Digitizing the Mind to Understand Mental Health

Eric J. Topol, MD: Hello. I am Eric Topol, editor-in-chief of Medscape. I am delighted today to welcome Tom Insel, a leading neuroscientist and psychiatrist, who has had a remarkable career trajectory in changing the ways we approach the mind.

Tom, how did you get into the area of behavioral science?

Tom Insel, MD: I was already interested in neuroscience and the biology of the mind when I started medical school. Psychiatry seemed like a great way to pursue that interest, but I got there too early; I think I was about 20 years before my time. There just weren't the tools in the late 1970s, early 1980s, to make that bridge between brain and mind.

I dropped out of psychiatry after about 5 years of clinical research. I had hit the wall on what I could do and where I could go. So I moved into neuroscience, retrained, retooled, and took a sabbatical at Johns Hopkins. Then I built my own lab at the National Institute of Mental Health (NIMH).

At that point in time, no one was interested in the biology of complex social behaviors. In fact, everyone told me I was making a huge mistake and would never get anywhere in that field. We just lucked into finding oxytocin and vasopressin and then some interesting animal models: voles, monogamous mice, and a bunch of animals that no one else in molecular biology was studying at that time.

In those days, people had never heard of oxytocin and vasopressin, and the idea that there were systems in the brain that were important for connections between people was completely taboo.

Dr Topol: It is amazing to reflect on how that research on oxytocin and social behavior still goes on today. Papers are coming out all the time that advance the remarkable work you did. Tell us how your career developed from there.

Dr Insel: When I got into the neuroscience of complex behavior, I was working at the cellular/behavioral level, while the NIMH was shifting gears and moving into molecular biology. At that point, it was made clear to me that the work I was doing was not work that the institute was going to fund, and I was quickly and not politely let go. I was fired.

Then, oddly enough, 8 years later, after a stint at Emory University where I set up a National Science Foundation center and a number of other things, I came back as the director of the same institute. Elias Zerhouni, then the head of the National Institutes of Health, recruited me. He asked me what it was like coming back. I said, "I'm coming back the same way I left when I was fired: with enthusiasm."

After being away from psychiatry for 20 years, neuroscience had grown a lot. Now the opportunity was there, with the revolution in genomics and imaging; our increasing understanding of neural circuits; and our growing ability to study mouse, monkey, and human brains. I thought this could be the time to work on that original mission of building that bridge between brain and mind.

It took 13 years to "move the cheese" within NIMH, to try to get people thinking about this in a deep way. It was a spectacular period; we saw so much happen.

Dr Topol: I watched the influence you had with respect to emphasizing genomics and taking a whole new approach to mental health. You challenged the psychiatric community on the Diagnostic and Statistical Manual of Mental Disorders, fifth edition (DSM-5). Why was it off-track?

Dr Insel: A real problem for psychiatry has been the lack of biomarkers, and our diagnostics have been built upon mostly subjective reports. The DSM is essentially a consensus document, with master clinicians getting together and voting on what criteria we should use as classifiers for major depressive disorder, or for posttraumatic stress disorder (PTSD), and so on. This process never included any biology or any kind of objective measures other than the consensus.

But I have to say, it worked quite well. I was around before we had a DSM, so I saw that this at least gave us a dictionary. It gave us a common language in which every term was defined, and that was a huge kind of progress. I was part of that original process when we formulated the DSM in 1983, and it transformed the field, but by 2010 we should have been able to do a little better than that. My concern was that we were putting out yet another edition of the same manual without having changed the paradigm at all.

Dr Topol: Your idea didn't go over well.

Dr Insel: It was definitely contentious. To be clear, I didn't have anything better to offer. All I said was, I get it: You need the money. It is an American Psychiatric Association publication, and I get that there is value in publishing it. It is like any other product. You have to revise it every now and then, similar to getting a new iPhone®. This is a great way to bring in some additional revenue.

But let's be honest about it. In terms of where we are 25 years later, we still do not have the data we need to be able to do this in a more objective, scientific way. Why don't we collect those data?

In 2008, we set up the Research Domain Criteria project, which was not a replacement for the DSM. It is a framework for research that tells us what kind of data we will need if we want to revise the way we do diagnostics. By the way, until we fix the diagnostic system, we probably will not be able to fix the therapeutics. We have to get this right. That led to this great debate within the field. Ultimately, I was reassured; most people have agreed that this is a problem.

Dr Topol: Over time, I believe there has been wide acceptance. You made the right call there. Now the question is how to get there.

Eventually you said, "I've done enough at the NIH." You went to Verily, and now you have moved on to run Mindstrong Health. Tell us about this current phase of your career.

Dr Insel: I came in with the bias that we could fix the diagnostic problem with genomics and imaging. I was wrong. We spent a lot of money on both of those efforts, and at the end of the day we found only a little evidence. It was not the transformative technology that we had hoped for, in the way that it had been for heart disease, for some neurologic diseases, and for other areas of medicine. We just weren't getting the specificity and the clarity that we thought we would get.

It turns out that those signals were no less complex than what we were seeing in the patients in the clinic. That is when I began to realize that we first need to get the phenotype right. Perhaps we are not even ready to know what genotype to look for because we don't yet have the behavior or the cognition right.

There has also been a parallel revolution in cognitive science; I want to include that with the sensor data and the information we can get with smartphones to get a real picture of what ecological behavior looks like in a continuous way, rather than seeing people once a month and having them fill out a form, which doesn't work very well. Getting real data in the real world is possible.

I have always said that science often progresses because of better tools, but the better tool for us was not, as it turned out, the 7T MRI machine. It was the smartphone, which every one of us has in our pockets. These are powerful computers that collect an enormous amount of information about us.

We can use these smartphones and ask, What are the signals that detect the beginning of mania, the beginning of depression, the beginning of psychosis? Perhaps this is how we can begin to fix the diagnostic pathway and get much better outcomes.

Dr Topol: Early on, there were such companies as Ginger.io, which looks at what you are texting, what you are voice-sending, and how much physical activity you have. Now you have taken that potential smartphone hub of data to a whole new level by trying to understand the keyboard interactions with the person. Can you tell us about that?

Dr Insel: We call this "digital phenotyping." It is a good way to describe three data areas. First are the things that Ginger.io and many other companies and a lot of academics have worked on: the sensors. Sensors would include the accelerometer, global positioning systems, and metadata about social and networking kinds of interactions. That is very interesting, but probably not that specific. It doesn't look that robust.

The second super-interesting area captures voice and speech. This offers profound insights about Parkinson disease, early dementia, and aspects of depression and mania, all of which confess themselves through changes in voice and speech.

Mindstrong is focused on a third area, which I don't believe anyone else has executed in quite the same way, and that is human/computer interactions: how we actually attack the keyboard, and the latency between hitting the space bar and hitting a character, or going delete-delete! or any of those things that we do on a keyboard, and how that changes over time.

Because there are so many data about this every time we use our smartphones, this is beginning to give us critical insights into cognition, mood, and other aspects of behavior, in the real world, in a continuous way that is highly objective, and it may be actionable. This may turn out to give us a set of biomarkers.

Dr Topol: It is extraordinary. There are many ways to digitize a person's mind, so to speak, whereby you could look at vital signs, facial recognition, or change in expression. How do you think this will evolve? Are we going to need a multimodal strategy to try to understand a person, particularly if the person has potential concerns about mental health?

Dr Insel: I believe it will be multimodal. What we have seen on the biomarker diagnostic side and on the intervention side has been a series of silos. No one has tried to bring them all together.

In this way, we could be providing interventions continually, whether through psychotherapy, peer support, crisis intervention, or all the things you can do on a phone. At the same time, we could make the assessment, and close that loop and figure out what's working and what isn't, and how to titrate the dose of the therapy. The opportunities are enormous.

Dr Topol: What is extraordinary to me is that we have such a large mental health burden, yet we have no way to ante up with the people who have the training and the capabilities to respond. It seems that we need to rally to get these other tools to help.

Depression is the number one form of disability in our society. Do you think that ultimately, not only through the digital phenotyping you have been describing but also with chat bots or other types of technology, that we can begin to address this very poor match-up of needs and support?

Dr Insel: That is a great question. I believe we have to be honest and say we don't know. There is a lot of hype about what we are doing with digital tools, and it is cool to talk about it, but whether it will work in the real world of healthcare is another matter. For example, we have some pretty good evidence that when people get depressed, they stop charging their mobile phones. Good luck there. That won't help with either the assessment or the intervention side.

We must prove this approach. I believe it is an urgent question because the burden is high and the needs are extraordinary, and no one likes the system we have now. No one wants the DSM, if you want to think of that as the problem, or the medications, or the therapies. People do not like what we are currently selling.

We need to come up with something that's better and perhaps something that's actually more user-friendly, where people have more agency. That could be super-interesting, but we have to remember that unlike diabetes, heart disease, and so many other areas of chronic illness, depression is a very tough customer because a fundamental aspect of depression is that you are helpless and hopeless. It gets in the way of you actually doing the things you need to do to get better. It's that catch-22 that is very tough to solve.

Dr Topol: That is a critical point, and no less the fact that the mood and state of mind interact with all the other conditions of man.

One surprise in recent years was the idea that you could talk to an avatar or a machine, and you would actually feel more at ease revealing your innermost secrets than if you had talked to another human. Do you believe that?

Dr Insel: I believe we are still learning about that. All of us are already dealing with bots. Even when we don't know about it, we are dealing with bots much of the time. I believe we will come to find it more acceptable. If there is one area of medicine where you would assume that the human interaction is going to be fundamental, it is psychiatry.

You would think that people need to talk to another person, but I hasten to point out that I was trained as a psychoanalyst, and what we were taught to do was to become a blank screen: a bot, if you will. We would have people sit on the couch not facing us, and we were told never to say anything that would reveal anything about ourselves, so that the patient could develop a whole series of fantasies and projections about us that then would become the basis of the therapy. Maybe that is what we will end up doing.

Dr Topol: That parallel is fascinating.

Dr Insel: In a very interesting way, we now have avatars that really are a blank screen. They are what Freud always wanted to create. The other thing about them, which is extraordinary, is that they can learn, and they do not have any built-in bias, if they are trained in the right way.

For example, if we could develop a tool that would help us give volunteers on the suicide hotline the skills of a master clinician for detecting suicidal risk, it would be fantastic. You could scale this up. You could do it in real time.

Imagine interpreting the voice and speech of people who call in and giving a red light, green light, or yellow light to college students who are working on the line, to cue their responses, giving them the skills in the same way that Waze (the community-based navigation app) helps a new driver in the city navigate as a master cabby would have done it in years past.

When we think about bots and artificial intelligence, and we think about the tools we could build, I believe we would use them not so much to replace people, but to replace tasks and to upskill people—giving them the skills similar to those of a master clinician without the training that goes into that. The Waze model for clinical care is pretty exciting, and we could do it. Even in the short term, I believe the technology at this point is ready to go.

Dr Topol: I like that Waze model, and I believe the insights you bring are quite extraordinary. It has been a delight to talk to you about this and hearing your optimism that some of these tech tools will help support a better model over the years ahead, never supplanting the importance of human-to-human interactions but hopefully making it a lot better. It is certainly a big issue that needs to be addressed.