(pulsing music) (digital music) >> Welcome back to the Seaport in Boston. You're watching theCUBE's coverage of Red Hat Summit 2022. A lot of bummed out Bruins fans, but a lot of happy Celtics fans. We're optimistic for tonight, Boston's crazy sports town, but we're talking tech, we're talking open source. Dr. Amanda Purnell is here. She's the director of data and analytics innovation at the US Department of Veteran Affairs, and Ben Cushing is the chief architect for federal health and life sciences at Red Hat. Folks, welcome to theCUBE, thank for coming on. >> Thank you for having us. >> So glad to be here. >> So we heard your keynote this morning, project Arches. Now you were telling us just briefly about your previous life as a clinician. >> That's right. >> That's really interesting, because you know what the outcome has to be. So talk about that project in your perspective. What the goals were and how you actually got it done. >> I could tell the long view. I'm a psychologist by training. I spent the first 10 years of my VA career providing care to veterans. So engaging in healthcare behavior change, providing training to providers and really trying to understand what is the care pathway for veterans, what's the experience of veterans along each of those touchpoints, and it became clear to me over time that there were opportunities for us to improve the transitions of care and provide better information at the right time to improve those decisions that are being made at the point of care. Ben and I were just talking before we began today, part of the core of the development of Arches was beginning with human-centered design. We wanted to interview and better understand what was the experience across the VA of many different stakeholders and trying to access meaningful information, understand in that moment what do I need to make a decision with a veteran or what do I need to make a decision with my care team and how can I improve the quality of care for veterans? And so, hundreds of interviews later, it became clear to us that we wanted to help those individuals already working for the VA to continue to improve excellence of care and one of those ways that we're trying to do that is using technology to make life easier for our veterans and for our clinicians. >> I always like to say, they say, "Follow the money." I like to follow the data. And you said something in your keynote about nurses have to have access to information and it just gets to an architectural question, because as a caregiver, you have to get insights and data and you need it fast, 'cause you're saving lives, but a lot of times, architectures are very centralized. They're monolithic and you have to beg, borrow, steal, break through blockers to get to the data that you need. How do you square that circle in today's world? Maybe you could talk about that, and then specific to Arches, how you dealt with that. >> I can dive into that a little bit. I have to say, Amanda had touched on this during the keynote, VA was one of the first, if not the first, healthcare organizations in the world to actually adopt electronic health records and because of that, they just have this incredibly rich amount of historical data and the challenge, as you pointed out, is gaining access to it. So there are a number of programs within VA designed specifically for that. And they are bringing data not just from the data warehouses, but also data from the electronic health records that are running inside of VA right now, and then also third party community data sets, as well as applications that run inside the VA. Now the value here really happens when you produce insights. Data by itself is useless. >> Lot of data out there. They're plentiful. >> You need to create knowledge and then you need that knowledge to inform your process that comes next. Those actions are really what matters. All of healthcare is process and activity and data is really just a historical record. I mean, all data that we look at is happening in the past and then as we're reading it, we're producing knowledge, again, to inform our process. Arches, the program itself, is right in that space at the knowledge layer of actually taking that data and turning it into actual insight and something that is usable and insightful for clinicians to affect the ability to deliver better care and also to actually improve their own working experience. A lot of the models that are getting built out are specifically designed to help their workflow, help them reach better outcomes for the veterans, but also for themselves, because if we can care for the providers, it'll certainly help them care for the patients even more so. >> So how does it work? I mean, from the provider's perspective, how was their life improved by Arches? >> That's a great question. We want to make it easier to access the information. So as Ben noted, the average person providing care in the field doesn't know how to code, doesn't know how to pull a unique request for an individual data point, and what we're trying to do with Arches is provide a user interface that allows for both a non-technical person and a very technical person to access information, and then what gets provisioned in front of a provider is something that is farther abstracted from the underlying data layer and more like here's a specific insight. So I use the example in my keynote of chronic kidney disease. So what's provisioned to the provider in that moment is this person is at higher risk for chronic kidney disease based on this basic information. So it's surfacing just the right amount of information to allow for that care pathway to be improved, but the physician doesn't need to see all of the layers of code underneath. They need to trust that it's worthwhile, but they don't need to know all the background abstractions. >> So it's a self-service, essentially, infrastructure in that sense. You're hiding the underlying complexities. You gave an example in your keynote of an individual who realized that they were under counting the probability of a potential disease for African Americans. >> Yes. >> I believe she just rewrote the algorithm. >> She did. >> Describe that process, because in a lot of organizations, injecting that new algorithm may have required new data sources, would take an act of the Pope to do. How did it work in Arches? >> This is what I get excited about with Arches is that we have the opportunity to empower enthusiastic people like Dr. Joshi to discover an insight and she's a talented informaticist, so she could do the technical work and provision a container for her to work in, for her to do the data analysis, the underground stuff that we're not letting the average provider have to cope with. We were able to provision the tools that she need, the environment that she needed to be able to test and develop the new insight, confirm that they're there and then begin to validate that and test it in other facilities. So our thinking is, how do we bring the resources to the users rather than saying to the users, "This is what's available. Good luck." (chuckles) >> So we've been talking a lot about, I'm sorry, go ahead. >> I want to add on to that. What we're actually experiencing inside of healthcare right now is the emergence of of learning health systems. >> Yes. >> And this is a great example of that. The terrifying number is, it takes 17 years for new knowledge that gets created with healthcare research, whether it's NIH or VA or elsewhere, it takes 17 years for those practices to make their way into practice. Generally the way that happens is through the education of new staff. And so the dissemination of that knowledge is just so freaking slow that we cannot move nimbly enough to take on that new knowledge and actually implement it in clinical space. What Amanda's describing is something that now happens in months. New knowledge getting produced and then actually getting disseminated out, both the insights, whether they are those probabilities, predictions and recommendations and the actual processes, which are getting automated, as well. So if you think about healthcare as just a process, you can automate a whole lot of that and we can move that needle really fast and actually take that 70-year number down to a couple months. >> In the early days when we were all talking about AI and getting excited about digital, I would often ask the question, will machines be able to make better diagnoses than doctors and to your point, Ben, that's not the right question. >> Exactly. >> It isn't the right question. >> The question is, how can machines compress the time to better patient outcomes- >> Yes. >> in concert with humans and that's what we're seeing now. >> That's right, it's surfacing those insights to start a conversation. >> We've been talking a lot about artificial intelligence for the last two days. As clinician or someone with a clinical background, how do you see the clinical experience changing as machines grow more intelligent? >> I think that there's a learning curve for people to feel confident in an artificial intelligence. It makes sense. So someone spent decades, perhaps, of their life obtaining medical training, doing fellowships, doing additional training that they have trust in that deep training. There are times, however, where a technology is able to surface something that we didn't know that we didn't know and it's important, as we make use of artificial intelligence, that we clearly validate it with independent means and that we clearly also bring in additional analysis to understand what are the elements and then test that new technology in an environment before we scale it widely, so that clinicians can see, yes, this was useful. If it wasn't useful, how can we make it even better? So it goes back to what we were talking about earlier that we have to bring in human-centered design to figuring out how do we make use of AI or machine learning models and make sure that there's trust in those models and that we can clearly articulate value for the clinicians and care teams on the ground. >> Is that a natural evolution of Arches? >> This is all built around it. Arches is the technology platform, but there's no magical technology that's going to change how humans interact. And so the way we think about each project is we think about what are the technological components and what are the human factors components? And we have to think about the entire care pathway. I'll go back to that example, the chronic kidney disease. She identified that we were under identifying African Americans for chronic kidney disease. So she changed the algorithm. Not only did she change the algorithm, we also had to think about who would be informed of those changes, how would that change, who would be connected to the veteran in that point of care and build out the care pathway in the care team and that's really how you actually influence an outcome. Surfacing an insight is important, but it's one part of a much larger picture. >> So what is Arches? You said it's a technology platform built on open source. At least, there's a lot of open source in there. And it's got API connectors to all the legacy technologies that you need it to. Can you describe, paint a picture of what it actually is? >> Arches is evolving as it should. So it's designed to meet the unique needs that aren't being met by other infrastructure in the VA. So we started first by identifying the need for cloud compute, so it's in the cloud, it has open source technology so that we're not stuck with any one provider and also has the ability to use containers to be able to move insights out of Arches to an enterprise solution. We're also bringing in multi-cloud strategy, which also something had been discussed quite a bit at this conference, to make sure that we're not saying only one cloud provider can be the solution for veterans' needs. Our mission is serving veterans and so we want to have access to all the technology and not just one and so we're looking at how do we expand the scope to make sure that we have the most variety possible so we can meet the needs of veterans. >> I can add a little bit to it, as well. Think of Arches as a program. It's an incubation space under the office of innovation. So it's a place where the governance allows for trying new ideas and really pushing the envelope for VA in general. There's not a lot of organizations, if any at VA, that allow for that type of incubation and so Arches is in a unique position to create new technologies and new novel approaches to solving big problems. And then the next step to that is moving the work from Arches out into the enterprise, as you called it out. So for instance, the system of engagement where the actual clinicians interact with patients, the model needs to find its way there and we can't do that in a way that disturbs the current workflow that the clinicians have. We need to be able to bring the model to where the clinician is, have those recommendations, probabilities and predictions surfaced to the clinician in a way that is precise to their existing workflow. They need it at the time they need it. Arches itself is not delivering that part of it. It's more like the place where the innovation happens and the incubation really occurs and then it's about taking this container, really, and moving out to other systems that are already deployed out to the hospitals, the edge, and in the cloud. >> And the federated governance occurs in Arches or elsewhere? >> It happens across the continuum. It's starting in Arches. the clinical validation that happens there is wickedly important, because the clinicians need to know that what they're working with is actually legit. And so when they know that the researchers and the clinicians who are involved in that incubation period have done their work, they can feel confident with the recommendations they're getting from the machine learning models that are getting deployed to one of them. >> So many questions, so little time. What's the business impact? How would you describe that? >> For me, it's an emotional impact. People have a sense of, "I have a place to develop a solution and I can get in there quick, and I can test out an idea. I could potentially partner with an external partner or if I have the talents and skills to do it myself." It's empowering all of those innovators who have great ideas to work together to test and develop and validate solutions, and they're not waiting years to get the idea off the ground. >> Amazing. >> Go ahead, bring it. >> Is Arches open source? >> Arches is a platform and it has open source component. So that the underlying infrastructure of technology is open source. >> Why was it important to you that this be built on an open source platform? >> It's important for us that we not marry ourself to any one technology and that we allow for, as much as possible, transparency and many different tools and the right tools for the right solution. So we didn't want to find ourselves connected to only one way of doing things. We want to have versatility to have the right tool for the right problem at the right time. >> I'm so sorry, we're out of time. This is so interesting and I really appreciate you here guys, coming on and sharing your insights for theCUBE audience. All right, keep it right there. This is Dave Vellante for Paul Gillin. We're in day two of Red Hat Summit 2022. You're watching theCUBE. (digital pulsing music) >> Due to the pandemic, the federal government declared a public health emergency, which created an urgency for healthcare coverage. >> One of the biggest-