Tracking Migraine Symptoms: A Longitudinal Comparison of Smartphone-Based Headache Diaries and Clinical Interviews

Abstract

Background/Objectives: By leveraging the capabilities of a smartphone-based headache diary, the objective of this study was to determine the amount of agreement between migraine-associated symptomatology during headache events and the symptoms documented during clinician-led intake interviews. Methods: This was a 90-day longitudinal, smartphone-based headache calendar study for participants diagnosed with migraine. Registered headache events were labeled as "definite migraine", "probable migraine", and "not migraine" in accordance with the International Classification of Headache Disorders, Third Edition (ICHD-3) criteria. Symptoms' agreement with clinician-led intake interviews (agreement percentages and kappa coefficients), symptoms' similarity between headache events within users (percentage), and amount of newly registered ICHD-3 symptoms per participant were calculated. Results: Twenty-seven participants provided 505 headache events eligible for analysis. The median agreement between recorded headache event symptomatology and clinician-led intake interview phenotyping ranged between 40% (for events fulfilling "not migraine" criteria) and 55.5% ("definite migraine") (p < 0.001). Higher intraparticipant headache event pair similarity was observed for "definite migraine" pairs (p < 0.01), along with a decreasing trend in similarity as the attack-pair headache distance increases. Over half of the participants registered at least one new ICHD-3 symptom during the study. Conclusions: Electronic diary registrations show substantial longitudinal variability in intrapersonal headache symptomatology, with the similarity of headache events declining over time. The registration of a new ICHD-3 symptom was the rule rather than the exception.

Keywords: diary; headache; mHealth; migraine; prospective; smartphone.

Figure 1

User interface for the submission of answers during the clinician-led intake interview. Blue tick boxes are checkboxes (multiple answers possible).

Figure 2

Headache registration interface for participants during the longitudinal phase of the study. (a) The overview of headache events registered in a timeline. (b) The headache registry module with different steps to be completed.

Figure 3

Overview of available headache event pairs per 5-day bin. The x-axis indicates the distance in days between the headache intervals of an event pair. Only events of the same participant with the same attack scoring can be paired. The vertical black dashed line indicates the upper headache-pair threshold.

Figure 4

Headache event symptomatology agreement with clinician-led intake interviews and its trend over time. Each data point in the subplots corresponds to a bias-free headache event (n = 505), color-coded by the ICHD-3 headache event label. The subplots are divided into columns on the basis of the symptom subset, showing either agreement with the clinician-led intake interviews (Rows 1–2) or the number of checked symptoms (Row 3). Subplots (A–C) use box plots to depict the distribution of intake agreement for different subgroups, with statistical differences determined by the two-sided Mann–Whitney U test. Subplots (D–F) illustrate the trend of this intake agreement for the study period, with a first-order regression line and a 95% confidence interval for each headache event label group. The third row (subplots (G–I)), formatted similarly to the row above, displays the count of checked symptoms over time, but uses a locally weighted regression to emphasize local trends. Finally, the subplot (J) represents a histogram that counts the available headache events for each label group over time, using 5-day bins. Note: p-values: **** = p < 0.0001, *** = p < 0.001, ** = p < 0.01, * = p < 0.05. Abbreviations: ICHD-3, International Classification of Headache Disorders Third Edition; Q1, first quartile; Q3, third quartile.

Figure 5

Per-participant mean IoU symptom similarity using paired headache event bins (bin size = 5 days). All subplots display median-aggregated IoU values of headache event pairs, using 5-day bins based on the headache distance of each pair. The headache event pairs were formed by combining headache events from each participant that have the same group label (i.e., “definite migraine”, “probable migraine”, or “not migraine”) and are color-coded accordingly. The columns of the first two subplot rows (subplots (A–F)) indicate the symptom sublists. In the first row (A–C), box plots illustrate the distribution of these per-participant mean-aggregated similarity values across different subgroups, with the two-sided Mann–Whitney U test indicating statistical differences. The second row (D–F) visualizes the similarity trend over time, represented by the distance of headache event pairs, featuring a first-order regression line and a 95% confidence interval for each classification group. The bottom subplot (G) is a histogram that counts the number of participants providing these mean aggregated values for each corresponding headache event label and 5-day bin. Note: p-values: **** = p < 0.0001, *** = p < 0.001, ** = p < 0.01, * = p < 0.05. Abbreviations: ICHD-3, International Classification of Headache Disorders, Third Edition; IoU, intersection over union.

Figure 6

Number of participants registering new canonical ICHD-3 symptoms (n = 27).