Tweets may help reveal Twitter users’ levels of loneliness and could thus be used to help target interventions that alleviate morbidities in this condition, according to results of a statistical analysis published in BMJ Open.
“Loneliness is associated with an increased risk of dementia and cardiovascular disease,” Sharath Chandra Guntuku, PhD, research scientist at Penn Medicine’s Center for Digital Health, told Healio Psychiatry. “It is also associated with high hospital utilization. If we can identify symptoms of loneliness early on, there is an opportunity to intervene to improve these users’ health outcomes and potentially better triage them before a hospital visit.”
Guntuku and colleagues collected approximately 400 million tweets posted in Pennsylvania between 2012 and 2016 that contained the words “lonely” or “alone.” They identified posts that seemed to refer to the public health meaning of lonely, rather than posts using it colloquially or as a joke. Further, they used natural-language processing to characterize diurnal patterns and topics of users’ posts, whether language seemed to predict manifestations of loneliness and associations with linguistic markers of mental health, they wrote.
The researchers found 6,202 users whose posts included words like “alone” or “lonely” more than five times during the review period. They then compared the entire Twitter timelines of these users with a matched group who did not use similar language in their posts. Users who used this language also reflected themes about difficult interpersonal relationships, substance use, psychosomatic symptoms, desiring change and insomnia. For instance, lonely users tweeted nearly twice as much as the control group and were significantly more likely to tweet at night. Linguistic markers revealed that they also had an “extremely high association with anger, depression and anxiety” compared with the control group.
The non-lonely group was more likely to engage in social interactions on Twitter by including other user’s Twitter handles in their own posts more often than lonely users.
The researchers predicted expressions of loneliness online with an area under the curve of 0.86 using a random forest model.
“While there are some existing stats that show about 40% of adults feel lonely, analyzing the words and phrases they use and themes they post online showed us what they are actually going through and the nature of this loneliness epidemic,” Guntuku said. “If about half of us feel that we don’t have anyone to talk to, we should do something about it.” – by Joe Gramigna
Disclosures: The authors report no relevant financial disclosures.