Identifying people with depression can be difficult even for trained professionals.
Now, researchers are looking at the possibility that a computer could do a better job.
A new study published this month in examined how effectively a computer program can spot signs of depression from social media posts.
However, experts have some concerns about patient privacy and appropriate treatment recommendations if this research starts to be used in the real world.
Depression affects millions of people in the United States every year. Sixteen million adults experienced one major depressive episode in 2015, according to the
Spotting depression in people can be difficult, with primary care physicians missing signs in patients about half the time, according to the
How the research was conducted
Researchers from the University of Vermont and Harvard University created a program that looked at Instagram data from 166 people.
The subjects included 71 people with a history of clinical depression.
The researchers looked for patterns in more than 40,000 of the subjects’ Instagram posts.
“Although we had a relatively small sample size, we were able to reliably observe differences in features of social media posts between depressed and non-depressed individuals,” Dr. Andrew Reece, a study co-author from Harvard University, said in a statement.
The researchers looked at how often a user posted photos, how many people were in the photos, whether or not they used filters, and if the picture’s saturation had been affected.
They found certain patterns were more present in people with a history of depression than other users.
“Our analysis of user accounts from a popular social media app revealed that photos posted by people diagnosed with depression tended to be darker in color, received more comments from the community, were more likely to contain faces and less likely to have a filter applied,” Dr. Christopher Danforth, a study co-author from the University of Vermont, said in a statement.
Danforth also pointed out that people with a history of depression were more likely to use the black-and-white filter and were more likely to post more often.
Additionally, photos that were darker with blue and grayer tones were more associated with users who had a history of depression.
Once they put these findings into an algorithm, the computer program was able to correctly identify about 70 percent of the depressed users.
The researchers acknowledge that this study is just a first step and that depression is complex, often coinciding with other conditions such as anxiety, bipolar disease, or chronic pain, among others.
Potential problems of an Instagram diagnosis
Ramani Durvasula, PhD, a psychologist and professor at California State University Los Angeles, said she thought the study was interesting. But she was skeptical that a computer program could help people with the condition.
“Depression is not just one thing. It’s pretty complicated,” she told Healthline.
Durvasula said she was also concerned that if social media companies — which are private businesses — start to use these programs to identify users likely to be depressed, it may not lead to those users getting appropriate treatment.
“Here’s the rub, is what do you tell people, if they are depressed? she said. “We don't always agree on what is the right treatment for depression.”
Durvasula said she was worried that the companies would sell data on potentially depressed users. These people could then start to get marketing for antidepressant medications without getting any information on counseling.
She also worried that these findings would not translate across multiple demographic groups and cultures.
“We always looked for a magical blood test,” for depression, she said. “I’m not sure this is going to be it.”
Pamela Rutledge, PhD, said she was fascinated by the study and how it linked back to some visual techniques used in psychology, like the Rorschach “ink blot” test.
“What I find really interesting about it [is] I find image very reflective of what goes on in the whole person,” Rutledge explained.
While she thought the study was interesting, she also said there needs to be far more research done to see if these findings would hold up for larger populations.
“I would be very cautious about going right to ‘We can diagnose people,’” Rutledge told Healthline. “Just as humans are fallible, tools are, too.”