OSF HealthCare Pilots EHR Tool That Surfaces Trends in Patient Data

Peoria, Ill.-based OSF HealthCare is piloting an internally developed clinical intelligence platform called CliniPane that surfaces meaningful trends in patient data from the EHR and presents it on the screen to clinicians but in a way that informaticists say is less intrusive than traditional clinical decision support alerts. OSF Innovation Senior Fellow Jonathan Handler, M.D., recently spoke with Healthcare Innovation about spurring innovation at OSF and the development of CliniPane. 

Healthcare Innovation: Before we talk about CliniPane, I was hoping we could talk more broadly about how OSF Innovation works to come up with and test new ideas.

Handler: I love being a part of OSF, because we really do have such a strong innovation culture. Any one of our employees can can fill out an innovation form if they have an idea. We actively solicit these sorts of things and encourage people to do that. We want to hear ideas from wherever they may come from. They’re vetted. We do a marketplace scan to try and figure out, has this been done before? Then we bring the ones that look promising forward, to see if we can make them happen. We have several different kinds of programs to drive innovation. We have a digital innovation development team that actually does some of the development work, especially on the software technical side. We have an Innovation Studio team that does something similar, but with very heavy user experience and design focus. We also have a commercialization engine attached to that Innovation Studio. We have an innovation academic innovate incubator, which is a partnership with five different universities in Illinois. Each of these programs has grant funding that can be applied for innovative ideas. So we have a host of ways to essentially encourage innovation, and then to actually execute on it, implement it, and then, where appropriate, commercialize it.

HCI: You mentioned these academic partnerships. I think I read that people from University of Illinois at Chicago and from Urbana-Champaign were involved in CliniPane. Was that one of the partnerships with academic centers around the state you just described?

Handler: Absolutely. We had the idea for CliniPane. But now, how do we execute it? We had a partnership at U of I at Urbana-Champaign to develop the very first prototype of the software application, and that would essentially interact with the electronic health record and be able to know which patient you are looking at right now, and who's the person that's looking at it, and then communicate with the software to get the data formatted and show it inside CliniPane.Then we needed some templates of how we want to display this information in a way that's really meaningful and impactful and easily understandable to a busy, sophisticated, clinical user. Our partners at the University of Illinois at Chicago worked together with our innovation team to develop that set of templates.

HCI: Are the target users of this both physicians and nurses at the point of care?

Handler: The target user of this is anybody who is working at the point of care. Our very first use case is deploying out to primary care physicians and advanced practice providers. We really hope to expand that much more broadly among our provider population, so we can support them in providing the best clinical care that they possibly can.

HCI: Could you talk about some of the limitations of traditional clinical decision support — or the capability of busy clinicians to surface the kind of information that this application is highlighting? In the traditional setting, what are some of the  challenges they face trying to get that information out of the EHR?

Handler: Generally, it's been my perception that there are two ways of interacting with decision, support. The first one, which was very popular maybe a decade ago or so, was to pop up an interruptive alert. It might say “you just prescribed a medication that this patient is allergic to. Are you sure you want to do this?” That seemed like a great idea. But they proliferated in many places, and clinicians felt there were too many of them, and too many of them were not helpful. Often they got alert fatigue and just clicked them away. 

The other way is that instead of pushing decision support to the clinician's eyeballs, the clinician had to do what I call pulling the information. You had to go somewhere. You had to make one or two clicks to see that information. And what I have found is that the minute something is hidden behind a few clicks, its usage drops off dramatically, sometimes nearly down to zero. So now we weren't interrupting the clinicians, but we weren't getting nearly the benefit that we hoped for from the decision support. 

What CliniPane aimed to do was to take a middle ground to say, what if we push it to the clinicians’ eyeballs, but not in an interruptive way. It's there. It's showing on the screen, but you don't have to dismiss it. You don't have to interact with it. You don't have to do anything with it. You can completely ignore it if you don't want to see it, but if you do want to see it, you don't have to do anything. It's just right there. We spent time thinking through how do we make it as minimally disruptive to the regular workflow as possible? We put a lot of work and design effort into that. We thought through when we do show something — even though it's not interruptive — how do we have as much confidence as we can that what we're showing might be of use to the clinician? So that was a lot of the design work that we put into it and and so far, it seems like we might have done something decent.

HCI: So some of the things you might show people are trending labs over time or blood pressure readings? 

Handler: Exactly. All of these are things that, given enough time, the clinicians could and probably would go find. But you don't have infinite time with a patient, and the clinician needs to be able to spend the time at the bedside of the patient. That's where we want them and that's where the patient wants them to be. We don't want them having to do all the work of clicking around to find what's important in the EHR. We'd like to serve that to them instead of making them explore and find it. So, for example, if there's a test result that is dangerously bad, we would like to surface that. However, what if there's a test result that's actually in the normal range, but it's at the far end of the normal range. And when we look at the trend, it looks like the next time they get tested, it will be outside the normal range if something isn’t done. Those are the kinds of things that we try and show to be not only reactive to what we now have seen happen, but to proactively prevent some things where that's feasible.

HCI: We are hearing more and more about how generative AI is starting to impact clinical decision support. Is there an AI element to this, or predictive analytics? 

Handler: We have done some work to incorporate AI, although some of these things are not live with the clinicians yet. We have done some work with classic machine learning predictors and incorporating those. We have also built a prototype implementation of gen AI interactions to help craft a prompt that may be more likely to get the response that you're looking for without you having to do all that typing. So we've done some work in the generative AI space with a lot more to come. We think there's a lot of opportunity there to do some real innovation. We are very excited about the promise that it holds, and we already have some prototype implementations, and we’re looking forward to rolling those out in the near future.

HCI: I read that one potential use for this would could be with external sources of data, such as from a remote patient monitoring system. Have any clinicians already experimented with trying to do that?

Handler: I have seen examples of people doing that kind of work in publications and in my outreach to my colleagues in the informatics community. We are capable of doing that with CliniPane, but we have not yet implemented that, because we're in our early days. But that is an exciting place to imagine that we would go forward. There's just this mountain of data that gets collected with remote patient monitoring. We could be aggregating that data and then surfacing it at the point of care in a way that's meaningful. It is an interesting informatics challenge, but it is something that I think CliniPane’s architecture and its associated technologies that we've put together are well tailored for.

HCI: Could you describe a few anecdotal responses you've heard back from the physician so far about their experience with CliniPane? 

Handler: My first and foremost concern was to make sure that we were not interrupting their workflow, that we were not consuming too much of their screen space. So far, it appears that we met those criteria from the feedback we've gotten. I was very excited that one of our our our beta test physicians, Dr. Justin Holschbach, gave us some comments about how he thought this was really helping with diabetes care. We have a hemoglobin A1C decision support card in CliniPane and it surfaces when the hemoglobin A1C is is abnormally high, and also when it's got trends that are concerning. We were really gratified to hear him say that he thought this was helping him in terms of diabetes care.

HCI: How are you envisioning rolling this out to more users within OSF? Do you go department by department and introduce it and see if there's interest in turning it on? How do you expect it to gain momentum?

Handler: Initially, we found some clinicians that we had worked with in the past and who were kind enough to try some of our other things that we did, and we said, “Would you please consider looking at this one?” and they were kind enough to do so. If they like it, we ask if they have any other colleagues that they think might also want to use it. 

HCI: So it is kind of spreading by word of mouth and reputation?

Handler: I love the spread by word of mouth, because that's when you get people who you know want the thing, and you're not imposing it on somebody. We started in primary care. We are looking at some other specialties where we could start branching out in terms of different kinds of clinical decision support. As we go beyond that, how about other disciplines? The nursing discipline, for example? Over the the next year, we are looking at branching out, both by specialty and also by the type of user.

HCI: Is this built on top of Epic?

Handler: We do use Epic. This is a program that lives separate from Epic, but it uses the interfaces that Epic provides in order to make what we think is a seamless integration.

HCI: So if this became a huge success within OSF, would you consider commercializing it so other Epic users could take advantage of it?

Handler: Those are conversations that we're having. I'm not the person who gets to make that decision. What I would say is it is absolutely my hope that one way or the other if we've done something good that helps clinicians, that we find a way to make it available to to other clinicians outside of OSF — whether that's commercialization or some other path.