This is the transcript of my recent podcast interview with MedNetworks CEO Larry Miller.
David E. Williams: This is David Williams, co-founder of MedPharma Partners and author of the Health Business Blog. I’m speaking today with Larry Miller. He is CEO of MedNetworks, a social network analytics company in health care. Larry, thanks for speaking with me today.
Larry Miller: It’s a pleasure David.
Williams: You had a big summit meeting yesterday in Cambridge, MA. How did that go over?
Miller: It went very well. I’m a little biased of course, but we had a full house, as many people as the fire code would allow, which is always a good sign. People I talked to thought the program was quite good, with the right mix of social network science and applications.
Williams: Tell me about the keynote discussions.
Miller: There were three keynotes: Juan Enriquez, James Fowler and Nicholas Christakis. Juan spoke very broadly about changes in what he calls code; moving from silicon based code to life sciences based code. James and Nicholas spoke more about social network science. James in particular traced the progress in the field over the last ten years and specifically the move towards understanding the genetics of social networks, which has exploded just in the last couple of years.
Nicholas talked about social networks in general, and then specifically about how social networks are “centers,” and how one can use social network effects to begin to forecast behavior. It’s a very exciting application that’s just begun to spread.
Williams: When people talk about social networks these days, the first thing that springs to mind is Facebook. But I understand that you’re quite a bit broader than that, looking at both the offline world and the online world. How do these worlds differ?
Miller: Just to be clear, we are not a Facebook or a social network site per se. We are mostly interested in mapping and analyzing existing social networks. We can do that with offline networks or online networks. We can work with Facebook and other social sites, we can work with cell phone traffic. Our primary business is looking at existing data rather than trying to invite people into networks.
Williams: What kind of customers do you work with?
Miller: There are four types. We work with pharmaceutical companies to help them understand and be more efficient in their promotions. We work with health plans in a number of areas. Probably the most important right now is as they begin to form Accountable Care Organizations and Medical Homes. We work with hospitals or hospital groups as they try to understand the referral patterns in their areas. Finally, we work with large corporations to help them understand the social networks of their employees. That’s not so much for health issues; it’s much more related to leadership and innovation, more organizational issues.
Williams: Let’s talk a little bit about Accountable Care Organizations and Medical Homes. What role do you play there?
Miller: We’ve worked in a couple of different areas. Different health plans are at different stages in formation of their organizations. Primarily we’ve been helping them form so-called ACOs.
It’s a fairly simple proposition. If you’re trying to form a group of physicians and other providers to care of a set population of patients and eventually to take some of the risk involved in caring for that population, how do you do it? How do you decide which physicians and providers form one unit?
You can just see who raises their hand or you can do it arbitrarily by geography. But what we suggested is to look at existing practice patterns. We’re very good at mapping physician networks, community by community, so we help them look at those community practice patterns, look at those networks and choose physicians who are already working together.
Williams: Would the customer for that service be a health plan or do the provider organizations themselves sign up with you?
Miller: That’s a good question. It depends a little on the part of the country. In most parts of the country, these so called ACOs are forming around health plans or in some cases hospitals. In a few areas, California being one of them, the provider groups are leading the charge. In California we’re working with providers and elsewhere we’re working primarily with health plans.
Williams: One of the areas of interest in network analysis in health care is the diffusion of knowledge among physicians or providers more generally. I’m thinking today about things like evidence based guidelines or quality measures. Do you play a role in analyzing or facilitating diffusion?
Miller: We do. In fact one of the things that networks are probably best at is diffusion of innovation. Yesterday at the conference both James Fowler and Nicholas Christakis made a critical point about networks, which is that they magnify whatever they’re seeded with. Of course that can be for good or bad, but in this situation, networks are particularly good at disseminating new information.
So if you’ve got a new practice guideline or a new best practice you’d like to spread through a network of physicians, how do you do it? Well, sending blast letters or faxes or e-mails is generally not terribly effective. They’re largely ignored and usually you can’t afford to visit everybody. That becomes prohibitively expensive. Again, networks and network influence turn out to work very well. When you look at the clusters of individuals, in this case physicians, look at influence patterns, you can use that information to help push information through new networks surprisingly quickly.
Williams: And how does this relate to patients? It sounded like you were doing some work with corporations about their employees, although that wasn’t necessarily health care related. When you’re analyzing physicians and their interactions with one another, does the patient come into play as well?
Miller: We’re doing some work in that area, too. One of the most dramatic examples is we’re working with a partner company, Healthways to map all the residents of an entire city in California; 300,000 people. Healthways will use the information to begin wellness programs for the entire population. The way we’re helping is once we map the residents of a city, we’ll look at connections, we’ll look at influence patterns and we’ll begin to understand if, for example, there is a smoking cessation program, where do you target that. Again, it’s prohibitively expensive to go to 300,000 people individually and work with them on specific programs, but using networks you can do it much more efficiently and we believe more effectively.
Williams: Turning back for a moment to the pharmaceutical industry, certainly the drug companies are known for being successful in influencing physician behavior. Obviously they’ve done that in the past without having this formalized kind of network analysis. Do you learn also from the activities that they do? Were they on the right track from their intuitive or pre-analytical approaches?
Miller: It’s pretty much what you would expect. A good sales representative in any industry, including pharmaceuticals usually knows his or her customers pretty well and the good ones often have a sense of what the networks look like in their communities. What we can do of course is make that systematic and spread it. We can understand these networks community by community and we can do it, if needed, for every community in the US. So, it’s a way of systematizing what the good sales representatives do already.
Williams: How has the network analysis field moved forward and what are some of the key hurdles that still need to be surmounted in order to make what you do even more successful?
Miller: There’s a long list and it’s still a young science. It’s been only 15 years old or so since it’s received substantial attention in the academic community, and even less in terms of applications outside academia.
We talked about a couple of areas at the conference that are challenges to network science. One is how can you increase the precision when you’re looking and trying to measure influence in a network?
Social influence is something that we and many others are quite interested in. One issue of course, when you’re looking at social influences, is how do you get rid of confounding factors? People are complicated. An important confounding factor is something we call homophily, that is birds of a feather flock together. If you’re looking at communities of people and you’re trying to understand changes in behavior, how much of that is social influence and how much of that is just people hanging out with people like them? There are very active research programs going on to improve the statistical methodologies there.
Another one –also raised at the conference yesterday– is how to guide a network toward specific ends, such as healthier behavior, once you’ve been able to map a network and understand the social influence. One way is to focus and intervene on key people. That’s a good start. But there is still plenty to be learned about how many people on a given network you need to guide to help move the entire network and how often you need to intervene to keep the ball rolling.
There’s a very substantial amount of theoretical work on these two issues. I’ll mention a final one that we’re very much involved in and it’s often seen when we’ve analyzed social networks. We’ve seen very large clusters, whether it’s consumers, physicians, etc.; 500, 1000, or even a couple of thousand individuals. The problem that we and others face is that what we know about social influence suggests that it really takes place in smaller groups; 100, 150, 200, maybe 250 people at the most.
So you’re presented with this very large network. How do you break it down? It’s something we’ve been very active with and we’ve developed some brand new methodologies to create what we call “communities” within these larger networks.
Williams: I hear about a lot of good things that can be done with this kind of work, like encouraging people to quit smoking. But it also raises the specter of whether these kinds of tools can be used on the dark side. I’m thinking about Orwellian approaches where Big Brother might not just be watching you but trying to influence behaviors within a network. Is that something to worry about, including in the health care field?
Miller: Privacy is always a concern, certainly for all of us. We work very hard to protect privacy of the data we use. But as I mentioned earlier, networks are very good at magnifying whatever comes through them, so one has to be very careful about using network influence to magnify something that’s for social good rather than not so good.
One might argue that the network analyses themselves aren’t doing anything new. We’re not, in some sense, doing different things than advertisers have been doing for years. In many cases those are good, but in some cases they may be manipulative. We are making that more systematic and probably more efficient. There’s nothing truly new. It’s simply a way of doing things more quickly and more effectively.
Williams: I’ve been speaking today with Larry Miller, CEO of MedNetworks. We’ve been discussing social network analytics and the role that his company is playing. Larry, thanks for your time.
Miller: I really appreciate it. I enjoyed speaking with you.May 10, 2011