Skip to content

Atrial Fibrillation Risk Scoring at Time of Stroke May Identify People at Low Probability of Cardioembolism

Key findings

  • Cardiologists at Massachusetts General Hospital retrospectively analyzed prospective registry data on 1,431 individuals who had an ischemic stroke, of whom 23% had a cardioembolic stroke
  • Atrial fibrillation (AF) risk at the time of stroke, whether calculated using the CHA2DS2-VASc, CHARGE-AF or Electronic Health Record–based AF (EHR-AF) scores, was significantly associated with cardioembolism
  • EHR-AF was the best-performing scoring scheme
  • In a hypothetical sample of individuals with undifferentiated stroke (that is, without diagnosed AF), the probability of cardioembolic stroke decreased substantially when sensitive thresholds of EHR-AF were used as a rule-out test

Atrial fibrillation (AF) is the predominant cause of cardioembolic stroke but may be subclinical at the time the stroke occurs. There's a critical need to discriminate cardioembolic from non-cardioembolic stroke because anticoagulation can prevent recurrent cardioembolic stroke, but has not proven effective—and may increase bleeding—in patients with unconfirmed cardioembolism.

Shaan Khurshid, MD, research fellow, Steven A. Lubitz, MD, MPH, of the Cardiovascular Research Center and Telemachus & Irene Demoulas Family Foundation Center for Cardiac Arrhythmias at Massachusetts General Hospital, and colleagues recently reported in Stroke that calculating AF risk at the time of ischemic stroke facilitates the identification of individuals with a low likelihood of cardioembolism.

Study Design

The researchers retrospectively analyzed prospectively collected registry data on 1,431 patients who had AF at the time they suffered an ischemic stroke between 2002 and 2011. All patients underwent routine assessment of ischemic stroke etiology, including continuous telemetry and electrocardiograms.

Trained physicians independent of the treating neurologists classified stroke mechanism according to the TOAST (Trial of Org 10172 in Acute Stroke Treatment) or CCS (Causative Classification of Stroke) criteria.

For each patient the researchers also calculated AF risk according to each of three scores:

  • CHA2DS2–VASc
  • Cohorts for Heart and Aging Research in Genomic Epidemiology–AF (CHARGE-AF)
  • Electronic Health Record–based AF (EHR-AF)


323 patients (23%) were judged manually to have had cardioembolic strokes. For all three AF risk scores, values were significantly higher in patients with cardioembolic stroke than those with non-cardioembolic stroke.

On regression analysis, all three AF risk scores remained associated with cardioembolism. EHR-AF performed best, classifying 71% of patients with cardioembolic stroke correctly.

The researchers report the performance of each AF risk score at selected sensitivity and specificity thresholds. For example, an EHR-AF score of 6.75 corresponded to 90% sensitivity and 32% specificity for cardioembolism with a positive likelihood ratio of 1.31 and negative likelihood ratio of 0.31.

Hypothetical Cohort

To investigate further whether EHR-AF is a useful cardioembolism classifier, the researchers assessed its performance in a hypothetical sample of individuals with undifferentiated stroke (that is, without diagnosed AF). They applied the test characteristics of the EHR-AF score at 90%, 95% and 99% sensitivity and specificity thresholds while assuming a cardioembolism prevalence of 5% to 30%.

In this cohort the probability of cardioembolic stroke decreased substantially when sensitive thresholds of EHR-AF were used as a rule-out test. The probability increased more modestly when specific thresholds were used as a rule-in test.

Applying the Findings to Practice

AF risk scores may be useful for guiding workups following an acute ischemic stroke. More intensive rhythm monitoring may be indicated for patients at sufficiently high likelihood of having an AF-related cardioembolic stroke, which may be estimated using AF risk scores. Future analyses are necessary to determine whether using AF risk scores to guide workups is a reasonable and cost-effective strategy for patients following a stroke.

Learn more about the Telemachus & Irene Demoulas Family Foundation Center for Cardiac Arrhythmias

Refer a patient to the Corrigan Minehan Heart Center

Related topics


MicroRNA miR-574-3p, which can be derived from whole blood, was associated with stroke in a community-based sample of more than 2,000 participants in the Framingham Heart Study's Offspring cohort.


Among asymptomatic patients, rates of stroke and the composite of stroke/death are better when carotid endarterectomy is performed by a vascular surgeon than by a neurologist, cardiac surgeon, thoracic surgeon or general surgeon.