Phone Data Can Show Signs of Depression

The more we learn about phone data, there's less and less noise for patterns to hide in.

Jul 15 2015, 2:45pm

Image: Robert Snache/Flickr

There are patterns in the noise of all that metadata that you and your phone are constantly generating, and one such pattern may be an insight into your mental health.

Researchers at Northwestern's Feinberg School of Medicine just published a paper in the Journal of Medical Internet Research demonstrating that cell phones collect enough data on us that they can even tell when we're depressed.

Analyzing the amount of time that the phone was used and GPS locations collected over two weeks for 28 individuals, and comparing that data to questionnaires that the individuals filled out, the researchers say they were able to identify depressive symptoms in the data with a remarkable 87 percent accuracy.

As the study states, depression is associated with several behavioral components: reduction in activities, or changes in sleep patterns. Phones that demonstrate that individuals are going fewer places, or whose schedules become irregular, could be exhibiting symptoms of depression.

Example GPS location data from the study. Each circle "represents a histogram bin which has a size of 500 by 500 meters" to help anonymize location tracking. Brighter colors indicate more samples caught by that respective bin.

Excessive mobile phone usage can be considered compulsive behavior and has also been linked to some symptoms of depression, according to the study. The study's lead author Sohrob Saebhe told me that the average daily usage for depressed individuals was about 68 minutes, while for non-depressed individuals it was about 17 minutes. (While that might seem low, the study's metric for "actual" usage likely differed from other studies of average daily phone usage, but that should not affect the correlation found within this study.)

Granted, the researchers didn't collect what the subjects were doing with their phones for that time, so they don't know if the subjects were chatting on the phone or scrolling through Facebook and getting (further) depressed by looking at other people's lives.

A discovery like this, just like the collection of all this data to begin with, is a double-edged sword. On the one hand, as Saebhe told me in an email, the data afford doctors an opportunity to "objectively and passively monitor people at risk of depression (e.g., people who have been diagnosed with depression in the past)," and even send "recommenders" to patients diagnosed with depression, recommending they get out of the house and use their phone less.

Getting recommendations to use your phone less from your phone is a little ironic, but it makes sense, and it certainly sounds like a better way to collect data than the methods used now.

"We won't need to bug [patients] with questions about their mood every day, as is done in longitudinal studies of depression," Saeb said. "The information gathered in this way will be useful to clinicians who are conducting a long-term treatment on patients."

On the other hand, as we're starting to learn, anything about you on the internet can and be used against you, if not in a court of law (although, yes there), then elsewhere, like in the marketplace. Companies like Google have already proven that there's a market for surreptitiously—but legally—collected information about your health. Phone companies have already demonstrated that they're willing to sell your metadata, which even if anonymized can be traced back to you.

The Northwestern study only worked with people who consented to being tracked and the data were kept encrypted on secure servers, Saeb told me, and he explained how the algorithms could be fine-tuned so doctors only got data like the number of new locations visited, rather than, say, the exact GPS coordinates themselves.

But doctors are ones we want to share at least some of this data with, and they're bound by HIPAA laws that are in place to protect doctor-patient confidentiality. The question is whether you want, say, the NSA to know that you are depressed, or if you have cancer, or if you want AT&T to tell its advertising partners that you're depressed. The information was already being generated, and the more we learn about that data, there's less and less noise for patterns to hide in.