Abstract Introduction While sleep disorders are implicated in atrial fibrillation (AF), the interplay and overlap of these disorders in AF risk remains unclear and a risk stratification challenge. Sleep-based clusters associated with AF can account for this complexity and translate to actionable approaches to identify at-risk patients. We hypothesized discrete phenotypes of symptoms and polysomnography (PSG)-based data differ in relation to incident AF. Methods Cleveland Clinic patients (age>18) who underwent PSG 11/27/2004-12/30/2015 were retrospectively examined. Clusters were identified using latent class analysis of 23 symptoms (e.g. related to sleep disordered breathing (SDB), sleepiness, REM intrusion, NREM parasomnias), Epworth Sleepiness Scale (ESS) score, and 24 measures of SDB and sleep architecture from PSG reports. Outcome was 5-year incident AF. Cox-proportional hazards models were adjusted for age, sex, race, body mass index, cardiovascular disease and risk factors, tobacco use, chronic obstructive pulmonary disease, anti-arrhythmic medication use, and positive airway pressure use. Results The sample included 43,433 patients: age 51.8±14.5, 51.9% (n=22,548) male, 74.4% (n=32,313) White, 7.3% (n=3,151) with baseline AF, and 4.3% (n=1,875) developed 5-year incident AF. Five clusters were identified and ranked by strength of AF association: 1) Hypoxemic (n=3,245): highest %time SaO2< 90% (T90), 2) Apneas+Arousals (n=4,592): most witnessed apneas, highest apnea hypopnea index (AHI), highest arousal index, least hypopneas, 3) Short sleep+NREM (n=6,126): shortest sleep time, longest REM latency, lowest %REM, 4) Hypopneas (n=2,661): most hypopneas, 5) Long sleep+REM (n=26,809): longest sleep time, shortest REM latency, highest %REM. Compared to ‘Long sleep+REM’, ‘Hypoxemic’ had 47% higher AF risk (HR=1.47,95%CI=1.27-1.69), and ‘Hypopneas’ did not differ (HR=1.05,95%CI=0.86-1.28). Conclusion Of five clusters identified, the ‘Hypoxemic’ subtype conferred the strongest AF risk with the highest degree of hypoxemia (highest T90, lowest minimum and mean SaO2), maximum end-tidal CO2, heart rate, and ESS score. Consistent with prior evidence of hypoxemia as an AF driver and cardiovascular risk of the sleepy phenotype, this constellation of symptoms and physiologic alterations illustrates risk in the clinical setting, providing potential value as a risk prediction tool. Future investigation should focus on external validation of these findings. Support (if any) Cleveland Clinic Neurological Institute Center for Outcomes Research & Evaluation Pilot Grant, Transformative Research Resource Development Award