 that IBM's IOD conference here in Las Vegas. This is theCUBE, our flagship program. We go out to the events, extract the signal from the noise. I'm John Furrier, the founder of SiliconANG, I'm joined by co-host Dave Vellante. And our next guest is Tom Daub, CEO of Centerstone Research Institute. Welcome to theCUBE. Well, thank you very much. So tell us a little bit about what you guys do first. Then we'll jump in to some questions. We know we've got some good case study data with what you're doing. So tell us about the company first. So Centerstone Research Institute is a mission-driven nonprofit organization we're in Tennessee and Indiana. And we focus on bridging the gap that exists between research and practice. So every day we learn mountains of new information that come out in research that affect both healthcare and also behavioral healthcare, mental health and substance abuse, which is what we focus on. And we're all about trying to bring that research closer to practice. And data, big data, is one of the ways that we've been doing that. So just take us through a little bit of how you get the data. Is it from your own sources or external sources? A range of sources. So healthcare in particular has not leveraged over the years all of the data that we collect. We're required for regulatory reasons, for payer reasons to gather mountains of information in healthcare. And only recently have organizations begun to really mine that information and use it to improve care. So we get data out of electronic health records as a primary source of information about the services that are provided, about patient outcomes, how people are doing in care, but we also bring in other sources of data. So for example, one of the things that we've done is look at the impact of just weather conditions on our care delivery model. So when people don't make it to the clinics, is it because of bad weather conditions and how can we take that into account and making our business process more efficient? So one of the things we learned, Dave and I were talking with GE and they do have a lot of oil and gas companies and they found that they've always stayed within their businesses of looking at their data, their data warehouses, but when they started using external data sets, in their case, oceanography information, the insights were amazing. So can you share with us an example of where you guys have had your normal data sets and then went outside your scope and brought in other kind of categorical data or industry data. Have you guys, can you share a little bit of insight there? Well, the real power I think in data is bringing together different data sources. So in most organizations, you have data that are siloed, so you might have in a healthcare organization, you might have clinical billing information in the electronic health record, you might have an HR information in HR systems, you might have financial information in financial systems, the general ledger. So by crossing silos and bringing that information together, you get more power. Now as you're talking about, we can also go outside of those systems and bring in external sources of data. So we've, for example, brought in weather information and we're able to look at the impact of that on whether people are able to make it into our clinics or services so we can manage our staffing more effectively. Oh, good. Go ahead. So you guys are essentially building a big data warehouse when you're housing this information. Take us back to the driver for that system, when did it all start, how are you doing it, maybe take us through sort of the case study of what you're doing. We started our journey, our analytics journey about five or six years ago. We were very fortunate as a nonprofit to get investment, philanthropic investment that helped us start this effort from the Josie Davis Foundation and also from the Ayers Foundation in Nashville. And that helped us get our effort off the ground. They funded the initial infrastructure that we purchased some of the software and hardware we needed. Also some of the training that we needed for staff. We had to grow our own data scientists, of course, they're hard to recruit. And we were able to bring people in who are interested in solving really important social problems and applying the latest data science methodology to address them. So first thing we did is we began to build all of that, the data infrastructure. And we did that by focusing on key business problems. As every healthcare organization faces today, we had several crises, that business crises that we faced. Payers were changing the way they did business. The healthcare environment is changing very rapidly. And that created business opportunities for us as an analytics to help solve those business problems and demonstrate our value. So, okay, so you put it in this infrastructure and then you said you had to homegrow your own data scientists, right? So what was that like? I mean, essentially, did you find a lead data scientist that could train everybody and transfer that knowledge? That's a major challenge for a lot of organizations. How'd you do it? Well, it's interesting. As a research institute, we took the approach of creating a center of excellence around analytics. Prior to starting our effort, we had three separate data groups within the organization. We had a group in our quality improvement area that focused on quality improvement data. We had a team in our information technology area that just did general reporting. And then we had statisticians, which were a my group at that time doing research work. We put that team together. It was very interdisciplinary. People with PhDs and computer scientists and quality folks, and they really learned from each other. So we have different people focused on different aspects of the work, but we are able to bring in people, attract people honestly that were interested in solving key social problems. They connect to the mission of the organization that we're focused on very important issues in mental health and substance abuse as an industry and as a society, and they wanna be a part of solving that. So that's actually helped us draw people in that a small organization honestly wouldn't usually have access to. Okay, so you got a mission, you got funding, you got tech, you obviously have domain expertise, you got data science. Was there any other ingredient that you needed to succeed? A crisis actually would be my answer to that question. We had, in our primary business, so Centerstone is a large behavioral health care provider organization. We serve about 75,000 people a year with mental health and substance abuse problems. And we have primary operations in two states, in Tennessee and Indiana. Both states in recent years have gone through significant environmental challenges. Contract changes, state regulatory changes, other things that happened that we're gonna make it very hard for us to be able to continue to do business as usual. And our analytics group was able to come in in both situations and help the business owners get the data to understand the situation, to make the good strategic and tactical decisions to manage through it, which helped us thrive and continue to do the work that we needed to do to meet our social mission in an era of shrinking funding and environmental challenges. So talk about the data science you mentioned. We had this conversation with Dr. Tim Buckman earlier, this comment online that said, hey, you know, he's a high-priced employee managing air traffic control of the hospital. But then he turned around and said, no, no, actually there's leverage in that because what he's doing is automating and bringing in more data points. So that was an interesting phenomenon. So again, in your organization, you have a scaling issue where you have a lot of people interested in helping. A lot of times the brain can't talk to the arms and legs, if you will, fast enough. So this is the social business phenomenon. How are you guys using the analytics to almost take that top-down approach but enabling an organic participation? Could you kind of thread that together? Because that's a challenge that many feel is critical. I want to enable organic and not be top-down oriented, but you still can do that. One of the things that I'm most proud of in our effort has been our ability and our success in getting staff involved, frontline staff, managers, people who are making decisions every day out in the field. And we may be doing some complex analysis and predictive models and so forth within our analytics team, but it's really about using that information. A model is only so good as it is being used on a day-to-day basis, and it was those business challenges that helped us drive adoption through the user base. So we had external pressures that helped us really drive the cultural change necessary within the organization to increase and improve our use of analytics throughout the management team. So you're five or six years in. What kind of results have you seen? We've seen tremendous business results. So as healthcare is shifting currently in health reform from an environment where they used to pay for fee for service, so you essentially see a client and you get paid for that, regardless of what your outcome is, it's shifting to payment for value. And what value is, it's really the equation dividing outcomes, your outcome, your unit of cure divided by your cost to achieve it. So for example, a new drug that can extend life by a couple of months and costs hundreds of thousands of dollars may provide a good outcome, but it may not be good value because of the increased expenses is so great. We've been very focused and we're talking here at this conference about how do we improve healthcare value? And behavioral health is really one of the areas that we have a great opportunity to do that because behavioral health, mental health and substance abuse is implicated in the very heavily in the increasing healthcare costs that we have as a country. So people show up in the emergency departments, hospital readmissions, those kinds of things all connect to mental health and substance abuse issues. Yeah, so essentially you described an ROI equation. Unit of care over cost to achieve, that's benefit over cost to achieve a benefit. Now, what I get excited about is that traditionally in the technology business, generally but specifically in healthcare, the best way to get an ROI was to cut the denominator. And what you're talking about is driving the numerator, that unit of cure. What metrics are you using to drive that numerator? So we have a number of different outcome measures that we gather. So in our setting when we see people and this is typical of mental health and substance abuse settings, we'll ask people questions about whether they're getting better or not. So we might see somebody who's depressed or has a substance abuse problem. We'll ask them questions, are you using drugs and alcohol less or is your mood improved? So we gather that information and then we're able to track over time the positive or in some cases negative outcome that is occurring with an individual. So as we track that, we're able to input that as data into our models to assess are we increasing outcomes and are we able to reduce costs as a function of that. One of the cutting edge things that we're working on today is actually modeling the decision making process that in particularly physicians and medical providers, other healthcare providers in our system use in care. So often in a fee for service model, you're incentivized to provide more care than somebody needs. And if you just stop when somebody gets well, it's amazing how much value that can create in healthcare delivery. So you're affecting both sides of the equation, right? Exactly, in ways other than just stopping spending here. And we've shown with some of our modeling that we've done, you can actually improve both. You can, we've shown that you can increase outcomes by roughly 42% and I think it's actually 60% and decrease costs by 42%. So one of the things we were talking about earlier is that modeling is great. I want to talk about that in a second on the auto, because that's automation, when you can automate things at scales. But the data accuracy is critical. So how do you guys make sure that the quality of the data, when you're going and doing the modeling, because you're modeling decision making, which is important, which puts new capabilities in the hands of real users. The issue of data quality is huge. And this is an area where I would probably depart from the conventional wisdom around data governance. Certainly data governance is an important thing. But in our experience, it is the first time you put data out for people to use from a production system, you're going to find problems with data quality. And the only way to resolve that is transparency. It's to take that data, make it available to people so they can see the problems that exist and figure out what they can do. So basically, if you control it, you're always going to have a quality snafu. Essentially, not again, sometimes I put words in your mouth, but there potentially could be a quality snafu versus empowering folks to iterate through. Is that what you're saying? I think you can get to the same place through both methods, but the data governance method, often you're trying to make sure everything's right before you move data out. And you have to do that in certain mission critical areas. If it's a life or death, if you're in an emergency room, you can't put out messy data. But in a lot of situations, you can afford to do this. And people use production systems in very interesting and creative ways to do their job that usually makes sense in the context of their job, but create messy data. And when you shed light on that, that stuff gets cleaned up. That's something I didn't realize when we started our work, is how powerful just that visibility and transparency factor would be in helping improve the data quality, which would ultimately feed back into our models. I'm interested in how you're organized to sort of address these opportunities, really, is what you're describing. We did an event with MIT in July. It was the Chief Data Officer Symposium. And there was sort of a big debate, really wasn't much debate amongst the Chief Data Officers, but there was certainly a big debate in the industry as to whether or not you should even have a Chief Data Officer. Is that person responsible for information quality? Should that be the CIO's job? As somebody who's got, you know, five, six years of messing around with this stuff, what do you think about that role? Should there be a data czar, if you will? Should that individual be part of the IT function? How do you guys handle it? I think at the right scale, there's a good reason to have a person who focuses on that role. In a smaller organization such as us, we are in the aggregate, we're about $130 million organization. Everybody's a data czar. Right. So, you know, that's sort of distributed. I wear that hat. Our analytics team wears that hat. Our business lead wears that hat. But what we did do that's interesting about that, as I said before, we brought those resources together out of IT, out of quality, out of research, and made an analytics center of excellence that focuses on solving these problems. Now, we work very closely with IT. We work very closely with quality. We work very closely with the business line managers in solving these questions. So it's really more about how you communicate and how you work together than any particular structure, I think. Question from Twitter here. Grant Case, one of our most important, his description, is an important member of the CrowdChat community has a question. How do you get your organization to buy into that iterative model for data? We didn't ask permission. And we didn't have much choice. So we started our initiative as a bit of a stealth initiative. Did you ask for forgiveness? Or did you get to that part yet? You didn't? Not yet. When we make mistakes, we ask for forgiveness. So what happens when we put that information out is there are often problems with it. We use a fairly agile, as an overused term, but a fairly agile approach. We want to get the data out there. We want to get people using it. That then brings ROI. It improves data quality, and it's an iterative improvement process. So you get to tend to value fast enough? We have to, and we've worked to have a culture where it's okay for the data not to be perfect. Now, I would separate, as I did before, if you're talking about life or death situations, you can't afford to do that. You have to make sure data quality is there before it gets out. But in a lot of situations, you don't have to do that. If you're just looking at financial information or service volume, those kinds of things, weather data, it's typically not life or death. So you can get that out there. And the noise is rarely going to be great enough to actually change the overall signal, the picture of the signal that you have. So one of the things we talked about earlier was situational awareness and context data in motion. Share with the folks just some experiences of how the data has impacted some of the, just the folks in the community. Any research has come out that's impacted lives? Can you share any insights there? Any examples? Well, I think there's sort of a business example and a clinical example that the business example, a couple of years ago in Indiana, the state was changing how they paid for services. And this happens every few years in healthcare and it's happening federally now. And it can be very disruptive to systems like us as we figure out, okay, well, the services that we were providing are no longer paid for. And the challenge is, it's not just a financial challenge, it's a challenge to how we serve our communities as a mission-driven organization. If we take a funding cut, a significant funding cut, that reduces the services that are available to people out there who need them. That demand doesn't go away, it's just people who aren't met. So we had to focus on using our data to inform management to make decisions about how do we reconfigure our business given this new set of rules that were handed to us by payers to still meet the social need that was out there. And we've done that very effectively. Now, on a clinical side, as I have just talked about, we've really been thinking about how do we get information into the hands of clinicians that helps inform their clinical decision-making. So this value issue, how do we help them know when to stop providing treatment when they've reached their maximum effectiveness so we don't continue to incur that cost and so we open up a slot so somebody else can get in to see that therapist or doctor. Tom's been great chatting with you on theCUBE here. Really appreciate your amazing work. I know you guys are a non-profit and again, the volunteer activity, we see that with social media, brings that social component together. It's really great stuff and big data allows you guys to scale and do more great, great stuff. We're here live inside theCUBE here. IBM's information on mental in Las Vegas, the hashtag is hashtag IBM IOD. You wanna join the conversation here on theCUBE with myself and Dave, go to the crowdchat.net slash IBM IOD, that's our new crowd chat application. You can sign in with LinkedIn or Twitter and join the conversation, ask questions and we'll address the board. We'll be right back with our next guest after this short break.