Verily has published a study that supports the use of smartphones as a means to measure symptoms of depression. For years, physicians lacked a way to measure clinically relevant behavior objectively outside of an office visit, causing data to be drawn only from self-reported surveys and other subjective sources. Verily’s study suggests that sensors and active tasks recorded by smartphones provide a low-burden, low-cost, and broadly deployable way to capture real-world data from patients that could augment clinical decision-making and move the field of mental health closer to scalable, measurement-based care.

"The COVID-19 pandemic has placed a much larger emphasis on mental health and the lens through which we view conditions like depression. Subjective measures aren’t as relied upon in other medical or scientific fields -- and it's time that we bring mental and behavioral health up to that same standard," says Vivian Lee, President, Health Platforms, Verily Life Sciences. "We set out to understand how novel sensing devices could be used to collect data that would be translated into clinically relevant measures, and we created a compensation-based study that would keep participants engaged and yield significant results to further mental health innovation."

Collecting the Study Data

The 12-week study "Toward a Mobile Platform for Real-world Digital Measurement of Depression"* evaluated 384 users in a controlled opt-in environment -- 313 with self-reported depression and 71 in a nondepressed control group.

The study evaluated passive monitoring from 20 phone sensor data streams (e.g., ambient audio level, location, and inertial measurement units) along with active surveys and voice diaries. From these data, the researchers derived 34 summary variables ​​capturing aspects of sleep, physical activity, sociability, mobility, affect and mood and compared them against the clinically validated Patient Health Questionnaire (PHQ-9) self-survey. Throughout the 12 weeks, participants were engaged through a chatbot interface that delivered gif images and celebratory content that encouraged continuous participation.

The Results: Understanding the connection between device-derived data and PHQ-9 

Of the 34 behavioral features selected for analysis, 11 showed a significant (P<.001) correlation with the weekly PHQ-9. For example, participants who visited many different locations in a given week showed less signs of depression. Analyzing the voice diaries showed that lower sentiment scores - a number showing how negative or positive the content of the voice diary was - were associated with more severe symptoms of depression. Another interesting finding was that using less emojis in outgoing messages was associated with more symptoms of depression.

Putting all 34 behavioral features together, the researchers created a classification algorithm to determine whether the participant was depressed (PHQ-9 score >10) or not (PHQ-9 score <10) in a given week. This model performed above chance with a mean area under the curve (AUC) of .656.

The Application: How can this extend broadly to mental health treatment?

By using mobile devices, Verily was able to gather more frequent data points about participants’ activity and well-being. The study ultimately demonstrated how passive and active monitoring together can be used to create potentially actionable depression insights.

For instance, by measuring real-word behaviors such as social app usage, time spent at home or other places, physical activity, or environmental factors such as ambient noise, clinicians can gain a more complete, continuous, and objective picture of their patient’s circumstances. In addition, learning about participants' normal daily activities and routines could lead to the development of an early alert system when unusual changes are detected.

The study also yielded insights about how to improve future remote-based mental health research. Using an interactive, lighthearted approach to engaging with participants proved beneficial according to two testimonials:

"The app made me feel like I had an everyday purpose. I looked forward to filling it out. I enjoyed the interaction."

"It put my mental health into perspective and I had to answer how I was feeling, not what people expect me to feel."

Next Steps for Verily

Verily’s science-backed, virtual care platform Onduo will evaluate these findings to understand potential application to the Mental and Behavioral Program it will be making available to health plan and employer clients. While patient privacy is always at the forefront of its development, when agreed to by individuals using the Onduo platform, the power of frequent, objective data points may help Onduo identify when individuals are in need of mental health support as part of its whole-person approach to care. This will allow the health technology brand to apply personalized resources and care management to individuals based on their specific needs.

The behavioral variables identified in this study may further be used in other clinical studies on depression or other mental health conditions. They can serve as additional clinical endpoints and real-world evidence to test the efficacy of novel therapeutics for mental illness. For example, most recently the Mood App had been deployed by our partner LivaNova in a randomized controlled trial on using neurostimulation for difficult-to-treat depression and bipolar disorder.

*Nickels S, et. al. Toward a Mobile Platform for Real-world Digital Measurement of Depression: User-Centered Design, Data Quality, and Behavioral and Clinical Modeling. JMIR Ment Health 2021;8(8):e27589