Brain Diseases, Central Nervous System Diseases, Headache Disorders, Headache Disorders, Primary, Migraine Disorders, Nervous System Diseases
Conditions
Keywords
Migraine, Smartphone, App, Prediction, Forecast, Detection, Individual, Machine Learning, Data Science, Algorithm, Behavior, Henry Ford Health System, HFHS, Lifegraph
Brief summary
This study is conducted at the Henry Ford Health System with Lifegraph's behavioral monitoring technology, to examine the relation between migraine attacks and behavioral and environmental changes as detected from the smartphone sensors. The investigators hypothesize that Lifegraph's technology can predict the occurrence of migraine attacks with high precision.
Detailed description
Migraine attacks can damage quality of life and lead to missed work days if not treated in time. These attacks last for about 4-72 hours, accompanied by headache and other symptoms. The time window for early intervention, which can potentially reduce the severity of an attack, lasts 2-48 hours before symptoms are starting to appear (10 hours on average). This time window is defined in the literature as the prodromal phase, when intervention during this phase can allow early treatment to improve the patient's condition and reduce the intensity and duration of the attack. Migraine attacks and the prodromal phase can be characterized by one or more behavioral or environmental symptoms, either causal or resultant. Some of them can be passively measured by the smartphone usage, such as changes in sleep, physical activity and weather. Lifegraph's smartphone application runs in the background of the subjects' personal smartphone, collects data passively and automatically, while rigorously maintaining privacy and with no effect to the daily use. Proprietary machine-learning algorithms analyze the collected data and turn it into behavioral channels, such as activity, sleep and mobility. The technology learns the personal routine of each user and detects changes in his/her behavioral patterns that can indicate an upcoming migraine. Eligible subjects will meet a neurologist, sign an informed consent, fill an initial questionnaire and install the Lifegraph application on their smartphone. The application requires a one-time registration process. During the study, subjects will self-report migraine attacks they experience through the smartphone application. Each report will include start time, end time and pain intensity. Data will be analyzed during the study in order to learn each subject's behavior and his/her migraine attacks. Subjects will be blinded to the app's migraine predictions to avoid expectancy bias.
Interventions
None listed
Sponsors
Henry Ford Health System
Lifegraph Ltd.
Study design
Observational model
COHORT
Time perspective
PROSPECTIVE
Eligibility
Sex/Gender
ALL
Age
18 Years to 75 Years
Healthy volunteers
No
Inclusion criteria
* Individuals who suffer from episodic migraine with 4-14 days of migraine per month (ICHD-3 patients). * Individuals who possess a smartphone - Android version 5.0 and above or iOS version 10.0 and above.
Exclusion criteria
* Individuals who are unable to sign the consent form. * Pregnant women. * Individuals suffering from headaches that do not meet the IHS migraine criteria or don't have moderate to severe chronic pain of VAS chronic grade 4 and above.
Design outcomes
Primary
| Measure | Time frame | Description |
|---|---|---|
| Assessing Lifegraph's predictive ability of migraine attacks before subjects report they experience an attack. | 3 months | Lifegraph has created a scalable and dynamic platform to accommodate different conditions, different types of patients with different types of data, concurrently. This platform converts the raw sensor data accumulating in Lifegraph's servers into behavioral and environmental features that have been found to be informative and helpful in generating insights relevant to migraines. The features are fed into machine learning algorithms that search for early signs of change, that may indicate an oncoming attack. These algorithms may be divided into population-based and personalized models. The study will develop a separate predictive model for each subject to predict the probability of experiencing a migraine attack during a particular interval (e.g. the next 12, 24, or 48 hours). Higher precision values of prediction will represent a better outcome. The precision is expected to be 50-70%, depends on the time passed since first installing the app and the number of reported migraine attacks |
Countries
United States
Outcome results
None listed