direct oral anticoagulant; AF indicates atrial fibrillation.

Figure 3.

Temporal trend in the use of A) warfarin, B) DOAC and C) aspirin in ischemic stroke patients with AF from 2010–2016 by sex. DOAC indicates direct oral anticoagulant; AF indicates atrial fibrillation.

Discussion

In the present study consisting of a large real-world sample of 24,040 hospitalized ischemic stroke patients with AF, we examined temporal trends in the utilization of aspirin, warfarin and DOACs since the introduction of DOACs into clinical practice. Moreover, we identified race-ethnic and gender disparities in the use of these medications, with significantly greater use of warfarin in Blacks and underutilization of anticoagulation in women. The uptake of DOACs into clinical practice has been established previously; however, our study analyzed prescription patterns over a longer time period when multiple DOACs were on the market, and with a specific focus on high-risk patients hospitalized with ischemic stroke.^{13, 14, 21–26}

Despite the easier use of DOACs and other evidence suggesting increased utilization rates of DOACS, we found only a 2% increase of anticoagulation during the study period. This confirms the large gap between evidence-based guidelines and clinical practice, particularly in patients with acute ischemic stroke. Prior population based studies focused primarily on outpatients have shown variable results on the effect of DOACs on overall anticoagulation rates.^{13, 14, 21–26} Furthermore, the use of DOACs has largely been directed at healthier, younger patients with lower stroke risk and better functional status.^{13, 14, 27} GWTG-S data from 2010–2012 showed that hospitalized patients with acute stroke due to AF were more likely to be discharged on a DOAC (dabigatran or rivaroxaban) if they had lower stroke risk.¹⁴ The SAMURAI-NVAF study on Japanese patients admitted with acute stroke due to AF from 2011–2014 demonstrated similar results in Japan.²⁷ In both studies, the overall rate of anticoagulation remained the same during the study period.

Similarly, our study focused on patients hospitalized with IS who were sicker and at greater risk of developing intracranial hemorrhage. To minimize risk, practitioners may have delayed initiation of anticoagulation, therefore, the proportion of patients on anticoagulation may be higher post-discharge and not captured in our study. This is supported by the overall stable use of aspirin at discharge, which would have been the alternative to treatment. Patients discharged on aspirin had more prolonged LOS, again reinforcing the hypothesis that these patients had more severe strokes. Moreover, risk for bleeding was the most commonly listed contraindication for anticoagulation in this cohort. Other patient characteristics, such as ambulatory status and discharge destination, may have posed additional influence on decision-making for anticoagulation, specifically DOAC prescription.¹⁴

In regards to racial disparities, our study showed that Blacks had the highest rate of aspirin and warfarin prescription at time of discharge compared to other racial groups. Even after adjusting for age, insurance status, stroke severity, stroke risk, renal function and hospital academic status, Blacks had a 22% greater chance of being prescribed warfarin compared to Whites at time of discharge. Although comorbid conditions, such as renal failure that may be more prevalent in African Americans²⁸,may influence the utilization of DOACs, we did not find a significant difference in DOAC utilization between Whites and Blacks after adjusting for renal disease. Furthermore, about half of the Black patient cohort had LOS ≤6 days and the majority of Black patients were functionally independent at hospital discharge, suggesting possibly milder stroke in this group. Therefore, the reasons for these disparities are not entirely clear. Financial access, even for insured patients, may have influenced prescription patterns due to costly copays and difficulty obtaining preapprovals.¹⁴ Although less likely, switching from DOAC to warfarin due to “medication failure” at time of the index stroke may have contributed to this difference in our study. Additionally, the presence of these disparities in treatment after adjusting for important clinical and sociodemographic factors emphasizes the possible influence of unconscious bias or lack of awareness of existing treatment disparities in practitioner decision-making.

Interestingly, PR-Hispanics had significantly lower aspirin and greater DOAC utilization at hospital discharge. PR-Hispanics also had more prolonged length of stay in the hospital and greater independence at discharge, which are likely the main drivers of the higher DOAC rates in this group. Since PR-Hispanics were further out from their stroke and more functionally independent at discharge, they may have been deemed safer candidates for DOAC use.

In regards to sex differences, women were older and more likely to have Medicare, greater stroke risk and more severe strokes. Despite this increased risk of stroke, women had lower prescription rates of aspirin, warfarin and DOAC at discharge compared to men. However, after adjustment for important clinical and systems of care factors, women had a 8% lower chance of being discharged on aspirin, 9% lower chance of being discharged on warfarin and no statistically significant difference in DOAC prescription compared to men. Existing data on patients with AF at high risk of ischemic stroke have shown variable results with respect to the sex disparity in anticoagulation utilization.²⁹ The GARFEILD-AF registry, a global registry of patients with AF at high risk of IS, showed no difference in rate of anticoagulation between men and women.¹⁷ In contrast, the PINNACLE registry showed that women with AF at high risk for ischemic stroke were more likely to receive aspirin instead of anticoagulation as compared to men after multivariable adjustments.¹⁶ A Danish registry of anticoagulant naïve patients reported that men were less likely to receive anticoagulation than women.¹⁹ The SAMURAI-NVAF study showed that stroke patients discharged on DOAC were younger and included more men.²⁷ Our observations are most consistent with previous GWTG-S results in which there was no difference in DOAC prescription between men and women at discharge from hospitalization for ischemic stroke.¹⁴ The gender differences we observed with warfarin specifically may have been the result of mitigating the increased risk of bleeding with warfarin, since women in this cohort had greater stroke severity and thus higher risk of intracranial hemorrhage. Nevertheless, despite the lower risk of intracranial hemorrhage with DOACs compared to warfarin, we did not observe a higher rate of DOAC use in women.

Our study had several limitations. First, our analysis was performed specifically on patients hospitalized with ischemic stroke secondary to AF and is not representative of practice patterns in all AF patients. Moreover, we were not able to assess follow up prescription patterns after hospital discharge, thus the actual rate of anticoagulation may be underestimated and the usage patterns of the various anticoagulants may differ. Second, missing data elements in the registry made it difficult to comment on the influence of several variables on decision-making for anticoagulation, such as prior anticoagulation status and quantitative measure of renal disease at discharge. Third, the scope of our registry did not account for other factors related to practitioner decision making, such as patient preference, or practitioner reasoning in medication selection for secondary prevention.

In conclusion, among patients hospitalized with ischemic stroke and AF, there has been a consistent increase in DOAC utilization coinciding with a decline in subsequent warfarin use. The uptake of DOACs did not influence the overall rate of anticoagulation in this population. Additionally, Black patients were more likely to be discharged on warfarin compared to Whites, PR-Hispanics were more likely to be discharged on a DOAC and women were less likely to be on anticoagulation overall at discharge from hospitalization for ischemic stroke. These findings stress the importance of further research to understand the drivers of these disparities in order to develop and implement educational and systems-based practice strategies for better secondary prevention of stroke in patients with AF.