Smartphones could be used to detect depression, Verily finds
Cell phone data can tell you a lot about a person: where they go, who they talk to, and how much time they spend scrolling on the internet. These metrics could also give insights into people’s mental health, according to research backed by Alphabet’s life sciences subsidiary.
Verily published a study in JMIR Mental Health, asking 600 participants to download an app for a remote, 12-week study. It compared 20 different data types including location, background noise, and sentiment from audio diaries, to PHQ-9 surveys, the current standard for assessing depression severity.
Researchers found that sentiment analysis from weekly voice diaries and sleep duration were the most strongly correlated with changes in reported depression.
Some of these findings were intuitive, said Dr. Natalie Lester, CMO of OneFifteen, an Ohio-based clinic that provides addiction treatment, and was created through a partnership with Verily.
For example, one hallmark of depression is changes in cognition, where people might perceive the negative aspects of something as being more true. Many people also experience challenges with sleeping.
However, the study didn’t necessarily prove these metrics work better than the PHQ-9, she said. It had a few other limitations, including that people were paid anywhere from $135 to $530 to participate in the study (although adherence was the same across the board). Additionally, it was limited to phones with an Android OS, and participants skewed young, white and female.
A possible benefit
Physicians interviewed by MedCity News saw a possible benefit for this technology, but agreed that it wasn’t quite ready for clinical adoption. For example, more research would need to be done to determine which signals are the most significant, and how to aggregate that information in a way that it could be useful for clinicians and patients.
“If it’s a tool that the clinician has to log into a new program and take several clicks to get into it, they get the raw data and have to analyze that, it’s of diminishing value,” Lester said. “If the clinician can almost passively absorb that because there is some algorithm in the background that’s screening … maybe I’m doing a telehealth visit and that information gets shared with me, that’s really valuable information. We’re definitely not at that point, but I think that would be the goal.”
Stephen Schueller, an associate professor of psychological science at UC Irvine and executive director of One Mind PsyberGuide, said there could be a clinical benefit to these models, but we’re probably a few years off from realizing it. While research points to the ability to use passive data collection to predict someone’s mental health status, the challenge is scaling it into something that is clinically meaningful.
“As a clinician, I can also tell you if a person is depressed, do I need an app to do that, so how would such a system fit into a system of care that would add value to a client?” he wrote in an email. “That’s the bigger question to me, is their value added to the individual contributing this data?”
These metrics also can’t be separated out from talking to patients, in the same way that a blood test or an x-ray alone couldn’t be used to make a diagnosis, said Dr. John Torous, Chair of the American Psychiatric Association’s Committee on Mental Health Information Technology. His team at Beth Israel Deaconess Medical Center built an app, called MindLamp 2, that takes in similar information to Verily’s app and is used by clinicians to manage patients’ care.
On an individual basis, it can be helpful in working with patients to understand patterns in behavior that they didn’t understand before, Torous said. But it varies from person to person. For example, some people might struggle more with different times of day, different medication, or even the impacts of air quality and pollution.
“You can help people untangle how their emotions are working,” he said. “Is it that you didn’t sleep well and your cognition was worse? … Patients like that because we can make these intangible mental health things a little more tangible.”
What about privacy?
Of course, most people would consider their location, phone activity, and their mental health incredibly personal. It begs the question — will people be comfortable sharing this information in the first place?
It depends, the researchers said, if the app is transparent about how people’s information will be used, and if there’s a clear clinical benefit.
For Torous, that meant building a smartphone app from scratch, and educating users specifically on their privacy rights, and how to identify apps that have good privacy practices. People are rightfully cautious, as the majority of apps available today don’t always protect people’s private information.
“It’s a pretty wild marketplace out there that doesn’t respect people’s privacy,” he said. “I would argue mental health data is the most sensitive, but it’s across the whole (digital health) space which makes it even worse.”
Schueller actually conducted a survey on 211 people’s comfort sharing their private health information along with mobile data. According to the results, published in JMIR, people were generally more comfortable sharing sensor data with their doctor than with an EHR system or family members, and were more comfortable sharing health information than personal data.
“So someone who has an experience of mental health issues might be more comfortable sharing their data with a professional who is providing care to them to help them manage their condition, but someone who has never experienced mental health might be less comfortable sharing data more from a large-scale surveillance, public health perspective,” he wrote in an email. “I think another aspect here is that a lot of these companies already have all this detailed type of information (location, speech, phone activity, purchases, browsing history, etc.) but we don’t always know how this data might be used or what it could reveal. So I think this is why trust, transparency, and transactional value is so important here.”
Verily isn’t the only one
Several other companies have tried their hand at using technology as a benchmark for mental health, but with limited success.
In its early days, Ginger started off by using cell phone activity, detecting patterns to gauge people’s mental health, before pivoting to offer text-based coaching and virtual therapy visits. Although it had drummed up 40 hospitals that were using the technology, adoption was slow, and the company’s former CEO told MedCity News in a previous interview that it wanted to offer a full service by working directly with the people who are paying for care. The company declined to comment further for this article.
Another company, Mindstrong, has also taken a similar trajectory: It still can monitor typing, swiping and scrolling to measure mental health symptoms, but it’s wrapped in a broader offering that includes therapy services and care plans.
Google’s competitor, Apple, might also be trying its hand at this technology. According to a recent report by the Wall Street Journal, the company is “using an array of sensor data,” such as mobility, physical activity, sleep patterns, and typing, to build algorithms to detect these conditions.
But in the longer term, it’s difficult to tell whether these projects will persist, or go by the wayside as so many others have.
Photo: phototechno, Getty Images