Background: Atrial fibrillation (AF) is the most common arrhythmia in clinical practice. The majority of adults with AF are symptomatic, and symptoms are major determinants of quality-of-life. We proposed a theoretical model of symptom perception that involves both symptom detection and symptom interpretation. In order to better understand AF symptom perception, the aim of this body of work was to identify AF-specific symptom clusters, characterize individuals within clusters based on sociodemographic and clinical variables, and determine whether symptom cluster membership was associated with healthcare utilization (AF-related emergency department visits and hospitalizations). Methods/Results: Data sets from the Standard versus Atrial Fibrillation spEcific managemenT strategY (SAFETY) Trial (n=355) and Vanderbilt Atrial Fibrillation Registry (VAFR, n=1,501) were used to conduct cross-sectional secondary data analyses of adults with clinically verified AF. Symptom clusters were identified using self-reported symptoms and two statistical approaches: hierarchical cluster analysis and latent class analysis. Regression analyses were performed with VAFR to determine associations with healthcare utilization. Three symptom clusters were found using cluster analysis and SAFETY participants, 2 symptom clusters using cluster analysis and VAFR participants, and 4 symptom clusters using latent class analysis and VAFR participants. Symptom cluster membership was associated with gender, age, AF type, BMI, heart failure, coronary artery disease, current use of anti-arrhythmic medication, and history of ablation. Although the clusters differed between studies, when the results from the different studies were compared the results were complimentary. The symptom clusters found with VAFR were associated with an increased rate of AF-related emergency department visits and hospitalizations, either when compared to all individuals without that specific cluster (hierarchical cluster analysis), or when compared to an Asymptomatic cluster of patients (latent class analysis). Conclusions: Clinically meaningful symptom clusters were identified that were associated with increased rates of healthcare utilization. Both modifiable and non-modifiable sociodemographic and clinical characteristics are associated with cluster membership.