An Odd Couple in Interdisciplinary Research

An Odd Couple in Interdisciplinary Research

with Linda Collins and Daniel Rivera

August 30, 2010

HostMichael Clevelandinterviews an odd couple in interdisciplinary research focused on optimizing behavioral interventions. Social scientistLinda Collinsfrom Penn State and chemical engineerDaniel Riverafrom Arizona State talk about their NIH Roadmap Initiative project.

Speaker 1: Methodology Minutes is brought to you by the Methodology Center at Penn State. Your source for cutting edge research methodology, in the social, behavioral, and health sciences.

Michael: Hello, and welcome to this episode of Methodology Minutes. Today, we are interviewing an odd couple – Linda Collins and Daniel Rivera. Linda is a professor of human development and family studies, and the director of the Methodology Center here at Penn State, and Danielle is professor of chemical engineering in the school of mechanical, airspace, chemical, and materials engineering at Arizona State University. Quite a mouthful. Daniel and Linda are collaborating on a research project funded by the NIH Roadmap Initiative. And welcome to you both.

Linda: Thank you.

Michael: I'd like to start just by some personal introductions. Daniel, do you want to start?

Daniel: Okay, yes. Well, I am a chemical engineer by training. First job out of graduate school was working for Shell Development Company, years back. And my main interests really to what is referred to as control systems engineering, which is actually within the engineering a multidisciplinary field. There is aspects of control systems that are seen in pretty much any engineering field, and the idea of the field or the main emphasis of the field is in looking at how systems change over time, and then coming up with mechanisms, controllers, that are then are able to do changes on a processor system to then improve the way in which they perform. So, again, this is a very broadly applicable type concept. And initially, when I got my job at Arizona State, I was looking at a lot of chemical process type related problems. And then, around the late nineties, got interested in what would be termed as nontraditional problems within control systems.

Michael: Okay.

Daniel: Looking at supply chain management and how inventories could be managed and supply chains, using, again, control engineering ideas, and that sort of led into some of the problems that we're looking now in behavioral health. And I guess we will talk a little bit more about that in the course of this podcast so.

Michael: Right.

Daniel: Yeah, so basically, yeah, have been twenty years at ASU, and it's been seven years since I've met Linda, and began my journey that has now led to not only the R21 project that we're going to be discussing, but also K25 from NIDA in looking at control engineering and related approaches to improving behavioral interactions.

Michael: Thank you. I'm happy to have you here, and welcome. Linda, maybe you want to tell the listeners a little bit about yourself?

Linda: Okay. Well, I am as psychologist by training. My undergraduate degree and my PhD are in psychology. My PhD is in quantitative psychology, which means that I am trained in research methods, statistics and measurement, as it's applied in psychology. In the graduate program I went to, there was a heavy emphasis on measurement. In fact, a lot of people from that program went to work at places like the Educational Testing Service, the organization that develops exams, like the SATs and the GREs.

Michael: Sure.

Linda: So that's the kind of training that I got in grad school. But very soon, I became interested in analysis of longitudinal data. And the first ... well I'm actually still working in that area and then, about maybe seven or eight years ago, I started to get interested in ... more in research design, in particular, designing research that could be used to improve behavioral interventions. Now, that has led into thinking more and more like an engineer. Actually engineering behavioral interventions, in much the way that somebody might engineer any other kind of a product. And so, the work I've been doing has been inspired by the thinking of engineers, although I'm just an amateur engineer, because I'm not trained in that way at all, but I started learning about this from Veejay Nyer at the University of Michigan. He is a terrific and terrifically inspiring engineering statistician. And then more recently, I started working with Daniel.

Michael: So, I have to admit that a psychologist and a chemical engineer working together does seem a little unlikely, at the surface, but as you both just described, it sounds like you had a natural progression towards this collaboration. Can you maybe talk a little bit about how the two of you actually came to collaborate together?

Daniel: So the way that I learned about the work that was going on here at Penn State was at Susan Murphy at Michigan, who was also one of the CO-PI's of the center, was at a workshop on data centric modeling and control in late December 2002, University of Minnesota. And she spoke about problems related to adaptive interventions, and mentioned that there were a lot of opportunities in that area, and so I spoke to her and she said you know you should talk to Linda. Because at that time, actually, there was a paper that appeared in 2004 on prevention science, which at that time had been I guess approved for publication. It was in press.

Linda: Right.

Daniel: On a conceptual basis for adaptive preventive interventions, so that sort of became my primer.

Michael: Okay.

Daniel: From my standpoint, I was seeing a problem where you are now deciding on individualizing prevention components over time, using failing variables in your ... again you have a time varying system for which you are then using decision policies to optimize or to improve outcomes. So this all was clearly or at least I could see conceptually the connections with what we do in control systems engineering.

Linda: Daniel's thinking has been so influential for Susan and me. I think a big light bulb lit over my head one time, although it took a while, but I saw, after working with Daniel for a couple of years, that people evaluating behavioral interventions are always thinking in terms of comparing a treatment and a control. And that is one way to look at it.

Michael: Sure.

Linda: But any engineer, instead, establishes a criterion, and tries to bring people up to that criterion. That's a very different way of thinking about it, and I think it's a much clearer way of thinking about it.

Michael: Can we step back, maybe, and talk a little bit more in general about what this would be, which is interdisciplinary research? I'm assuming that both of you have had other types of collaborations that are interdisciplinary. Can you talk about the advantages that you see to this type of research?

Daniel: Yeah, in my case, I mentioned earlier how I had already sort of branched out from sort of a traditional chemical process control. What I saw as extremely beneficial, in particular going into work that involved behavior psychology and other fields that are outside the domain of whatever strict engineering was just the opportunity to really work on a new class of problems in a very open field. But, in terms of advantages, as you know you're being able to work on interesting problems, social significance, that the field is not too crowded and you can really make a good inroad. So certainly, when this opportunity came up, the fact that I'm one of the few people doing this kind of work was very appealing. And then of course the fact that Linda, Susan, the center, just a great infrastructure here for doing this kind of research and very open, that's often an important factor too that great people to work with.

Michael: Sure. Right, right. Linda, do you want anything?

Linda: Well, I think like Daniel, I was at a point in my career where I was getting a little tired of what I was doing and I was ready for something new. And to me, a big advantage of interdisciplinary research is that you learn so much. I just feel like I have learned so much working with Daniel. And of course, as I said before, I don't have any training as an engineer, so it's been just kind of learning a little bit as I go, and I realize that I've just barely scratched the surface of engineering. But what I've learned I guess it's not so much specific facts about engineering, but more as I said before, a way of thinking about problems, that I like. And as it turns out, it's very natural for me.

Michael: Sure.

Linda: So I've enjoyed that part of it very much, and I feel also that, I feel excited about the opportunities to improve behavioral interventions, because now that we've made some progress over the last seven years, I'm seeing that this isn't just a pipe dream. That we really can improve behavioral interventions. That we can develop a new generation of behavioral interventions across a wide spectrum of health areas that can have a much larger public health impact.

Michael: Right.

Linda: And so that, to me, is very exciting.

Michael: Right, I agree. What about the flip side of that question? What are some of the disadvantages or the drawbacks of combining engineering and psychology? It's got to be hard ... on many levels

Daniel: I'd say the biggest drawback in a way is sort of an unavoidable thing, I guess it's sort of a benefit in a way, and that's confusion. I mean, that just comes with the territory, because you're really sort of going outside of a comfort zone, you're going outside the defined silos of knowledge, and so certainly, you have to go out on a limb in terms of learning new things and being able to understand the mindset of different fields, and that can be very challenging.

Michael: Sure.

Daniel: But then it wouldn't be interesting research otherwise. It wouldn't be interesting interdisciplinary research that were not the case.

Michael: Mm-hmm.

Daniel: And I would say that in my case, I happen to be fortunate that I have a K25 from NIDA that includes a training program.

Michael: Mm-hmm

Daniel: So I ... and there's a very good quantitative psychology program at ASU, so, I think I've sat through maybe seven, eight different courses.

Linda: Mm-hmm.

Daniel: And then I've also been participating in a postdoc seminar that Lori Chasson runs.

Michael: Sure.

Daniel: So I've had these structured experiences-

Michael: Right.

Daniel: That have helped me become better acquainted with methods.

Michael: Let's talk about the actual project that you were collaborating on. Just beginning first with the acknowledgment that this project is funded by the NIH Roadmap Initiative. Can you talk a little bit about that initiative in general first?

Linda: The NIH Roadmap Initiative is an initiative that spans all of the centers and institutes within NIH. It's designed not to be affiliated with any one center or institute. And whatever work you do for roadmap project has to affect public health broadly.

A few years ago, there was a research funding announcement from the roadmap, for development of methodology for the behavioral sciences. And that's what Daniel and I applied to. We applied for funds to develop, over the course of four years, our thinking about using methods from engineering control to optimize behavioral interventions. And the behavioral interventions could be in any public health area.

Michael: Can you go a little bit more into detail about what this project entails?

Daniel: Sure, yeah. The main idea was to look into a diverse series of interventions and see to what extent the dynamical systems and control theory could be relevant to these interventions. So what we did is that we proposed a panel of behavioral experts with whom we would meet, and of course for the first two years of the project, to take one of the interventions that had been working on and through a structured interview process, determine how it could be tasked as a dynamical system that would then be amenable to some kind of optimization. And in the process of the discussions with them, we have been ... so again in going through this process of describing the interventions as dynamical systems, and then we've been getting ... really been doing a couple of things. One is in some cases developing interventions, like in the case of the discussions with Danielle and LeAnn here. They interventions that they developed. So we've done some simulation work in support of that. And then we're talking now actually of doing some further work with that. And then in some of the other cases, we are getting secondary data sets that we are analyzing.

For instance, in the case of Jarred Younger from Stanford school of medicine. He's got a pilot study. He's got two studies related to use of Naltrexone and fibromyalgia patients, that involve intensive longitudinal data so we've been analyzing those.

So, I think that that was the idea. First, being able to meet with this panel to look at these diverse problems and now and those dynamical systems. And then in the course of the latter two years of the project, then being able to actually try out some modeling approaches that would give us models for which we could then say, let's use these models to develop and optimize intervention.

Michael: Okay.

Linda: So the idea is that first, you need to be able to express the intervention as a dynamical system.

Michael Yes.

Linda: And then once that is done, it is potentially possible to control the system.

Michael: Okay. So Linda, as a follow-up question to that, how do these relate to or correspond to what are called adaptive interventions? Are they along the same line, am I thinking correctly here?

Linda: Yeah, you definitely are thinking correctly. So, adaptive interventions, most generally, are interventions that can be adapted to characteristics of the individual or the environment.