Penn researchers helped develop an algorithm to help predict depression from Facebook posts

Research from the Penn Center for Digital Health and the World Well-Being Project suggests that social media posts can be used to screen individuals for depression.

In a study published in the Proceedings of the National Academy of Sciences, researchers from the University of Pennsylvania and Stony Brook University found that certain language used in the posts could predict future diagnosis of depression in their medical records.

"There's a perception that using social media is not good for one's mental health," principal investigator H. Andrew Schwartz said to Eureka Alert. "But it may turn out to be an important tool for diagnosing, monitoring, and eventually treating it. Here, we've shown that it can be used with clinical records, a step toward improving mental health with social media."

Although nearly 1,200 people volunteered their data, researchers narrowed in on a final sample of 683 participants who provided archives of their Facebook posts, 114 of whom were already diagnosed with depression in their electronic medical records. The study encompassed a total of 524,292 Facebook posts over the span of many years, and for depressed participants, researchers only analyzed posts prior to their first diagnosis, the PNAS report says.

The researchers categorized certain words frequently used in the Facebook posts as “depression-associated language markers.” These words reflect feelings of hostility, loneliness, sadness, rumination and increasing self-reference. The prediction model also analyzed aspects such as the length of the Facebook posts and the frequency of the posts being made. 

"Social media data contain markers akin to the genome," Johannes Eichstaedt, the study's founding research scientist, explained to Eureka Alert. "We can comb social media data to find these markers. Depression appears to be something quite detectable in this way; it really changes people's use of social media in a way that something like skin disease or diabetes doesn't."

Eichstaedt also is a postdoctoral fellow in Penn’s Psychology Department.

According to a public release published by Eureka Alert, the World of Well-Being Project had already been researching how words reflect individual feelings for the past six years. Eichstaedt expanded the ongoing research in 2014 into how language on social media can pinpoint mental health issues directly.

To lead the study, Eichstaedt and Schwartz collaborated with Robert J. Smith, Raina Merchant, David Asch, and Lyle Ungar from the Penn Center for Digital Health along with data scientist Patrick Crutchley and machine learning expert Daniel Preot¸iuc-Pietro.