Atrial fibrillation (AF) is the most common cardiac rhythm disorder affecting up to 6 million Americans, with associated healthcare costs of over $6 billion per year. AF disproportionately affects the elderly and is a major risk factor for stroke ad cognitive decline. Despite the enormous burden caused by the AF epidemic, the natural history of AF remains largely elusive due to ascertainment challenges and the use of clinical classification schemes that are impracticable in real world clinical populations. We have leveraged electronic medical record (EMR) phenotype algorithm methodology developed by the Electronic Medical Record and Genomics (eMERGE) Network to assemble a population cohort that includes 3,345 incident AF patients from Olmsted County, Minnesota from 2000-2010 (Olmsted AF Cohort). The current AF classification scheme promulgated by the American Heart Association (AHA), American College of Cardiology (ACC), and the European Society of Cardiology (ESC) relies on duration of rhythm and spontaneous conversion and patients are considered to have paroxysmal, persistent, or permanent AF. However, the accuracy and reliability of this classification scheme is limited due to the high rates of asymptomatic AF episodes thus underestimating the burden and duration of AF which results in misclassification. Ambulatory telemetry can reduce misclassification, but this is unrealistic in population studies and underscores the futility of classification schemes that rely on or incorporate assessments of AF burden. Therefore, the goal of the current application is to develop, validate, and implement an EMR based algorithm to identify AF progression in the Olmsted AF Cohort. Based on preliminary data, we propose a progression model that groups patients based on initial anti-arrhythmic therapy status and uses time to recurrence and treatment change/escalation due to arrhythmia activity as well as development of permanent AF as metrics of progression. Patients who are left in AF or whose incident event occurred with known provocateurs (e.g., cardiac surgery or thyroid disease) will be considered separately.
In aim 1, we will characterize patterns of AF progression in a population-based cohort of incident AF patients by developing and validating an EMR algorithm with dissemination of the final algorithm to the larger research community via Phenotype Knowledge Base (PheKB) (www.phekb.org).
In aim 2, we will test the hypothesis that the EMR-algorithm predicts all-cause mortality and that there are clinical risk factors associated with progression. The R21 mechanism provides the opportunity to apply novel methodologies to challenging phenotyping problems that are a major barrier in AF research. The successful development of a highly accurate AF algorithm that predicts outcomes and has identifiable risk factors will revolutionize large scale population research for this burdensome disease.
Atrial fibrillation (AF) is the most common cardiac rhythm disorder affecting up to 6 million Americans and that disproportionately affects the elderly. However, limitations of current AF classification schemes are a barrier in population research. We propose to use longitudinal data from electronic medical record systems to develop a cost-effect algorithm that will characterize AF progression.
