Smartwatch app to track social interactions of stroke survivors may help maximize recovery

A smartwatch app designed to measure social interactions of hospitalized stroke survivors may enable new treatments to preserve or enhance cognition, social engagement and quality of life after a stroke, according to a preliminary study to be presented at the American Stroke Association's International Stroke Conference 2026. The meeting is in New Orleans, Feb. 4-6, 2026, and is a world premier global event dedicated to advancing stroke and brain health science.

Researchers developed a machine learning app called SocialBit, which is compatible with Android smartwatches, and can identify social interactions in both people with and without neurological conditions. The researchers noted that other devices to track social interactions are focused on people without disabilities. SocialBit is currently only available for use in research projects.

According to the American Stroke Association, the loss or change in speech (dysarthria) and language (aphasia) profoundly alters the social life of stroke survivors. Yet, research has shown that socializing is one of the best ways to maximize recovery after a stroke.

My previous research has demonstrated that stroke survivors who are socially isolated or have a smaller circle of friends and family have worse physical outcomes at 3 and 6 months after a stroke. We created a tracker of social life customized for stroke survivors. Tracking human engagement is crucial, and social isolation can now be identified in real-world situations. This may be addressed by notifying the patient, family members, caregivers and health care professionals of social isolation."

Amar Dhand, M.D., D.Phil., study lead author, associate professor of neurology, division of stroke and cerebrovascular disorders, department of neurology, Mass General Brigham, Boston

Amar Dhand, M.D., D.Phil., study lead author, associate professor of neurology in the division of stroke and cerebrovascular disorders in the department of neurology at Mass General Brigham in Boston

Researchers recruited 153 adults during their hospitalization for an ischemic stroke.

Participants wore a smartwatch with the SocialBit app while they were in their hospital rooms, between 9 a.m. to 5 p.m. daily, for up to 8 days (some of which may have been after transfer to a rehabilitation hospital). The app logged the amount of socialization time according to acoustic patterns from the participant and/or another person talking, indicating social engagement. During the same timeframe, members of the research team watched a livestream video of the participants and logged the same minute-by-minute social interactions of the participants with others.

The researchers found:

• Compared with human observers, SocialBit was 94% as accurate in recognizing social interactions.
• In patients with aphasia, SocialBit maintained accuracy at 93%.
• SocialBit's performance remained consistent despite TV noise, side conversations, different environments (rehabilitation unit versus hospital) and across various Android smartwatch models.
• Participants who had a more severe stroke had less social interaction, with about a 1% drop in total social interaction minutes for each 1-point increase in the NIH Stroke Scale, a standardized tool used to assess the severity of a stroke.

"I was surprised by how well the app performed for people with aphasia. We used SocialBit to capture sounds instead of words to protect privacy, and this feature ended up being helpful for people with limited language skills," Dhand said.

"The SocialBit app may also help people recover from brain injuries. It can support therapies like speech, occupational and exercise therapy," he said.

Future research could use SocialBit to measure how many people are at risk for social isolation while in the hospital and after they leave. It could also explore how this isolation is related to depression and other mental health changes that can happen after a stroke.

"We can also test if this app can help with other brain injuries and in healthy aging to keep and improve brain health over time," Dhand said.

One limitation of the study was that the detailed evaluations of social interactions were only tested in hospital or rehabilitation settings.

Chair of the American Heart Association Stroke Council and chair of the writing group for the Association's 2024 Guideline for the Primary Prevention of Stroke Cheryl Bushnell, M.D., M.H.S., FAHA, said, "This research is fascinating in its capture of social interactions, which I presume can distinguish between conversations from case managers, nurses, therapists and the care team from non-hospital personnel. If not, then the amount of social interaction could be dependent on the size of the care team (academic teams have more staff, and trainees spend more time in the patient's room), or the nurse-to-patient ratio. If the app does distinguish hospital from non-hospital personnel, then distance from the hospital and the number of family and friends become major factors in the degree of social interaction. Regardless, there are multiple interesting ways this app could be used in future studies, including measures of quality of hospital care and social interactions at rehab facilities and nursing homes." Bushnell is also a professor and director of the Center for Transformative Stroke Care in the department of neurology at Wake Forest University School of Medicine in Winston-Salem, North Carolina, and she was not involved in this study.

Study details, background and design:

• 153 adults (average age of 66 years, ranging from 26 to 94; 53% were men, 46% were women, 1% other) were hospitalized with ischemic stroke. Among all study participants, strokes were mild (median score of 2 on the NIH Stroke Scale), as was cognitive impairment (mean score of 23.6 on the Montreal Cognitive Assessment).
• Participants were hospitalized at Brigham and Women's Hospital between June 2021 and March 2025 for the initial treatment of stroke. No participants were in the intensive care unit, where visitation was restricted. Some patients who stayed in the study were moved to Spaulding Rehabilitation Center during part of the assessment period.
• Participants were not included in the study if they were receiving end-of-life care, had a prior diagnosis of dementia, or if they or their decision-maker could not speak or understand English well enough to complete surveys or confirm informed consent to participate in the study.
• Between 9 a.m. and 5 p.m. daily, participants wore a smartwatch equipped with the SocialBit algorithm. Using acoustic patterns indicating human speech, the device quantified the number of minutes per day participants engaged in social interaction with another human. These were compared with ratings from human observers monitoring livestream video during the same time periods.