Technology's Latest Quest: Tracking Mental Health
You can now count your steps, measure your glucose
levels, monitor your blood pressure and track your caloric intake from
your phone or high-tech wristband. But for those dealing with depression
rather than diabetes, or trying to keep tabs on their bipolar disorder
rather than their weight, the pickings are slimmer.
There
are apps that track mood by asking users to fill in surveys, ones that
give tips on breathing or thinking positively, others that remind people
to take their antidepressants or other medications and even some that
turn cognitive treatment into games. The problem is that many lack a
substantial basis in research, and the kind of information a user can
provide can be severely limited. Another problem is that many patients
are either not motivated or not self-aware enough to respond accurately,
says Philip Resnik, a computational linguistics professor at the
University of Maryland, who is studying how natural language processing
can shed light on mental health.
Researchers are
beginning to link to mental health certain types of data that people
can’t track or identify in themselves. Just as an EKG is a more
effective tool to diagnose cardiac disease than asking a patient how his
or her heart feels, these new data sources could radically change our
ability to track mental health.
Technological
tools to flag mental health problems could have a huge impact on these
populations, which could use an easily downloadable app for early
screening instead of jumping through hoops to find a qualified clinician
just to start the process.
In addition, moving to tech
could help better reach a younger generation. Roughly 20 percent of
teenagers (ages 13 to 18) in the U.S. experience severe mental disorders
each year, according to NAMI, while suicide is the third-leading cause
of death for those ages 15 to 24. “A lot of psychology still takes place
on paper,” says David C. Cooper, a Mobile Health Program psychologist
at the
National Center for Telehealth and Technology (T2),
which leads initiatives for the Department of Defense that use
technology to deliver psychological health care options to the military
community. He was referring to traditional therapy tools like the paper
journal he asks patients to keep between sessions, and the notes he
takes during those meetings. With 70 percent of military personnel under
the age of 30, the generation of veterans Cooper treats is starting to
expect tech as part of their treatment.
“If I give patients a piece of paper, they’re going to look at me like I’m some sort of Luddite,” Cooper tells Newsweek.
Cooper has helped develop apps like PTSD Coach,
Mood Tracker, Breathe to Relax and Virtual Hope Box that translate
standard, analog mental health care practices into ones and zeros. But
using new kinds of data and technological capabilities to track mental
health is a daunting task. Unlike measuring glucose levels, which can
easily help patients (and their doctors) understand where they stand,
there is no direct way to measure depression, anxiety, bipolar disorder,
schizophrenia or PTSD.
Carol Espy-Wilson, a professor
of computer and electrical engineering at the University of Maryland, is
working with Monifa Vaughn-Cooke in the Department of Mechanical
Engineering, and Resnik, the computational linguist, to come up with a
complete set of measures—physiological markers like heart rate and skin
temperature, along with patterns based on vocal features, facial
expressions and language use—that could help track mental health.
Espy-Wilson
started by looking at the Mundt database, collected in a study from
2007 that looked at depression and speech patterns. The Mundt study
recorded participants, all of whom were undergoing treatment for
depression, as they spoke freely and assessed their depression on the
standard Hamilton Depression Scale.
Espy-Wilson’s study
built on those findings. She looked in particular at six patients whose
assessments showed the greatest variation in mental well-being week to
week. She found that when they were depressed, their speech tended to be
slower and their vowels “breathier,” and that their voices’ “jitter and
shimmer”—a measure of variability in duration and amplitude of
sound—increased.
Meanwhile, Vaughn-Cooke is running a
study with healthy participants, which she’ll later repeat with others
who have been diagnosed with depression, prompting them with questions
like “How was your day?” and “What was the saddest part of your day?”
The responses are recorded with both video and audio. The former will be
analyzed for emotion using facial recognition software, while the
latter will be sifted through to identify vocal patterns as well as
transcribed and analyzed as text.
That’s where Resnik
will come in. He has already been looking for “signals in language use
that help produce insight into people’s mental health status,” he tells Newsweek.
In other words, his goal is to connect speech or writing, whether it’s
an essay or a tweet, to something that can identify problems with a
person’s mental health—something like the Hamilton Depression Scale, for
example.
“Once [we start] understanding how all of these
different predictors relate to each other, we can then develop
algorithms to better predict when a depression patient is going into a
relapse,” Vaughn-Cooke says. This will “not only improve quality of life
but also reduce incidence of suicide, relapse and readmission to
treatment facility.”
Ideally, all this will be
streamlined into one app that would collect this information while a
person is going about his day, potentially asking him to record
responses to questions periodically in addition to continuous passive
tracking. The result will be a tool, a Siri on steroids, that can track
mental health outside of formal treatment or between therapy sessions.
“You want people to get the kind of attention they need when they need
it,” says Espy-Wilson.
Glen Coppersmith, a research
scientist at Johns Hopkins University’s Human Language Technology Center
of Excellence, has been studying Twitter in an effort to understand
just that. His work focuses on quantifying signals—whether from the text
itself or from the tweets’ metadata (e.g., the time and location of
tweets and degree of interaction)—that are relevant to mental health and
could potentially lead to intervention. He and his colleagues have
already
figured out how
to parse people’s Twitter feeds and identify whether they’ve been
diagnosed with depression, bipolar disorder, PTSD or seasonal affective
disorder.
In early November, Coppersmith will run a
weekend-long hackathon at Johns Hopkins. A few dozen attendees—computer
scientists, statisticians, psychologists and others—will work in groups
on a data set from Twitter, trying to learn more about subtle cues and
patterns that could cumulatively indicate something about users’ mental
health.
Some efforts have already made use of social
media to track mental health. The Durkheim Project, run by a team from
Dartmouth University and the U.S. Department of Veteran Affairs,
launched in July of last year. The goal of the project is to analyze
data from the social media accounts and mobile phones of veterans who
opted in to try to find ways to predict suicide risk. Until recently,
“most of the available signals about mental health state through
language were not accessible—[like] the conversation by the watercooler
at work or at the dinner table at home,” says Resnik.
But
now that many of our interactions happen in public forums, like
Twitter, “we’re starting to have data that’s relevant that we can derive
insights from, that we can build technology on,” says Coppersmith.
One
company, Ginger.io, has released an app that uses phone sensors to
track mental health in patients who are participating at the suggestion
of one of their health care providers. The app is built on the
hypothesis that the ways people use their phones can provide important
information about mental health. It runs in the background and builds a
model of its owner’s behavior patterns.
It can then “notice” and flag any behavioral patterns identified in the Diagnostic and Statistical Manual of Mental Disorders
as potential markers of depression. An increase in missed calls or
texts, for example, could indicate reduced social interaction. Changes
in sleep cycles often accompany depression and bipolar disorder, so the
app uses sensors and activity to track sleep. GPS and other
movement-tracking sensors can be analyzed for patterns that might
indicate lethargy.
“None are perfect diagnostics,” says
co-founder and CEO Anmol Madan, but when risk factors are picked up by
the app, it can “send them a message, coaching, a question or set up
interaction with a nurse or provider.” It’s a “first level of triage,”
he says, but the intervention, communication and support patients need
is still done by people, psychologists and other mental health
professionals.
Closing the circuit between the
technology and the professionals and larger health care system is part
of the challenge, Stacie Vilendrer, an M.D./MBA candidate at Stanford
University who has
studied ways to harness technology to benefit patients, tells
Newsweek.
Beyond the difficulty of asking longtime clinicians to change their
ways, there are concerns about patient privacy and the potential for
liability. “For example, if an app collects information that a patient
is suicidal, dumps this information into a portal that the physician has
access to, and the patient commits suicide without any physician
intervention, it would seem the physician is still liable,” Vilendrer
says.
It can also be a disheartening challenge for
researchers and entrepreneurs to find ways to work within the U.S.’s
complicated health care system, she tells Newsweek.
But
“mental health is something that has touched every single one of us at
some point in our lives,” whether it’s a personal experience or watching
family or friends go through it, says Coppersmith, who predicts that a
spike in research activity around technology and mental health is
coming. “I don’t know how you can’t attack this problem. This is the one
everyone should care about.”
