O066 Generalized linear model for characterisation of nocturnal arrhythmia

Abstract Background Generalised linear models (GLM) based on point processes have been previously shown helpful for characterising dynamics of sleep-disordered breathing (SDB) events using sleep stages, body position and history of SDB events, and characterising period limb movements. Episodes of non-sustained arrhythmias may occur during sleep in patients with underlying cardiac conditions, and the point process theory may help model their occurrence over time. Objective This study aims to develop a (GLM) to analyse the temporal patterns of nocturnal arrhythmia (NA) and their relationship with sleep disturbances and architecture. Methods This study analysed 7341 polysomnography-derived ECG recordings from the MESA and SHHS datasets to detect non-sustained arrhythmias. NAs were defined as any instance where ECG R-R intervals dropped by 30% of the baseline and then returned to 90% of the baseline. we developed our GLM framework to characterise NA events as a function of SDB and sleep arousal events, sleep stages, as well as previously occurring NA episodes. Results The likelihood of NA occurrence during light NREM sleep was, on average, 10% higher than during REM sleep (p=0.022), and 16% and 40% higher than during deep sleep and wakefulness, respectively (p<0.001). The influence of SDB on NA occurrence in men was 4% higher than in women (p=0.032). Participants older than 75 years showed nearly an 8% higher impact of sleep arousal compared to those younger than 65 years. Conclusion A GLM-framework characterises NAs dynamics and their association with sleep disruptions and architecture.