Weather factors associated with recorded headache events and user-level meteorological recording profiles in smartphone headache diary data: a retrospective observational study

Abstract Background Weather changes are commonly reported as headache triggers, but the relationship between specific meteorological conditions and headache occurrence remains inconsistent across studies. We investigated if weather variables and user-level weather profiles were associated with recorded headache events using smartphone-based headache diary big data. Methods This retrospective observational study analysed 714,911 hourly-recorded headache events from 10,248 users of the Zutool ® smartphone application who were presumed to have episodic migraine. Hourly meteorological data from the Japan Meteorological Agency were linked to recorded headache events by meteorological station. A negative binomial regression model with a log link was used to examine if weather variables were associated with hourly-recorded headache event rates. Principal component analysis followed by varimax rotation and k-means++ clustering was used to identify if user-level weather profiles were associated with recorded headache events. Results The mean age of the 10,248 users was 38.4 ± 11.8 years, and 90.2% were female. Longer sunlight duration (incidence rate ratios 1.109 [95% confidence interval, 1.008–1.221]), faster average wind speed (1.035 [1.009–1.062]), more rainfall (1.021 [1.011–1.031]), barometric pressure (BP) decrease from 6 h before to the current hour (0.990 [0.980–0.997] for BP increase) and lower BP at 6 a.m. on the next day (0.984 [0.981–0.987] for higher BP) were associated with significantly higher recorded headache event rates. Compared with type III time-series BP change trajectory (increasing before headache records), type I (remaining low around headache records; 1.088 [1.045–1.132]), type II (decreasing before headache records; 1.218 [1.172–1.265]), and type IV (remaining high around headache records; 1.060 [1.015–1.107]) were associated with higher hourly headache record rates. User-level clustering suggested five exploratory weather profiles associated with recorded headache events. Conclusions Using large-scale smartphone-based headache diary data linked to hourly weather data, we found that recorded headache event rates were associated with several weather factors, although the magnitudes of these associations were generally modest. Exploratory user-level clustering identified five weather-associated recording profiles, suggesting that recorded headache events may occur under heterogeneous meteorological trajectories shaped by both broader BP patterns and short-term weather deviations; however, their clinical and practical relevance remains uncertain.