Modeling of the Effect of Alcohol on Episode Patterns in Atrial Fibrillation

Growing evidence shows that alcohol triggers paroxysmal atrial fibrillation (PAF) in some patients. How-ever, there is a lack of methods for assessing the causality between triggers and atrial fibrillation (AF) episodes. Accordingly, this work aims to develop an approach to episode modeling under the influence of alcohol for the purpose of evaluating causality assessment methods. The alternating, bivariate Hawkes model is used to produce episode patterns, where the conditional intensity function $\lambda_{1}(t)$ defines the transitions from sinus rhythm (SR) to AF. The effect of alcohol consumption is characterized by a body reactivity function, defined by the base intensity $\mu_{1}(t)$, which alters $\lambda_{1}(t)$. Different AF episode patterns were modeled for alcohol units ranging from 0 to 15. The mean AF burden without alcohol was 17.2%, which doubled with 9 alcohol units; the number of AF episodes doubled from 12.9 with 8 alcohol units. The aggregation of AF episodes tended to decrease after 6 alcohol units. The proposed model of alcohol-affected PAF patterns may be useful for assessing the methods for evaluation of causality between triggers and PAF occurrence.