People with depression show distinct patterns of online activity, according to a study published in Scientific Reports. Twitter users who said they had a diagnosis of depression were more active on Twitter in the evening, less active in the early morning, and ruminated more on Twitter from midnight to around 6 a.m.
Major depression is one of the most common mental illnesses around the world and is associated with a range of negative outcomes like increased risk of suicide and disease. While the underlying mechanism of depression remains a topic of study, one contributing factor seems to be poor sleep. In particular, depression has been repeatedly linked to disruptions of the circadian rhythm — the body’s internal clock that regulates the sleep/wake cycle.
A research team led by Marijn ten Thij set out to explore differences in the daily activity cycles of people with and without depression using a unique source of data — social media activity. This approach would offer them clues to the behavioral and cognitive activity of a large data set of people and allow them to estimate circadian rhythm cycles.
The researchers identified 688 Twitter users who had explicitly tweeted about receiving a diagnosis of depression. They then analyzed these users’ past tweets and compared them to the tweet histories of a random control sample of 8,791 Twitter users with no mentioned diagnosis of depression.
The researchers found that the depressed and non-depressed Twitter users’ activity levels followed a similar circadian rhythm — as demonstrated by the same pattern of ups and downs in activity throughout the day. Both groups’ activity levels peaked around 9 p.m. and dropped during the early hours of the day between 3 a.m. and 6 a.m.
However, there were significant differences in the groups’ activity levels at specific times of the day. Namely, the depressed group was much less active on Twitter between 3 a.m. and 6 a.m., 9 a.m. and 10 a.m., and 1 p.m. and 2 p.m. Users with depression were instead more active in the evening, between 7 p.m. and midnight.
To see what types of content might be driving these differences, the researchers looked closer at the tweets being posted during these times. Specialists in cognitive behavioral therapy (CBT) analyzed the content of the tweets, looking for certain words categories related to self-reflection and rumination.
It was found that depressed users (compared to non-depressed) tweeted more words related to rigid thinking and questioning (e.g., could, should), and fewer words related to positive affect (e.g., happy, love) during the time period from midnight to 3 a.m. They also tweeted more words in the personal pronouns (e.g., I, myself) and negative affect (e.g., angry, cry) categories between 4 a.m. and 6 a.m. — which the authors say may suggest higher emotionality among the users with depression. Finally, they used more words in all categories between 5 a.m. and 6 a.m., suggesting heightened rumination and self-reflection during these early morning hours.
Ten Thij and colleagues note that comparing the two groups’ activity levels revealed no evidence of a phase-shift. In other words, the patterns of activity point to similar bedtime and wake-up times among both groups. Instead, differences in activity levels were seen at specific times of the day. Finally, greater use of the self-reflection and rumination word categories further suggests that Twitter users with depression post tweets containing more “depressogenic” language.
“These results suggest that diagnosis and treatment of depression may focus on modifying the timing of activity, reducing rumination, and decreasing social media use at specific hours of the day,” the authors say. They also note that additional study into the word categories used could offer “insight into differences in broader language use between depressed individuals and the general population.”
The study, “Depression alters the circadian pattern of online activity”, was authored by Marijn ten Thij, Krishna Bathina, Lauren A. Rutter, Lorenzo Lorenzo‑Luaces, Ingrid A. van de Leemput, Marten Scheffer, and Johan Bollen.
