Multiple sclerosis (MS) affects around 2.8 million people worldwide. People living with MS can experience many symptoms, including neurological symptoms affecting cognitive function and movement, depression, fatigue and poor sleep. Some symptoms can be managed, but because clinic visits are often infrequent, treatment can sometimes lag behind what people need.
Smartphones and wearables can track various aspects of health passively and continuously. This creates an opportunity to use data from people’s devices to predict when symptoms might worsen. The month’s Tap into Research looks at a study testing whether machine learning models can predict short-term changes in MS symptoms using data from smartphones, wearables, and self-reported questionnaires.
Participants took part for 12-24 weeks and used their own smartphones. The research team collected call activity, location, and screen use. Participants also wore Fitbits provided by the study, which measured heart rate, step count, and sleep patterns.
Alongside this passive monitoring, participants completed a short symptom questionnaire on their smartphone three times per day.
Every 2 weeks, participants completed a validated questionnaire on depressive symptoms. Every 4 weeks, they completed questionnaires on neurological symptoms, fatigue, and sleep quality. These questionnaires were used as the “gold standard” to judge how well the models predicted symptoms.
Models that included smartphone and wearable data were better at predicting short term worsening in depression, neurological symptoms, fatigue, and sleep quality than models that did not use this data.
The study also found:
- Using smartphone and Fitbit measurements from the period just before symptom rating gave the best predictions for depression, neurological symptoms, and sleep quality.
- For fatigue, also including data from further back in time improved predictions.
- For three of the four symptoms, adding the daily self-reported questionnaires did not improve predictions. The exception was depression: the question “How depressed do you feel?” helped predict future depressive symptoms.
This study shows that data collected from smartphones and wearables during everyday life can help predict when MS symptoms might worsen. This could allow people with MS to monitor their symptoms more easily at home and help clinicians provide the right support at the right time.
Future research could explore how well this approach works outside a study setting, where people would need to keep using their devices, without the close monitoring that helped maintain adherence during the study.
Read the paper: Longitudinal digital phenotyping of multiple sclerosis severity using passively sensed behaviors and ecological momentary assessments: real-world evaluation
0 Comments