A study involving more than 53,000 app users suggests headache diaries could help forecast migraine, although wider clinical benefits remain unproven.
Artificial intelligence could help people living with migraine anticipate difficult days by identifying patterns in their recorded headache history. A peer-reviewed study published in Neurology Open Access reported 91.2% precision in forecasting next-day migraine, offering evidence that routinely logged symptoms may contain useful warning signals.
Researchers analysed 770,473 daily reports from 53,065 users of the Nerivio app, collected between January 2020 and July 2025. They evaluated seven machine-learning algorithms using diary entries, questionnaires, demographic information and local weather data. The strongest-performing model achieved 81% overall accuracy and 80% sensitivity, according to developer Theranica’s September announcement.
Precision has a specific meaning. The 91.2% figure describes how often a day predicted to involve migraine actually did. It does not mean the system correctly predicted 91.2% of all days or detected that proportion of all migraine days. Those questions are measured separately by accuracy and sensitivity.
A person’s headache history over the preceding month provided the strongest predictive information. In particular, average headache severity across the previous 30 days was more influential than symptoms reported shortly before an attack. The finding suggests that consistent diary entries could contribute to forecasting as well as documenting past experiences.
The technology is already incorporated into Nerivio’s Your Day Ahead feature, according to Theranica. Its stated purpose is to help users prepare for an anticipated migraine day; it does not diagnose migraine or recommend changes to prescribed treatment. prnewswire.com
Important questions remain. Participants used a single therapeutic device and tended to be frequent, treatment-engaged patients. Results therefore need confirmation in broader migraine populations. The forecasted outcome also included mild headache pain, so the headline figure should not be interpreted as predicting only severe attacks.
Forecasting offers a potential planning aid, but predictive performance alone does not establish that using the feature prevents attacks or improves treatment outcomes. Further clinical evaluation will help determine how much practical benefit these forecasts deliver.
Sources:
