Mobile Data and the Search for the 'Holy Grail' in Obesity Research

; Donna Spruijt-Metz, PhD


January 26, 2016

This feature requires the newest version of Flash. You can download it here.

Editor's Note: In this One-on-One interview, Medscape Editor-in-Chief Eric J. Topol, MD, talks to Donna Spruijt-Metz, PhD, an expert in pediatric obesity, about how mobile data collection and wearable sensor technology can potentially reduce the epidemic of "diabesity" in the United States.

Searching for Better Data Collection

Eric J. Topol, MD: Hi. I'm Eric Topol, editor-in-chief of Medscape, and I'm very pleased to welcome Donna Spruijt-Metz from the University of Southern California. We're going to be talking about her program in mobile health and obesity. Donna, tell us a little bit about your background and how you got into working in mobile connected health.

Donna Spruijt-Metz, PhD: My background is in obesity research, predominantly childhood obesity. A long time ago, before the term "mobile health" existed, I was doing an in-lab study where I brought overweight Hispanic and African American kids into our lab, where they had 2 days of after-school exercise, nutrition classes, and motivational interviewing. We collected data like good researchers, pre- and post-intervention. As this intervention unrolled and I was on the ground, in the trenches with these kids, I realized that we have all of these data on them that we could be using, but we don't even look at them until the trial is over.

Not only that. I was talking to one of the kids for the motivational interview—and we're doing a randomized controlled trial, so everybody has to get the same treatment—and I knew that she was about to head into a full hour of talking about sugar-sweetened beverages and why they're bad. I had just spoken with her and knew that she doesn't drink them. She had other vices, but not sugar-sweetened beverages. I thought to myself, we're going to lose this kid and we're not using the data that we have. And there weren't any accelerometers in the first and last weeks. We could be having that data all the time. There's got to be a better way, and that's what got me started.

Dr Topol: So that was the impetus to bring in mobile data collection from the real world in these kids representing minorities?

Dr Spruijt-Metz: Right.

Real-Time Interventions

Dr Topol: What have you learned from this?

Dr Spruijt-Metz: The kids that I work with love the real-time attention. When we intervene using real-time new technologies, they are in communication with us. One of my real-time interventions is what we call Just-In-Time Adaptive Intervention. "Just in time" because it's in context, which is so important. It's when you need it. We are intervening in these kids' sedentary behaviors—not having them come in and telling them, "You must be less sedentary," but sensing that they've been sedentary for a certain amount of time and texting them to let them know that it's time to get moving.

Dr Topol: Do you have to give them a smartphone?

Dr Spruijt-Metz: It depends. For some of the research that we're doing now, we use our own phones, so they have to have an Android phone, for instance, to be in a particular study. For others, we give them phones. We also have deployable sensors in the home to try to understand their behavior in the context of the family.

The Holy Grail of Obesity Research

Dr Topol: Recently there was a study that you may have seen in Cell Metabolism,[1] where the Salk Institute in La Jolla made a smartphone app that helps to collect information on all of the food that one eats during the day. They found that people eat all day long, not just in three meals, and particularly in the evening. Wouldn't it be nice if we had a way to passively detect what people are ingesting?

Dr Spruijt-Metz: That is, in obesity research, the wicked question.

Dr Topol: It's the Holy Grail that's missing?

Dr Spruijt-Metz: It's the Holy Grail. There are a number of different efforts around that. There are two best measures that we have. There is doubly labeled water: You drink isotope-labeled water and then give me all of your pee for the next 10 days. That's fun and kids love it—no. Also, it gives us energy in and energy out, but we don't know what they're eating.

Right after that gold standard is 3-day dietary recall, which is unreliable at best. It's not so much that people lie, but they don't remember what they ate. They don't know what was in the food. Even if we call them with the best dietitian support for 3 days, they're not reliable. It's downhill from there.

In a couple of labs in Hawaii and Pittsburgh, they have been trying to do photo capture and then machine learning of recognition of the food on the plate. I think it's a rabbit hole.

Dr Topol: It's a tough one, isn't it? Well, let's hope that someday that technology will evolve. I've seen some things like trying to look at the conductance of tissues. People are definitely trying it, but as we say, this is a difficult feature.

Dr Spruijt-Metz: Well, the question is, I'm beginning to think, do we need to know? To understand the metabolism, it's important to know what people eat—exactly what they eat. Then you might want to do a controlled feeding study. But to understand and change behavior, we might not need to know that. I don't know. My jury is still out.

The Challenge of Changing Behavior

Dr Topol: I did want to get into that with you, Donna, about changing behavior, the ultimate challenge. The hope is that, through mobile connectivity, there would be feedback loops and a new way to intervene. Not just getting data on things like activity; you would also have the ability to gamify it, to have managed competitions with peers and neighbors, Facebook friends, whatever, and you would be able to incentivize it, whether it's with finances or other means. Do you think that we're going to be able to crack the case? Obviously, obesity is such a dominant problem and it's horrible to see the rate in children. Could this be a way, with these other tools that are on top of smartphone connectivity, to finally start to tackle the behavioral issue?

Dr Spruijt-Metz: I think that only a smartphone will never be enough. It's just a gadget.

It needs to be smart. It needs to have some good artificial and real intelligence behind it. I think that it'll always be a combination of smartphones and sensors, either wearable or deployable, so that we know what people are doing and can adapt in real-time to their personal needs. It's personalized. Precision behavioral medicine is what we're getting, what we can do now in real-time with these technologies. Do I think gamification will help? I don't know. I've developed some games. I've worked in that field.

Dr Topol: Especially with kids, you would think.

Dr Spruijt-Metz: Developing a good game is very difficult. It has to be darn slick before it's going to work. Even Angry Birds—at some point, people get tired of it. I think gamification has something else to offer that's very cool, though, which is that you can use it as a tool to teach things to kids without having to test them.

Dr Topol: Or feeling that they're having to go through educational maneuvers.

Dr Spruijt-Metz: You can just watch what they're doing from the back end, and once you sort that out from game mechanics and how well they play it, you can see where they're struggling and where they're not struggling.

Dr Topol: What about managed competition with other kids and their peers? Is that a strategy that you're using?

Dr Spruijt-Metz: I'm not sure that competition is always the way to go. I think it's good for some people some of the time. [Northwestern University professor of preventive medicine] Bonnie Spring has done some really great work with adults where it's anonymized. You have teams, but you don't really know who is on the team. I think it can work for kids. It has to be done really well.

Dr Topol: It hasn't been tested yet, right?

Dr Spruijt-Metz: Well, you can't test it. When you start talking about mobile health and embedding games and competitions, it's all in how you do it and what other things are surrounding it. If I make a terrible intervention and make competitions out of it, and then it doesn't seem to help, then is competition no good or the other elements? That's where some of [Pennsylvania State University researcher] Linda Collins's work, looking at fractional factorial designs of what works and what doesn't within the context of a larger intervention, is important. So I can't give you a good answer for that.

Capturing Environmental Factors

Dr Topol: What about the other factors that are not necessarily captured well today? We talked about the actual ingestion of food, but what about sleep and other environmental factors that we don't do so well at quantifying or even capturing at any level? What do you think about those?

Dr Spruijt-Metz: I think we're getting there with sleep. I like the Lullaby system. I love their work. I like the deployable systems. I think that some of the sensors that are combined—sensors in one's room and a wrist-worn sensor—can work. It's not going to give you all of the things that you can get with deep sleep monitoring, but I think it's going to give you enough to know where people fail and what's disrupting their sleep. What I really like about the new sensing modalities for sleeping is that they take context into account.

Dr Topol: I've seen one that's pretty extraordinary because it compares favorably with polysomnography. The ones that are under the mattress and just detecting gross movements are not necessarily very accurate.

But I think you're right. I think that's going to be something that could be integrated. One of the problems that we have today with mobile apps is that they're each developed as one-offs; we haven't gotten the master integrative dashboard and virtual coach and that sort of thing. Hopefully, over time, that's going to be the way that this evolves.

The Big Picture: Reversing the 'Diabesity' Epidemic

Dr Topol: Now let's talk about the bigger picture. We have a massive problem of diabesity, and I'm not sure whether the data indicate that we're making any headway. What's going to finally get our arms around this serious problem of epidemic proportion?

Dr Spruijt-Metz: I'm going to go back to what you said about sleep. It's an environment that we live in. We live in time and we live in context. That's what makes mobile technology so great, because it can capture all of that. I think we are beginning to get a handle on [diabesity] because it's not continuing to go up—although on the very, very tip of it, morbid obesity is going up. But we've sort of stopped the flow a little bit, stopped the bleeding. I do think that we can get a handle on it. If I didn't think that, I would do something else with my life.

Dr Topol: Have you been discouraged over the years?

Dr Spruijt-Metz: I'm always discouraged and I'm always encouraged. I think the tipping point is going to be understanding the interpersonal, interindividual variability—how I interact with you. We have a neat new study that the National Science Foundation just funded, using deployable sensors in the home to understand family interactions around eating in the families of obese children. It uses some simple beacons placed around the home—I can give you some examples—and a phone and wearable sensors that will be able to tell us eating behavior, only the eating episodes. I don't care what you're eating right now; I just want to know if you're eating. I want to know if you're eating, if you're stressed, and if you're angry. I want to know if you're alone. I want to know who you're with.

Dr Topol: Besides the sensors that are in the phone, do you have things on the refrigerator and the pantry?

Dr Spruijt-Metz: Yes.

Dr Topol: So anytime it's opened up, you're sensing that.

Dr Spruijt-Metz: And I know who it is.

Dr Topol: You have a unique identifier.

Dr Spruijt-Metz: Yes.

Dr Topol: So at least on some level, you're watching habits and behavioral things, right?

Dr Spruijt-Metz: Right. It's that whole cascade of interactions around eating. The funny thing is, to come back to the wicked question, we don't know yet what people are eating. We do know that they're eating too much sugar. In general, we know what's wrong with the American diet, but on the individual level, it's very difficult to know.

Dr Topol: What about the kid who says, "I don't want Dr Donna Spruijt-Metz to know what I'm eating, so I'm going to have my sister go to the pantry and the refrigerator to get the food and bring it to me and game the system"? Are we going to have that type of issue from sensors that are not granular enough?

Dr Spruijt-Metz: It is always possible to game the system. I am working with a couple of very smart engineers from the University of Virginia, and I'm pretty sure that they'll have some fail-safes.

Dr Topol: Okay, good. I think you're bringing up many central points here about the evolution of sensors as they get more and more refined, informative, and highly accurate. You've also brought up where the holes are, whether it's the context or the inability at this point to track actual food ingestion. You also provocatively mentioned that maybe it's not so important after all. We can get all of these other pieces in there.

Dr Spruijt-Metz: The reason that it might not be important is because, how well have our messages done about eating less sugar, substituting sugar, or eating less saturated fat? How well have we done with those messages? Not fantastically well.

We seem to stumble from one diet to another. I remember when we first immigrated back to the United States, [the message] was that low saturated fat was so important. No fat. Fat was bad. We were driving cross-country and we would buy a low-fat muffin. It's 900 calories of refined disgustingness.

Dr Topol: Yes, and then you get hungry a few minutes later.

Dr Spruijt-Metz: Right. We need to know the mechanisms and how the metabolism deals with those foods—that's something else entirely. But I don't know if those messages are helping us intervene. What we do know from behavior change and psychology is that stress during mealtime, for instance, will make many people eat too much, too fast, or both. We can certainly intervene on that now that we have these real-time technologies.

Limitations of Mobile Data Collection

Dr Topol: One other question is about the selection of kids into your studies. It isn't truly random, the ones who you can find that want to be in these studies, right? Are you selecting out a group that is potentially not representative?

Dr Spruijt-Metz: Yes.

Dr Topol: Is that a problem?

Dr Spruijt-Metz: Yes, but show me a research study that doesn't have that problem.

Dr Topol: Well, the only way that people have gotten around this is by getting to hundreds of thousands, if not millions, of people through passive data collection. That has the potential to override some of the concerns when you're doing smaller studies that require active participation. Just the fact that they're participating selects people who potentially have plasticity in their behavior. I just bring it up because I wanted to get your views.

Dr Spruijt-Metz: Yes, it's a problem.

Dr Topol: I understand fully that this is not easy, and it's far better than not studying it. It's always just one of those things to acknowledge as a potential limitation.

Dr Spruijt-Metz: It is a potential limitation. I serve the underserved. I work with predominantly low income African American and Hispanic families and kids. Do we get everyone that we try to recruit? No. It's very difficult.

Dr Topol: It's clinical research in academia.

Dr Spruijt-Metz: That is clinical research in academia. Do I see hope for changing that with precision medicine and the new cohorts coming down the pike? To some extent, mobile recruitment is going to help us, but for any study that requires intensive measurement, directed measurement, and informed consent, there's always going to be some kind of selection bias.

I feel really strongly that as much as we're all so excited about big data—I'm excited about big data— and we're all so excited about the data that's just lying around waiting for us to use it, like Fitbit data, it doesn't have directed data collection. It can't answer some of the most important scientific questions.

Dr Topol: I couldn't agree with you more. That's a very good point to finish on. No matter which way we go, there are always issues that we have to keep in the center of our thoughts, because there is no perfect way to assay what's really going on and to change it.

Dr Spruijt-Metz: Yes.

Dr Topol: Donna, thanks so much for joining us on Medscape One-on-One.

Dr Spruijt-Metz: My pleasure.

Dr Topol: Thanks to you. I'm glad we had the chance to explore such an important topic about childhood obesity and the use of mobile sensors as a potential way to understand it better and, in the future, intervene to try to get a better handle on what is a very serious public health problem. Thanks for joining us.


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.