Social isolation is now understood to rival the effects of cigarette smoking, hypertension, hyperlipidemia, obesity, and physical inactivity as a risk factor for poor health.

Researchers at Brigham and Women’s Hospital have developed SocialBit, a smartwatch-based sensor designed to track the number and duration of daily interactions of the person wearing it. The research team is currently conducting a study to validate the accuracy of SocialBit in detecting social interactions in a sample of stroke survivors during the period after stroke when they are being observed in the hospital.

Kelly White, an MD candidate at the University of Connecticut and formerly a research assistant in the Department of Neurology at the Brigham, Amar Dhand, MD, PhD, an attending physician in the Department, and colleagues describe the app and the study protocol in BMJ Open.

Background

SocialBit tracks social interactions by sampling ambient audio. For privacy, the application never stores the raw audio but instead stores a series of audio features to be classified by a machine learning algorithm. The algorithm detects social interactions by analyzing how the wearer’s vocal acoustic behavior, such as pitch and intensity, varies over time.

The SocialBit user first creates a voice profile, reading five short sentences aloud for 30 seconds. Thereafter, sound features are classified as social interaction or not social interaction, and SocialBit quantifies the number of social interaction minutes per day. The app detects interactions whether or not the wearer is contributing distinguishable words to the conversation.

SocialBit is designed to accommodate a range of neurologic abilities. There is no need for the conversation partner to wear a sensor.

Study Protocol

The researchers began on June 15, 2021, to recruit adults with acute ischemic stroke at the Brigham for a prospective, observational study. They include patients with a variety of neurological deficits, including aphasia, dysarthria, cognitive changes, and paralysis.

The plan is to enroll 200 patients for up to eight observation days (up to five days at the Brigham and up to three additional days if the patients are transferred to Spaulding Rehabilitation), plus three months of follow-up assessment. The researchers aim to monitor patients for about eight hours daily and collect 16 to 24 hours of data per patient.

Inpatient observation allows research assistants to establish “ground truth” by observing patients via video livestreaming. During each minute of their shift, the staffer records what is happening in the patient’s room, recording answers to questions such as, “Was the patient talking to another person? How many people was the patient talking with? Does the conversation contain a foreign language? Was the TV on? Was there a conversation happening that the patient was not part of?”

In addition to determining the accuracy of SocialBit, the study examines (a) the impact of physical and cognitive impairment on social interaction time and (b) the associations between social isolation and stroke outcomes.

Looking Ahead

SocialBit may prove to be a useful basis for social therapeutics, detecting social isolation with high accuracy and sensitivity so healthcare providers can intervene to improve stroke recovery. It might also provide real-time feedback and coaching to patients, families, and clinicians.

The app can be downloaded onto WearOS devices, and in the future, it may be possible to download it onto other platforms, such as the Apple Watch.