How we can use smartphones to diagnose and treat depression
HealthRhythms uses mobile phone data to monitor symptoms of anxiety and depression, giving doctors valuable insight into their patient’s mental health and offering new interventions and treatments.
When you’re depressed, it isn’t all in your head. Depression can disrupt sleep, instigate irregular eating and make you less active and social. It can completely disrupt your internal body clock.
The good news: we can monitor these behavioral changes to more effectively diagnose depression. Digital health company HealthRhythms is working to do just that. Using data passively collected from smartphones with privacy sensitive and HIPAA compliant sensing and processing, the New York-based company can assess daily behaviors to paint a comprehensive picture of an individual’s mental health. Soon it plans to go even further: last fall, the company received a $2.1 million grant from the National Institutes of Health to develop an automated recommendation engine capable of delivering personalized, real-time health interventions and behavioral change suggestions for people with mental health conditions. And while HealthRhythms currently provides software directly to healthcare clients to support their patients, it plans to make a version of their system available to the general public soon.
Here, co-founder and CEO of HealthRhythms Tanzeem Choudhury, a TED Fellow and Director of the People-Aware Computing Lab at Cornell University, walks us through how HealthRhythms works, how it fits into the broader personalized-medicine movement and how she thinks we’ll treat mental health in the future.
How does HealthRhythms work?
Currently, the best way to measure mental health outside the doctor’s office is to ask patients questions and let them report their own experiences. These questions look at a person’s overall behavioral patterns: Has your sleep been disrupted? Do you eat regularly? Are you moving too slow or too fast? Have you lost interest in socializing? Most screening tools simply combine answers to such questions to form assessments.
The core idea behind these assessments is that when someone’s mental health is declining, there’s a disruption in their normal behavior. If that disruption persists for two or three weeks, or even longer, it’s often seen as an indication that someone’s emotional well-being is compromised.
At HealthRhythms, we use the sensors on smartphones to sense many of these behaviors, and the way a person typically uses their phone provides a baseline of behavior and routines. We monitor a user’s speech and sleep, as well as how they use their device, which can tell us about cognitive changes. What apps do they use? How many texts do they send? What are their typing habits, including typing speed and word length?
The app also tracks movement and mobility: Where do they typically go? How much time do they spend at home? Are they walking faster or more slowly? With such data, HealthRhythms detects when routines get disrupted, which can indicate that the person is depressed. For instance, when someone gets depressed, their walking rate slows down, and in the manic phase of bipolar disorder their pace of activity speeds up. We can look at sociability too — based on app usage, SMS and voice patterns — to determine whether someone is becoming more isolated.
Who is using HealthRhythms?
The app focuses on depression and anxiety across all age groups due to their high prevalence across the human lifespan, so we address everything from youth anxiety to geriatric depression. Some behavioral symptoms related to sleep, mobility, activity and sociability are relevant for a range of mental illnesses at more specific times of life. In our research, for example, we have looked at schizophrenia, bipolar disorder, mood disorder and depression. However, depression and anxiety are the most common disorders across all ages.
How does HealthRhythms suggest behavioral changes?
The key here is knowing the habits and timing of a person’s routine when they are well, so that if the person begins to experience a mental health condition, we can make suggestions that are meaningful and seem doable to the user.
For instance, the app can suggest engaging in social behavior if it senses growing social isolation in the user. If someone is spending more and more time at home, the system might say “It’s a nice sunny day today, a great day to get a coffee from your favorite café down the block.” For activity, we generate suggestions that could change behavior but won’t cause disruption in lifestyle. By looking at the times a person typically takes a walk, the routes they normally take and the weather conditions, for instance, we can suggest a walk that fits with their basic routine.
For sleep, it’s important to align bedtimes with specific body types. HealthRhythms can estimate someone’s body clock type based on sleep habits — such as the difference between sleep habits on work days versus free days, and the types of app used when they wake up. The app then automatically suggests a sleep pattern aligned to that person’s body clock.
How might these interventions evolve over time?
Lots of apps out there try to engage and motivate alternate behaviors, and offer lots of feedback. But this can overwhelm the user. How many things can you actively keep track of, and how many changes in behavior can you realistically make on a conscious level? Our long-term vision is to make the interventions disappear into the background, modifying a person’s behavior with minimal effort by the user, because we think it will make treatment more effective.
Our long-term vision is to make the interventions disappear into the background, modifying a person’s behavior with minimal effort.
Here’s an example of an almost undetectable intervention: Often, when a person gets stressed or anxious, their heart rate will increase measurably. So we designed a piece of wearable technology called Emotion Check. When Emotion Check senses that a user’s pulse is beginning to race, which suggests anxiety, it will provide very subtle tactile feedback that mimics a slower heartbeat. This feedback can actually slow the heart rate down and reduce anxiety levels.
In this vein, we’re thinking about how to leverage the brain and body’s automatic reactions, and how to design technology that intervenes in minimal ways. Instead of shouting instructions, we’re nudging the user’s behaviors with biofeedback.
How does HealthRhythms work with doctors, nurses and other health care professionals?
When our app identifies behavioral changes in patients that could be unhealthy, we raise alerts for their doctors. In working with those doctors, the key is not to overwhelm them with data but to deliver timely indicators of mental health changes in their patients, and provide a succinct summary of the behavioral changes that could raise red flags.
We also speak the language of clinicians rather than expecting them to learn our tracking lingo. Our behavioral-health API easily integrates with health systems applications and electronic health records, making it easy to incorporate our behavioral health measures into existing systems.
How does HealthRhythms fit into the broader movement to personalize medicine?
Precision medicine is a big initiative. We’re getting more and more information about the biological makeup of unique individuals, how disease manifests in individuals and how treatments work at the level of the individual. Traditionally, the obstacle to personalized medicine has been easy access to such data and measurements as genomics, hormone levels and various other physical manifestations of illness. These standard measurements do exist; the question is really how you use this information to personalize treatment.
But in the context of current mental health practice, such objective measurements don’t exist — it’s typically self-reported. Yes, within certain populations of mental health patients, neural changes do occur, but some of these factors are still not very well understood. Meanwhile, there is no physical test for diagnosing mental illness. By gathering personalized measurements, HealthRhythms will enable mental health care to be on more equal footing when it comes to gathering data to create personalized care.
What is your long-term vision for HealthRhythms?
Mental health is not an isolated condition; it affects all of health. About one half of chronically ill patients have co-occurring depression or anxiety. Research shows that if a patient has one of these conditions along with another disease, their chances of a second heart attack goes up, for example, or their cancer relapse rate rises. The cost of treating someone with co-morbid depression is at least double or triple the cost of treating someone without depression. That’s why we say we deliver mental health as a service — we want to add a layer of mental health care across all of health care.
The TED Fellows program supports extraordinary individuals at work on world-changing projects, helping to raise international awareness of their work and maximize their impact.
