Popular algorithm shown to be inaccurate in predicting atrial fibrillation

Early identification of patients at risk for atrial fibrillation, a quivering or irregular heartbeat, is critical to minimize complications such as blood clots, heart failure and stroke. But, a widely accepted atrial fibrillation risk prediction algorithm did not accurately predict incidence of the condition when it was applied to the electronic health records of a large group of patients.

That’s the finding of researchers who recently published results of their study in the journal JAMA Cardiology, which retrospectively examined the EHRs of patients who did not have atrial fibrillation but returned for follow-up care at least three times within two years.

Researchers used the EHR to follow the health of nearly 35,000 patients over five years to find out how many developed atrial fibrillation. In the process, they discovered that a popular risk prediction model—when applied to the EHR—under-predicted the incidence of atrial fibrillation among low-risk subjects while over-predicting the incidence among high-risk subjects.

Also See: Docs vastly outperform computer algorithms in diagnostic accuracy


Given the fact that more than 6 million people in the United States have atrial fibrillation, resulting in 130,000 deaths annually, the failure of the risk prediction algorithm to accurately predict the condition has significant repercussions as the number of U.S. patients is projected roughly to double by the year 2050.

The algorithm was developed by investigators on the Cohorts for Heart and Aging Research in Genomic Epidemiology (CHARGE) trial. However, Dawood Darbar, MD, chief of cardiology at the University of Illinois Hospital and Health Sciences System, warns that the AF risk prediction models should be used with caution in real-world EHR settings.

“Developing models using large cohort studies is great, but because electronic health records are so ubiquitous, there is a need to develop and validate new risk prediction algorithms that will be extremely useful for patients in primary care settings in order to identify patients at high risk for atrial fibrillation,” says Darbar. “We need to develop these models in the real world using EHRs. Stroke is the major complication associated with AF. If we can prevent even one stroke in a patient, that’s obviously a huge impact in terms of reducing the morbidities.”

He adds that risk models need to be “race-based or race-specific,” taking into account health disparities between African Americans, Hispanics and Caucasian populations. “Algorithms have to also apply to ethnic minorities,” according to Darbar, who points out that Hispanic Americans have a high risk of developing strokes.

Predictors in the CHARGE-AF model included age, race, height, weight, systolic and diastolic blood pressure, treatment for hypertension, smoking status, type 2 diabetes, heart failure, history of myocardial infarction, left ventricular hypertrophy and PR interval.

“The models performed poorly in our EMR cohort, illustrating the difficulty of applying risk models developed within prospective cohort studies to a real-world EMR context,” conclude researchers in their article. “Risk models for the development of AF or other complex disorders are unlikely to be widely used in clinical care unless they can be incorporated into EMR systems. Risk models, therefore, should be derived from and validated in different EMR cohorts, with the goal of prospectively and automatically identifying individuals at high risk for AF and implementing personalized strategies for primary prevention.”

Darbar admits that the study has some limitations, including inconsistent EHR data-entry procedures and an “indication bias” such that individuals who developed atrial fibrillation likely had more clinical encounters than those who did not.

For reprint and licensing requests for this article, click here.