Investigating the effects of weather on headache occurrence using a smartphone application and artificial intelligence: A retrospective observational cross‐sectional study

Abstract

Objective To investigate the relationship between weather and headache occurrence using big data from an electronic headache diary smartphone application with recent statistical and deep learning (DL)‐based methods. Background The relationship between weather and headache occurrence remains unknown. Methods From a database of 1 million users, data from 4375 users with 336,951 hourly headache events and weather data from December 2020 to November 2021 were analyzed. We developed statistical and DL‐based models to predict the number of hourly headache occurrences mainly from weather factors. Temporal validation was performed using data from December 2019 to November 2020. Apart from the user dataset used in this model development, the physician‐diagnosed headache prevalence was gathered. Results Of the 40,617 respondents, 15,127/40,617 (37.2%) users experienced physician‐diagnosed migraine, and 2458/40,617 (6.1%) users had physician‐diagnosed non‐migraine headaches. The mean (standard deviation) age of the 4375 filtered users was 34 (11.2) years, and 89.2% were female (3902/4375). Lower barometric pressure (p < 0.001, gain = 3.9), higher humidity (p < 0.001, gain = 7.1), more rainfall (p < 0.001, gain = 3.1), a significant decrease in barometric pressure 6 h before (p < 0.001, gain = 11.7), higher barometric pressure at 6:00 a.m. on the day (p < 0.001, gain = 4.6), lower barometric pressure on the next day (p < 0.001, gain = 6.7), and raw time‐series barometric type I (remaining low around headache attack, p < 0.001, gain = 10.1) and type II (decreasing around headache attack, p < 0.001, gain = 10.1) changes over 6 days, were significantly associated with headache occurrences in both the statistical and DL‐based models. For temporal validation, the root mean squared error (RMSE) was 13.4, and the determination coefficient (R2) was 52.9% for the statistical model. The RMSE was 10.2, and the R2 was 53.7% for the DL‐based model. Conclusions Using big data, we found that low barometric pressure, barometric pressure changes, higher humidity, and rainfall were associated with an increased number of headache occurrences.

Fremanezumab and Non-High-Dose Galcanezumab for Comorbid Cluster Headache in Patients with Migraine: Three Cases

“…Furthermore, the increased release of CGRP and other substances upon stimulation of the trigeminal ganglion and the increase in CGRP in jugular venous blood during CH attacks suggest a close involvement of the trigeminal nerve and CGRP in CH [ 24 ]. The hypothalamus, trigeminal nerve and CGRP are closely involved in both migraine [ 29 ] and CH. Migraine and CH may also be accompanied by clinically overlapping findings; up to 46% of CH patients can show migraine-like features, and migraine patients can have typical trigeminal autonomic symptoms [ 17 ].…”

Common challenges in the development of prediction models

“…This is the case because isolating an exposure–outcome relationship requires careful consideration of any biasing factors (i.e., selection bias, information bias) and the inclusion of any confounding influences in the statistical model. Thus, although incredibly promising, it is not entirely clear how the identified associations in the model proposed by Katsuki et al 1 should be interpreted. For example, are barometric pressure changes uniquely related to novel headache attack risk apart from the propensity to experience barometric pressure changes (i.e., other simultaneous weather patterns, the difference in individuals across locations)?…”

“…The impressive work published by Katsuki et al 1 presents a substantial advancement in studying weather and headache attack risk. All prediction model development studies face similar challenges in evaluating discrimination, calibration, and external validation.…”

“…There is a distinct difference between methods used to generate inferences versus those used to generate predictions 6 . The elegant effort reported by Katsuki et al 1 provides a hybrid of such methods, with several aspects of the modeling based on theoretical considerations (e.g., differences in sex) and others based on ad hoc data‐driven methods (e.g., time‐series clustering, feedforward neural network). Typically, interpreting any associations codified into a prediction model is challenging because maximizing the association between a set of predictors and an outcome differs from maximizing the interpretation between any single exposure variable and an outcome.…”

Trainee highlights

“…He is presently applying AI technology to his research with the Japanese Headache Society, Dokkyo Medical University, Saitama Neuropsychiatric Institute, and more. On a personal note, he is also a singer‐songwriter named Cuty Katsuki 2 …”