Twitter data may help shed light on sleep disorders

Researchers from Boston Children's Hospital and Merck have built the beginnings of "digital phenotype" of insomnia and other sleep disorders based on data from Twitter. This study, published today in the Journal of Medical Internet Research, is one of the first to look at relationships between social media use and sleep issues, and- based on assessments the sentiments expressed in users' tweets - gives preliminary hints that patients with sleep disorders may be a greater risk of psychosocial issues.

The study - led by Jared Hawkins, PhD; David McIver, PhD; and John Brownstein, PhD, of Boston Children's Informatics Program and researchers at Merck - is the product of a Boston Children's/Merck collaboration on social media and sleep announced in 2014.

Insomnia and other sleep issues affect between 50 and 70 million Americans. Apart from their impact on productivity, accidents and risky behaviors, chronic sleep disorders also contribute to diabetes, cardiovascular disease and depression.

Historically, population-level research on sleep disorders has relied on survey methods such as the Behavior Risk Factor Surveillance System. However, such methods are time- and resource intensive, expensive, suffer from long lag times before reporting and are not generalizable to the larger U.S. population.

Research based on social media data may help overcome these limitations. To support such research, Hawkins, McIver, Brownstein and their colleagues sought to develop a "digital phenotype," or baseline profile of what a person suffering insomnia or other sleep disorders "looks" like on social media.

"Sleep deprivation and chronic sleep disorders are not well understood," said Brownstein, who directs the hospital's Computational Epidemiology Group. "We wanted to see if we could use new forms of online data, such as Twitter, to characterize the sleep disordered individual and possibly uncover new, previously-undescribed populations of patients suffering sleep problems."

The research team used publically available anonymized data from Twitter to create a virtual cohort of 896 active Twitter users whose tweets contained sleep-related words (e.g., "can't sleep," "insomnia"), or hashtags (e.g., #cantsleep, #teamnosleep), or the names of common sleep aids or medications. They then compared data from that cohort to those of a second group of 934 users who did not tweet using sleep-related terms. The team examined each user's:

age
total number of tweets
total numbers of followers or people followed
number of favorite tweets (that is, the number of tweets by others that the user had favorite)
length of time on Twitter (that is, how long the user had had an active Twitter account)