Skip to main content
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2019 Oct 1.
Published in final edited form as: Stroke. 2018 Oct;49(10):2529–2531. doi: 10.1161/STROKEAHA.118.022826

Incidence of Atrial Fibrillation in Patients with Recent Ischemic Stroke versus Matched Controls

Jens Witsch 1, Alexander E Merkler 2, Monica L Chen 2, Babak B Navi 2, Kevin N Sheth 1, Ben Freedman 3, Lee H Schwamm 4, Hooman Kamel 2
PMCID: PMC6205716  NIHMSID: NIHMS1503408  PMID: 30355110

Abstract

Background and Purpose:

It is unclear whether atrial fibrillation/flutter (AF) newly diagnosed after ischemic stroke represents a pre-existing risk factor that led to stroke, an arrhythmia triggered by post-stroke autonomic dysfunction, or an incidental finding.

Methods:

We compared AF incidence after hospitalizations for ischemic stroke, hemorrhagic stroke, and non-stroke conditions using inpatient and outpatient claims between 2008–2015 from a nationally representative 5% sample of Medicare beneficiaries. We used validated ICD-9-CM codes to identify AF-free patients hospitalized with ischemic or hemorrhagic stroke, and matched them in a 1:1 ratio by age, sex, race, calendar year, vascular risk factors, and Charlson comorbidities. We then matched the combined stroke cohort in a 1:1 ratio to patients hospitalized for non-stroke diagnoses. We used survival statistics and Cox regression to compare post-discharge AF incidence among groups.

Results:

We matched 2,580 patients with ischemic stroke, 2,580 with hemorrhagic stroke, and 5,160 patients with other conditions. The annual post-discharge AF incidence was 3.4% (95% confidence interval [CI], 3.1–3.7%) after ischemic stroke, 2.2% (95% CI, 1.9–2.4%) after hemorrhagic stroke, and 2.9% (95% CI, 2.6–3.1%) after non-stroke hospitalization. Ischemic stroke was associated with a somewhat higher risk of AF than hemorrhagic stroke (hazard ratio [HR], 1.5; 95% CI, 1.3–1.8) or non-stroke conditions (HR, 1.2; 95% CI, 1.1–1.3). The latter association attenuated in sensitivity analyses limiting the outcome to AF diagnoses made by cardiologists (HR, 1.1; 95% CI, 0.8–1.5) or limiting the outcome to a minimum of two AF claims on separate dates (HR, 1.2; 95% CI, 1.0–1.5; P = 0.09).

Conclusions:

New diagnoses of AF were more common after hospitalization for ischemic stroke than after hospitalization for hemorrhagic stroke or non-stroke conditions, but all hospitalized patients had a substantial incidence of new AF diagnoses after discharge and differences were attenuated when using more stringent definitions.

Keywords: Arrhythmias, ECG, Ischemic Stroke, Atrial Fibrillation, Etiology, Monitoring, Thromboembolism, Stroke


Atrial fibrillation/flutter (AF) is often newly detected after ischemic stroke.1,2 This may mean that AF preceded and caused the stroke, or may be explained by shared risk factors3 or stroke-induced disruption of autonomic pathways (Supplemental Figure I). There are scant data on rates of AF after ischemic stroke in comparison with other conditions, particularly hemorrhagic stroke, which may also increase the risk of neurogenic AF.4 We used claims data from a large sample of elderly U.S. patients to compare new AF diagnoses after hospitalization for ischemic stroke versus hemorrhagic stroke or non-stroke conditions.

Methods

Design and Population

We performed a retrospective matched cohort study using inpatient and outpatient claims between 2008–2015 from a nationally representative 5% sample of Medicare beneficiaries (Supplemental Appendix). The Weill Cornell Medical College institutional review board approved our study and waived the need for patient consent. The data cannot be shared by the authors because of a data use agreement but are publicly available via application to CMS.

Measurements and Patient Matching

We created a 1:1:2 matched cohort of patients hospitalized for ischemic stroke, hemorrhagic stroke, or non-stroke conditions. First, we used 2008 as a run-in period to exclude patients with diagnoses of ischemic or hemorrhagic stroke, defined as intracerebral or subarachnoid hemorrhage. Strokes were ascertained using a validated ICD-9-CM diagnosis code algorithm. Second, we excluded hospitalizations if a diagnosis of AF had been made before or during that hospitalization. We defined AF as any inpatient or outpatient claim with ICD-9-CM codes 427.3x in either primary or secondary diagnosis code positions, as typically done in most studies; these codes have been validated to have a positive predictive value of 89%.5 Third, using 1:1 matching without replacement, we matched patients hospitalized for ischemic stroke to patients hospitalized for hemorrhagic stroke. We matched by age, sex, race, calendar year, Charlson comorbidities, and vascular risk factors (Supplemental Appendix). Fourth, we matched the overall group of stroke patients to patients hospitalized for non-stroke conditions. Our final cohort thus consisted of patients hospitalized with ischemic stroke, hemorrhagic stroke, or non-stroke conditions matched together in a 1:1:2 ratio. We then ascertained new inpatient or outpatient diagnoses of AF after discharge from the index hospitalization.

Statistical Analysis

Survival statistics were used to calculate incidence rates of newly diagnosed AF. Cox proportional hazards models were used to determine associations between stroke status and post-discharge AF.

Sensitivity Analyses

First, we censored patients on the day before the start of any post-discharge heart-rhythm monitoring. Second, we made our outcome definition more stringent by requiring an AF diagnosis on ≥2 separate dates after discharge. Third, we included only AF diagnoses documented by a cardiologist. Fourth, we created a rough proxy for cryptogenic stroke (Supplemental Appendix).

Results

We matched 2,580 patients with ischemic stroke, 2,580 with hemorrhagic stroke, and 5,160 patients with other conditions (Table 1; Supplemental Appendix).

Table 1.

Baseline Characteristics of Patients, Stratified by Stroke Status.

Characteristica Ischemic
Stroke
(N = 2,580)
Hemorrhagic
Stroke
(N = 2,580)
No
Stroke
(N = 5,160)
Age, mean (SD), y 78.5 (7.7) 78.5 (7.6) 78.5 (7.7)
Female 1,761 (68.3) 1,761 (68.3) 3,518 (68.2)
Race
 White 2,327 (90.2) 2,327 (90.2) 4,653 (90.2)
 Black 170 (6.6) 132 (5.1) 327 (6.3)
 Other 83 (3.2) 121 (4.7) 180 (3.5)
Coronary heart disease 368 (14.3) 377 (14.6) 743 (14.4)
Hypertension 2,297 (89.0) 2,300 (89.2) 4,597 (89.1)
Diabetes 581 (22.5) 592 (23.0) 1,176 (22.8)
Congestive heart failure 65 (2.5) 66 (2.6) 131 (2.5)
Peripheral vascular disease 38 (1.5) 40 (1.6) 78 (1.5)
Chronic kidney disease 76 (3.0) 81 (3.1) 157 (3.0)
Chronic obstructive pulmonary disease 104 (4.0) 106 (4.1) 210 (4.1)
Valvular heart disease 66 (2.6) 74 (2.9) 140 (2.7)
Tobacco use 113 (4.4) 112 (4.3) 225 (4.4)
Alcohol abuse 114 (4.4) 115 (4.5) 229 (4.4)

Abbreviations: SD, standard deviation.

a

Data are presented as number (%) unless otherwise specified.

The annual post-discharge AF incidence was 3.4% (95% confidence interval [CI], 3.1–3.7%) after ischemic stroke, 2.2% (95% CI, 1.9–2.4%) after hemorrhagic stroke, and 2.9% (95% CI, 2.6–3.1%) after non-stroke hospitalization (Supplemental Figure II). After adjustment for demographics and vascular risk factors, ischemic stroke was associated with a higher risk of AF compared to hemorrhagic stroke (hazard ratio [HR], 1.5; 95% CI, 1.3–1.8) or non-stroke conditions (HR, 1.2; 95% CI, 1.1–1.3) (Table 2).

Table 2.

Associations between Stroke Status and Post-Discharge Atrial Fibrillation.

Analysis HR (95% CI)
Ischemic versus hemorrhagic stroke
 Primary analysis 1.5 (1.3–1.8)
 Censoring post-discharge heart-rhythm monitoring 1.5 (1.2–1.8)
 AF diagnosed by cardiologist 1.9 (1.2–2.9)
 ≥2 AF claims 1.5 (1.7–2.3)
 Proxy codes for cryptogenic stroke 1.7 (1.1–2.7)
Ischemic versus non-stroke controls
 Primary analysis 1.2 (1.1–1.3)
 Censoring post-discharge heart-rhythm monitoring 1.2 (1.0–1.3)
 AF diagnosed by cardiologist 1.1 (0.8–1.5)
 ≥2 AF claims 1.2 (1.0–1.5)
 Proxy codes for cryptogenic stroke 1.8 (1.3–2.6)

Abbreviations: AF, atrial fibrillation.

Discussion

In a large sample of Medicare beneficiaries, we found a higher rate of newly diagnosed AF after hospitalization for ischemic stroke versus for hemorrhagic stroke or non-stroke conditions. However, differences were not substantial, especially with more stringent endpoint definitions.

Previous work on post-stroke AF detection has been limited to patients with ischemic stroke. Such studies cannot establish a causal link between AF and the preceding stroke. Our study addresses this knowledge gap and our findings are consistent with a scenario in which an ischemic stroke occurs and the cause is only elucidated later when pre-existing paroxysmal AF is finally detected. This supports screening patients with otherwise unexplained stroke and starting anticoagulant therapy if AF is found.68 However, our findings also suggest that, in some cases, post-stroke AF may reflect background risk and may not have been directly responsible for the stroke, especially when AF is first detected years later, as often seen in our study. This argues against assuming that AF detection automatically reveals the proximate cause of stroke. However, the detection and treatment of these cases of subclinical AF may be important for future stroke prevention.9

Our study has several limitations. First, its validity depends on the accuracy of ICD-9-CM diagnosis codes. We used validated algorithms and performed sensitivity analyses using more stringent definitions, such as requiring cardiologist documentation of AF or accounting for post-discharge heart-rhythm monitoring; the administrative codes used to identify cardiologists have been validated10 while codes for post-discharge monitoring have not. We may have missed cases of AF that were not documented, especially in patients with multiple comorbidities. Second, physicians may have looked more carefully for AF or more often labeled nonspecific findings as AF in patients with a recent ischemic stroke. Patients with ischemic stroke were probably more thoroughly investigated for AF during hospitalization, which might have reduced the frequency of later outpatient detection. Third, our study may not be generalizable to younger patients. However, the burden of AF and stroke rests mostly on patients >65 years of age. Fourth, we were unable to reliably identify patients with cryptogenic stroke, and post-stroke AF detection may be higher in this subgroup than in other stroke patients, as suggested by our sensitivity analysis using a proxy for cryptogenic stroke.

Although 10–20% of patients with ischemic stroke are found to have AF,1,2 the pathogenic significance of these new AF diagnoses remains controversial. We found a higher rate of de novo AF diagnosis in patients with a recent ischemic stroke than comparable patients with hemorrhagic stroke or hospitalization for other reasons, but the magnitudes of these differences were modest. Our findings support the practice of AF screening in patients with unexplained ischemic stroke, but also indicate that such AF cases may not necessarily have been present before the stroke, so the discovery of post-stroke AF should not curtail thorough evaluation and treatment of other stroke risk factors.

Supplementary Material

Supplemental Appendix

Acknowledgments:

None.

Sources of Funding: Dr. Kamel receives funding from NIH/NINDS (grants K23NS082367, R01NS097443, and U01NS095869) and the Michael Goldberg Research Fund.

Footnotes

Disclosures: Dr. Kamel serves as an unpaid consultant to Medtronic and iRhythm. Drs. Kamel and Schwamm serve as steering committee members for the Stroke AF trial funded by Medtronic.

References

  • 1.Sanna T, Diener HC, Passman RS, Di Lazzaro V, Bernstein RA, Morillo CA, et al. Cryptogenic stroke and underlying atrial fibrillation. N Engl J Med 2014;370:2478–2486. [DOI] [PubMed] [Google Scholar]
  • 2.Gladstone DJ, Spring M, Dorian P, Panzov V, Thorpe KE, Hall J, et al. Atrial fibrillation in patients with cryptogenic stroke. N Engl J Med 2014;370:2467–2477. [DOI] [PubMed] [Google Scholar]
  • 3.Healey JS, Alings M, Ha A, Leong-Sit P, Birnie DH, de Graaf JJ, et al. Subclinical atrial fibrillation in older patients. Circulation 2017;136:1276–1283. [DOI] [PubMed] [Google Scholar]
  • 4.Samuels MA. The brain-heart connection. Circulation 2007;116:77–84. [DOI] [PubMed] [Google Scholar]
  • 5.Jensen PN, Johnson K, Floyd J, Heckbert SR, Carnahan R, Dublin S. A systematic review of validated methods for identifying atrial fibrillation using administrative data. Pharmacoepidemiol Drug Saf 2012;21 Suppl 1:141–147. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6.Freedman B, Camm J, Calkins H, Healey JS, Rosenqvist M, Wang J, et al. Screening for Atrial Fibrillation: A Report of the AF-SCREEN International Collaboration. Circulation 2017;135:1851–1867. [DOI] [PubMed] [Google Scholar]
  • 7.Mairesse GH, Moran P, Van Gelder IC, Elsner C, Rosenqvist M, Mant J, et al. Screening for atrial fibrillation: a European Heart Rhythm Association (EHRA) consensus document. Europace 2017;19:1589–1623. [DOI] [PubMed] [Google Scholar]
  • 8.January CT, Wann LS, Alpert JS, Calkins H, Cigarroa JE, Cleveland JC Jr., et al. 2014 AHA/ACC/HRS guideline for the management of patients with atrial fibrillation. Circulation 2014;130:e199–267. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9.Mahajan R, Perera T, Elliott AD, Twomey DJ, Kumar S, Munwar DA, et al. Subclinical device-detected atrial fibrillation and stroke risk: a systematic review and meta-analysis. Eur Heart J 2018;39:1407–1415. [DOI] [PubMed] [Google Scholar]
  • 10.Kamel H, Chung CD, Kone GJ, Gupta A, Morris NA, Fink ME, et al. Medical Specialties of Clinicians Providing Mechanical Thrombectomy to Patients With Acute Ischemic Stroke in the United States. JAMA Neurol 2018;75:515–517. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplemental Appendix

RESOURCES