Of the 24 Fellows honored today with the MacArthur Foundation’s 2013 “genius” awards, about half labor on the frontlines of science and technology. One of those is Susan Murphy, a 55-year-old statistician at the University of Michigan. In this interview she talks about her passion for using math to improve health care for people with mental illness. She and the other recipients earn a $625,000 prize to be paid out over five years (an increase from the $500,000 given in past years.) See a full list of winners here, and scroll to the bottom of this post for details about the other science researchers.
[An edited transcript of the interview follows.]
What is the greatest challenge currently facing us in the health care arena?
Achieving and maintaining behavioral change. You can have a great pill—a magic blue pill that will control your blood pressure—but if you don’t take that pill every single day as many times as required, your blood pressure will not be controlled. That’s behavior change. And it’s difficult to change. If you need to take that pill for a decade, that is a different thing entirely than taking it for a day or two.
Tell me about how your methods helped clinical trials compared with those that simply compare two treatments.
The idea is to allow you to look at the entire sequence of treatments [a person experiences in care]. We see how people do in [a] first treatment, and if they do well, we randomize them to different kinds of maintenance therapies; if they don’t do well, we randomize them to different kinds of secondary treatments—augmenting their current treatment or secondary treatments. Then you get all the people who didn’t do well on the first treatment or people who did do well on the first treatment and those groups are much more representative [of a population with a certain malady].
Our SMART [for sequential, multiple assignment, randomized trial] trials are now being applied all over the health field. They are listed on my Web site. Let’s think about a trial for little kids who have autism. These are kids who at age five are still nonverbal—this is a big problem because verbal ability is strongly related to success later in life. These kids have not responded well to past treatments. Initially they were randomized to two interventions designed to improve their ability to communicate. One is focused on using spoken language in play activities—something relevant to the child. The other would be focused on the kid using something like an iPad, where the kid could press it and [it would] speak for the child. Then what happened is the little kids were tracked over time, and they saw whether or not the interventions were improving their ability to communicate, either verbally or though the iPad. The iPad device is expensive, so normally you wouldn’t give it to the child initially. The kids in the study arm that were being encouraged to express themselves verbally in play, if those kids were still struggling to communicate, then that kid was rerandomized between using the iPad-like device or even more intensive treatment—seeing the therapist more frequently. The ultimate goal was to think more about what therapies you should start with.
Right now I know you're working on ways to deliver personalized health care via mobile devices. What's your approach there?
We are working on the innards—the algorithms that would be behind an app that you would see on your phone. You can imagine an app that would be a recovery app “coach” application—helping you deal with trying to stay off drugs, for example, or trying to deal with cravings that you feel as you are moving through life and you are trying to keep your weight down. The goal is to tailor whatever strategies the coach recommends to you; so, as time goes on [and] you change, we want the app to change with you.
Where's that research now?
These apps are already being tested, but they are not at a point where they are individualized. We are at the point where we are developing algorithms to better tailor care to a person. Apps for medical use are already out there. Strategies are there. The question is: How do we individualize it to the person? Another thing that is important here is it’s really difficult to change your behavior. The coach on the phone has to engage you and keep you motivated, so a lot of these strategies are about maintaining interest to stay the course with behavior change.
Walk me through the nuts and bolts here. You would input data about yourself and it would spit out text messages at specific intervals or something like that?
These algorithms, they work much like the algorithms that work on the Internet. When you go on the Internet and go to a big company site, they give you ads that have been personalized to you. How do they do that personalization? They explore (or randomize you between) different appropriate recommendations and they see which ones you pick. In our situation the little coach app would try different behavioral strategies—maybe ones with high probabilities that they would work for you. [With the] variety of behavioral strategies, it would record how well you do—for example keeping your weight off or staying off drugs. The behavioral coach might suggest you text a friend or indicate exercise that was helpful in the past. The coach might suggest you look at pictures of your family and children and think about why you have made this decision to stay off drugs.
So, what would you need to tell the "coach" then?
Now there are more and more sensors being worked into phones—activity sensors that record how active you are, and sensors can be built into phones that record how much social activity you use the phone for—so, there is a lot of information the phone records already, like with diabetes. Wireless sensors on your body could communicate with the phone. The goal is to have health care that doesn’t interfere with your life.
Why are you interested in mental illness and behavior change in particular?
People who have these difficult conditions, they have been stigmatized for years. I think that I should try to help the people who need my help most.
What do you plan to do with your award?
I’m going to use it mainly to support the development of these algorithms. It requires a lot of effort, a lot of programming and different people have to be involved: computer scientists, statisticians, clinical scientists. All that money is going to go toward that endeavor. That’s all I really want to do. What can I say?
As a woman working in mathematics, what do you think can be done to further encouarge women to enter into STEM careers?
I like math. It’s the only thing I’m really good at. I can’t be anything else. I think having really engaging junior high math teachers and potentially high school math teachers is key. Certainly it was for me with one of my teachers. I still remember her name; I named a number of my pets after her. That teacher doesn’t need to be female necessarily, but he or she needs to allow teenagers to experience how much fun math can be.
You work at the University of Michigan. Do you have any thoughts on what statistical applications could be employed to help the people of Detroit?
Poverty can be viewed as a chronic, relapsing disorder that requires a sequence of interventions to help people get out of poverty and stay out. So from my point of view, helping people get out of poverty is just as important as helping people manage their diabetes or helping people improve their recovery in substance abuse. Poverty is a similar type of problem in that people move in and out of poverty repeatedly over time. Welfare policies can often be viewed as a sequence of treatments, so we need to think about how best to sequence those treatments to help people stay out of poverty and move into the middle class.
Other science researchers receiving 2013 MacArthur grants:
- Sara Seager, a Massachusetts Institute of Technology astrophysicist working on a theoretical framework for determining the characteristics of planets beyond our solar system
- Angela Duckworth, a University of Pennsylvania psychologist, studying grit and self-control’s place in educational achievement
- C. Kevin Boyce, a paleobotanist at Stanford University exploring the connection between the remains of ancient plants and present-day ecosystems
- Phil Baran, an organic chemist at The Scripps Research Institute in La Jolla, Calif., re-creating natural products in the lab for pharmaceutical applications.
- Jeffrey Brenner, a primary care physician in Camden, N.J.
- Craig Fennie, a materials scientist at Cornell University
- Carl Haber, an audio preservationist at Lawrence Berkeley National Laboratory
- Dina Katabi, an M.I.T. computer scientist examining wireless networks
- Julie Livingston, a public health historian and anthropologist at Rutgers University
- David Lobell, an agricultural ecologist at Stanford
- Sheila Nirenberg, a neuroscientist at Cornell working on prosthetics
- Ana Maria Rey, an atomic physicist at the University of Colorado at Boulder