Objectives To examine the hypothesis that surprising experiences of headache triggers are associated with daily headache activity. Background Little is known about the specific environmental or behavioral interactions that might trigger a headache attack in a prone individual. We propose that headache trigger exposures can be usefully characterized, not only by their mechanisms of action, but also on the degree of “surprise” they present to the individual. This hypothesis is based on elements of information theory: that unusual events and experiences carry more information than common events and experiences and that headache attacks are associated with reactions to uncommon or unexpected biopsychosocial exposures. Methods A secondary analysis of the Headache Prediction Study, this prospective cohort study followed N = 95 individuals with episodic migraine who contributed 4195 days of diary data. Information was collected on daily levels of several common headache triggers: number of caffeinated beverages, number of alcoholic beverages, stress (Daily Stress Inventory), and mood disturbance (Profile of Mood States). The probability of observing variations in each trigger was used to estimate the “surprisal” of experiencing each trigger, and this information, measured in bits, was used to predict headache attacks. Results Participants experienced a headache attack on 1613 of 4195 days (38.5%). Each of the triggers was modestly related to the probability of experiencing a future headache in a complex manner that involved interactions between current headache status, current levels of the trigger, and lagged (previous) levels of the trigger. However, when expressed as a surprisal, the associations were simplified and strengthened. For example, each of the individual trigger surprisals exhibited a meaningful association with the development of a future headache attack (expressed as a 1 SD change in surprisal), with odds ratios ranging from a low of 1.11 (95%CI: 1.00 to 1.24) for alcohol to a high of 1.30 (95%CI: 1.14 to 1.46) for stress. The individual surprisals could be summed for total trigger surprise and exhibited a reliable association with new onset headache, OR 1.35 (95%CI: 1.17 to 1.49). This score exhibited superior discrimination between headache and non‐headache days from all of the individual triggers (ignoring base rate, AUC: 0.61; AUC: 0.71 with base rate). Conclusions Diverse headache triggers can be uniformly described using their probability distributions. Rare values of headache triggers, or surprising values, were found to have consistent associations with headache activity across a variety of triggers. This finding, if validated in external data using other triggers, has potential to be used to conceptualize the influence of a wide range of headache triggers.

A substantial pool of candidate trigger factors could be considered in the creation of forecasting models. However, because mechanistic information about causal factors that precede a migraine attack is not well understood, and such factors are difficult to measure, empirical models that are based on trigger factors that are merely associated with the onset of headache activity are likely to be the focus of forecasting efforts in the near future. Of such factors, stress has considerable empirical support and has been used to successfully forecast future headache attacks within individuals over time. However, at present, existing models possess only modest levels of discrimination and lack strong resolution in generated predictions. Current headache forecasting models represent an important first step in accurately predicting future headache activity. However, to utilize these models in a preemptive treatment paradigm where the risk of headache is treated prior to the actual experience of pain, these models must achieve greater precision with good calibration and generate predictions that are clinically actionable by individuals in their real-time home environments.