Development of A Migraine Trigger Measurement System Using Surprisal

BackgroundIndividuals who experience migraine continually seek to understand the causes, or "triggers," of their attacks. Many triggers have been hypothesized, and it is not uncommon for individuals with migraine to be advised to consider a vast number of potential migraine triggers ranging from foods, weather influences, stress, mood states, certain behaviors, and sleep, among many others. Information-theoretic measures such as "surprisal" offer a novel approach to quantifying the unpredictability and diversity of trigger exposures on a single standardized scale. ObjectiveThis study aimed to quantify the within- and between-person variability of migraine trigger exposures using surprisal and entropy measures and to evaluate their potential for stratifying individuals based on trigger exposure patterns. MethodsThis longitudinal daily diary study included participants diagnosed with migraine who completed twice-daily electronic diaries reporting exposures to a range of potential headache triggers. Surprisal values were calculated to quantify the unexpectedness of individual trigger exposures, while entropy values captured overall variability in trigger domains such as sleep, mood, daily stressors, dietary behaviors, and environmental encounters. ResultsN = 109 individuals enrolled in the study and self-reported 187 different headache triggers for up to 28 days, resulting in 540,876 headache trigger measurements. Participants exhibited substantial heterogeneity in surprisal and entropy values across trigger domains, reflecting diverse patterns of exposure both within and between individuals. Morning measures of sleep and mood showed moderate entropy, while evening measures of dietary patterns and environmental encounters exhibited greater variability. A small number of principal components explained most of the variability in surprisal values, suggesting that only a few dimensions might offer the ability to characterize trigger exposure across the variables. ConclusionsThese findings reinforce the utility of surprisal measures for capturing nuanced patterns in the vast array of headache trigger data and support their potential as a measurement tool for stratifying trigger exposure either at the day or individual level.