Can your smartphone tell you if you have depression?

Originally posted by Carina Storrs on CNN.com on July 15, 2015.

Researchers at Northwestern University in Illinois tested an app they developed called Purple Robot. It uses data from a number of sensors in the smartphone that detect location, movement, phone usage and other activities to assess if a user is likely to have depression.

“The main reason for the development of the app is to see if we can objectively and passively identify if people are depressed,” said Sohrob Saeb, a postdoctoral research fellow at the Feinberg School of Medicine at Northwestern University who is one of the developers of Purple Robot.

In the study of Purple Robot, Saeb and his colleagues at Northwestern and Michigan State University looked at GPS or phone usage data among 28 participants for two weeks.

They found that Purple Robot could identify 87% of the participants in the group who were determined to be at risk of depression according to PHQ-9, a nine-question test for depression, based only on GPS data on how much users moved between their regular locations. The more users moved around, the less likely they were to fall into the at-risk category.

In addition, by identifying the participants who used their phone the most, including everything from texting and playing games but not talking on the phone, Purple Robot could detect 74% of those in the at-risk group. Data on both GPS and phone usage were not available for enough participants to let the researchers see how well Purple Robot performed when both data sets were available, Saeb said.

However, PHQ-9 is only a screening tool that tells you if you have an above-average chance of having depression and is not enough to diagnose depression, said Dr. Scott Monteith, clinical assistant professor of psychiatry at Michigan State University, who has not been involved in developing or researching Purple Robot or other smartphone apps.

The way the test was used in the study, with a low cutoff score, it probably incorrectly identified many of the participants as being at risk of depression who were not, he added.

To get a better idea of the effectiveness of Purple Robot, the researchers are going to do a study involving more participants over a longer period to see if the app can detect changes in behavior over time, Saeb said. In addition, the group will see if they can improve Purple Robot’s ability to spot depression by including additional data, such as how long people talk on the phone and who they talk to.

Depression is a debilitating illness that affects about 17% of people at some point in their lives. Meanwhile, it is estimated that by 2025 more than 5 billion people in the world will have a smartphone, and their sensing capability will be above and beyond that of today’s iPhones, Androids and Blackberries.

There are probably hundreds of apps that promise to improve your mental health, from offering tests to gauge your depression risk to providing information about depression treatments. Others, like Purple Robot, are in the development stage.

Optimism and DBSA Wellness Tracker are two of the apps on the market that track your mood. Ginger.io goes further and analyzes data such as how much users move around on the weekends and how long they talk on the phone, as well as users’ reports of their health, to alert them and their health care providers about concerns with their behavior and mental health. Ginger.io, which is in use at about 30 medical centers, is available through health care providers and as part of research studies.

However the problem with all the apps that are designed to warn about depression risk is that their effectiveness has not been demonstrated, Monteith said.

It is not clear how good these apps are at picking out people who have depression, Monteith said. What’s more, it is not clear how these apps would be “embedded into a broader continuum of care” to ensure that a person or their doctor went from getting an alert from the app, for example, to that person getting a diagnosis of depression and getting proper care, he added.

Even if researchers can get a better handle on the effectiveness of these apps, there are still numerous questions regarding risks, especially about the data they collect not being secure and private, Monteith said.

“The data from these types of apps could potentially end up in anyone’s hands, if the data are moved offshore, which a lot are,” said Monteith, who co-wrote a recent article on health care data privacy. Another way data security could be compromised is that when a company is bought, the buyer may not have to adhere to the original terms and conditions about how the data are used, he added.

Experts including Monteith worry that once data get into the wrong hands, that could potentially jeopardize a person’s ability to get a job, get life insurance or get a loan.

The best way to keep data secure, at least from hackers and thieves, would be to make sure the users control their data, such as by keeping it stored encrypted on their phone, and have apps analyze the data on phones, and never have it sent back to the app developers or other companies, according to Dr. Deborah C. Peel, leader of Patient Privacy Rights, a nonprofit advocacy organization. Monteith is on the advisory board of Patient Privacy Rights.

As for Purple Robot, some of these concerns may not apply for now. Saeb and his colleagues work with encrypted data. However if they eventually make the app public, if they can demonstrate its effectiveness, they would have all the data on secure servers at Northwestern. This type of data centralization, even on secure servers, is a “honeypot” for hackers, Peel said.

So far, the analyses that Purple Robot is doing are really only for research purposes, Saeb stressed. In addition to the work he is involved in, there is also research on whether the app can pick up signs of bipolar disorder among users.

The app gets its name because the color of Northwestern University is purple, and because the developers hope the app can act like a robot and automatically alert a user of his or her mental health risk and also make recommendations to possibly mitigate the risk, such as using the phone less or getting out of the house, Saeb said.

Despite concerns surrounding these apps, Monteith said he is “totally in favor of research [on them], that’s what we need to do.” However, he urged that researchers consider both effectiveness and risks in their studies. “We need to look at what the FDA looks at” when deciding whether to approve medical treatments and devices, Monteith said.