[MUSIC] We hope that you've enjoyed this particular course, talking about organizational readiness for change, concepts of leading change, governance, and clinical decision support, project management, strategic planning, and change management itself. At this time, we hope that you've gotten excited about the field of health informatics, and that you've been able to collaborate with other classmates in the online platform. One of the articles available this week is from the New England Journal of Medicine. And it's written by several former ONC coordinators who write about some of the path ahead. The article is titled, The HITECH Era and the Path Forward. And in it, Peter, as you know, they write about quote, vexing issues of interoperability, usability, privacy, security, and data stewardship. Of these vexing issues, can you speak a bit to which ones you find most on your plate these days? >> Gosh, these vexing issues are my life at this point. This is what I live very routinely. I've called out a couple. We've done a lot of work to define data stewardship and what that means, and how do we define systems of truth. How do we want to use the data? What should the data flow be? I think that is all the time now, it comes up for me. Not just do we have some system, but how does the data actually flow? What's the system of record? What are the systems that are going to reflect the system of record? And have we really thought through all their processes? And then I also spend a lot of time talking people off the ledge. >> [LAUGH] >> From keeping data in a silo, and making sure that, hey, instead of sort of thinking of yourself as being an owner, think through what it should mean to be a steward. And if you're a steward, you are getting the data out to the people who need it, to make use of it. You're still getting recognized as the person who knows most about it. But that's really important from a data stewardship standpoint. >> And I think a lot of headway has been made. But I'm sure you still get feedback occasionally. Friends of mine that are taking on clinical research, we have this enterprise wide system. Why can't I just have all of the data that I need the way? But it's not so easy, and it takes years to execute on. And the issues of privacy and security really are utmost. >> It is enormously important, and I think this is an area where there is a lot of great innovation occurring, where we really do want to free the data. But we also want to make absolutely sure that we are being respectful of the privacy of this data. These data are different then any other industry in that regard. And boy, you don't need to spend long. I do get contacted directly, sometimes by people who have very sensitive medical conditions. But it's often employees, who want to make sure that I understand that they're worried that we are exchanging data with so many partners in this process. And so these dilemmas of how to use it become very hard. And one of the things that we did, we actually were a part of help starting a company called Protenus, which uses very advanced machine learning techniques to do outlier detection. They do sort of graph database analysis. And they traverse that graph to identify outliers, in terms of identifying people who are inappropriately accessing records. And that sort of advanced capability is very exciting, because we're actually starting to automate some messages back out of that system to people who are incorrectly accessing records. And so we're trying to make this a more automated approach. >> Yeah, very exciting application of machine learning, and we've talked a bit about privacy, security, data stewardship. What about interoperability? When it comes to vexing, that one certainly takes the cake, right? >> It is such a hard problem. And you have to look back at the last 30 years of organizations that have worked on trying to get standards. And one of the things that that has made me realize, is that it is not just a standards issue. You can create a standard that doesn't get adoption, especially if you get people involved in creating the standard that aren't actually using the systems. And so you can do a lot of standardization that really doesn't get uptake, and doesn't drive change in the industry the way you would want. There's so much excitement, appropriately at this point, about FHIR. And that's something that comes up in this and other informatics courses. But one of the things that I also realize is that even FHIR alone is not going to be enough. There's going to be, needs to more explicitly specify what's in some of the payload contents of FHIR service messages. And more importantly, I think that our major vendors need to be in a position where they're deploying FHIR in a way that we can plug apps into that vendor. So in our case, we're an Epic customer. And I was very involved with Epic at pushing them to create an app orchard. And I also created an advisory body for those who are customers of Epic, to give advice into that process. because we really want to enable the best possible platforms for innovation. And I think that's the kind of mix that we need, because we don't want one vendor trying to do everything. We want to create this ecosystem where lots of people have great ideas, and they're able to create interoperability with the platforms. And we've gotta get the right value proposition between those startup companies and the major vendors. And so I think that that is an exciting way forward. We all have figured out that there's some pretty amazing things that we can plug into our smartphones and our tablets in this process. And really that sort of platform plus plugin strategy, with defined APIs, is what has driven lots of the major software systems, Twitter, Facebook, Amazon, etc. And that's really what we've gotta have for health care. >> Yeah, certainly an exciting time, but those vexing issues continue to be, they're exciting challenges to have, right? Great progress has been made. But like the authors write, with health care transformation, there continues to be this need to address especially these five hot topics. And usability, we've touched on quite a bit, as well. And I go back to one of our original discussions during this course, and being able to recreate the trifold, [LAUGH] and how usability is so important. And there are some solutions in this optimization phase, but it's certainly still a challenge, and the clicks are plenty full. >> It is, and what's really interesting on the usability is we're all so excited about artificial intelligence at this point, and being able to bring advanced machine learning models into healthcare. That is, the very exciting part of my job at this point now is figuring out how we use the data to create those models. To what extent do we do internal work? When do we use third party companies? When do we use our existing platform companies? And so there's really interesting conversation and efforts going on there. But this vexing issue of usability actually comes up right away with artificial intelligence. And it's so important to understand, exactly where are we bringing this insight into the clinical workflow? And where is the clinician? Who is the clinician or other person? Who's the right person to bring this message to? What's the timing, what's the venue? And interestingly, it's not necessarily in the EHR that's most valuable. because if you've got a really good algorithm, and we've done some great work with Suchi Saria, one of our computer scientists, who's just spun out a company called Bayesian Health, that we're counting on to do sepsis detection and other kinds of clinical deterioration. But she and the whole team that's been involved have really thought a lot about the human-system interface there. And so we make sure that the human beings get a chance to interact with the machine learning model to shape what's important, and ignore certain patient characteristics if they're not relevant. And also, it is explains itself, so you get into this whole topic of explainable AI. If it's just a black box, and you're just trying to tell the clinician this patient is in trouble, they're going to say, why? What is the reason that you can use my net? And then, finally, it's giving them information back from a usability standpoint, in context without extra clicks, so that they can easily make the right decision and get that documented in the record. >> Yeah, in terms of user centered design, explainable AI, very exciting. But how much transparency and information does the clinician need at the point of decision making, versus how much is too overwhelming? Really difficult, [LAUGH] and big questions. >> Yeah, that's a new and exciting space for us. And that's something that we're all going to be figuring out. AI is already getting way over-hyped. And yet we know that there'll be, I think that in the next few years, we'll go from having a handful, to dozens of models being in production in our systems. And so the leadership skills that hopefully everyone in this class has learned are all really important to help us through this transition. Yeah, that's a great point, Peter. And as rates of adoption for telemedicine rise, as we continue to see progress in ACOs, clinically integrated networks, as we introduce voice recognition software perhaps into the hospital lobby, there's plenty of change that individuals need to continue to lead. And I think it's an exciting time to be in this field. And we've certainly touched on a lot of topics during this course. Of course, it's just the tip of the iceberg. [LAUGH] So we hope you enjoyed the content, that you've gotten some great insight into health care IT transformation that's currently underway, and that you all look forward to being a part of leading change in health informatics.
