How AI-Based Tools Are Improving the EHR Experience at Univ. of Iowa Healthcare

University of Iowa Health Care has seen improvements in clinicians’ perceptions of the EHR by working with two AI-based solutions: Evidently, which improves access to and summarizes both internal and external data, and Nabla, an ambient scribe tool to help with note generation. Jason Misurac, M.D., part of the health system’s informatics team, spoke recently with Healthcare Innovation about these deployments and next steps. 

HCI: Could you describe your role at University of Iowa Healthcare and how you got involved in these AI developments involving the EHR?

Misurac: I'm one of our associate chief health information officers. I'm a pediatric nephrologist by trade, but I'm also a clinical informaticist, so I have a little bit of variety in my professional life, which has been rewarding. What really motivated us at University of Iowa to look for a solution like this was that we did the KLAS Arch Collaborative in 2023, and what we found was that the integration of external data was a sore spot for our users.

HCI: Was it surprising when that rose to the surface as an issue?

Misurac: Well, the University of Iowa is primarily a referral center, so we have a fairly large catchment area all across the state of Iowa and into some neighboring states as well. I's very common for our patients to have traveled some distance to see us and to be receiving their primary care or even other specialty care elsewhere. That means that there's data in multiple places. There are scanned faxes, which are still pretty commonly used in the field of medicine. There is data from Care Everywhere, which is Epic’s background health information exchange. And then there's our own generated data from University of Iowa itself. So while this wasn't something that we already had as a high priority to address, when we saw these results, they weren't that surprising to us, given the nature of our institution and the patients who come to see us.

HCI: Does some of the data come through a health information exchange?

Misurac: Yes, sometimes it does. The mechanics of accessing that is all through that one interface in Epic called Care Everywhere. Even if it's not coming from another Epic institution, all of the HIE data comes through Care Everywhere, so I don't have to go to an external web page to access that. But it still means that as a nephrologist, if I want to see a patient's trend of serum creatinine over time, I may have to go to three different places to see that — the scanned faxes, the Care Everywhere data, and then our own lab results tab.

HCI: How did Iowa get involved with this company Evidently? Was there a process of co-developing or customizing a solution?

Misurac: We started looking for tools that could address the problem of external data integration, and pretty quickly we found that Evidently seemed to be the most promising to meet our needs. I will say the partnership with Evidently has been rich. It was more than a year ago that we did the pilot — and almost a full year since we went enterprise-wide with deployment. From the very beginning, the Evidently team has been receptive to and hungry for feedback and we've seen improvements as a direct results of our feedback. 

HCI: Does the data aggregation or summarization happen all within the Epic workflow? Or is it an application that sits to the side? 

Misurac: It’s an application that sits as a tab within Epic.

HCI: So using your own clinical work as an example, is this saving you time? Are there things that now you have access to easily that you might have had to dig around before to find? Both external data but also within your own health IT ecosystem? 

Misurac: All of the above actually. No matter the source of the data, the summaries inevitably are useful just to say: who is this patient that I'm going to see? What’s the update since I last saw them? There also are a number of views of data that incorporate the data from all the sources. For example, you can see all the clinic notes, whether they were here, there or elsewhere. You can see a table of the lab values that incorporates all of the data sources in that one table, or blood pressures. Again, I'm a nephrologist, so I'm thinking about hypertension in my patients. And if I want to see all of their blood pressures, not just the ones done here, that's another view that that is within Evidently.

HCI: Does this also have the potential to help with documentation and/or with billing?

Misurac: Those are two very different questions, although related. I’ll start with documentation. About the same time that we introduced Evidently, we also introduced ambient documentation at the University of Iowa and and our vendor for ambient is Nabla.

Nabla obviously has access to everything you discuss with the patient, and Evidently has access to everything in the chart. And those are two Venn diagrams with some overlap, but not a ton, right? So the the idea of combining those two sources of information, neither one of which is comprehensive, to create a note, has been very attractive, and we're now starting to see some prototypes of that from Evidently. They’re starting to do some testing of their note drafter tool. So using the data from the chart, this will draft a note, and if there's been an ambient encounter, it will capture that output as well and incorporate that. That has the potential to be richer than a note that you could generate from either ambient or chart summary alone.

HCI: Does that require those two companies to work together?

Misurac: It requires that you have exported the Nabla note into Epic, because once something is exported into Epic, then Evidently has the ability to access it. But I will say that there's another way that this could work, where you send the patient summary from Evidently to Nabla, so that Nabla has the context when generating that note. And I've seen a prototype from Nabla doing that. So these two companies have talked to each other, and they're building the APIs in the background for sending that patient summary from Nabla to Evidently, trying to fill in the holes of that Venn diagram.

HCI: I read that Evidently has something like a conversational chat AI that enables users to ask the chart questions in natural language. Has University of Iowa started deploying that?

Misurac: We have a small group of people testing it. Similar to the note drafter tool, that is one of their newer, more advanced features. You can think of it a little bit like an LLM that you ask questions. A common question for me is: tell me about his kidney history, and it will give me a summary, or tell me about the treatment of the blood pressure, and it can give me the details of that specific patient's blood pressure treatment over time.

HCI: Are you seeing these tools used differently in different parts of the health system, like in emergency or primary care or by the hospitalists? 

Misurac: That’s where the ability to customize is really useful. Because if you're in primary care, you may want to see different things up front than you would if you're in the emergency room, so you can choose different summaries. I have a nephrology-focused summary in addition to the general summary. Then you can rearrange which data elements are higher up or lower down on the page based on what you like to see. We're seeing a fair bit of use in the emergency room and some of the emergency room residents really have taken to this and to Nabla. Both GI hepatology and rheumatology have been heavy users of this, and I think that makes sense, because those are data-heavy fields where you may have a lot to review to get an accurate picture of your patient.

HCI: Have you surveyed users to get a sense of the impact on satisfaction with the EHR tools? 

Misurac: Yes, in the spring of this year we repeated the same Arch Collaborative survey we did in 2023 and saw a significant improvement in our net EHR experience score.

HCI: So could that improvement be from the impact of both Evidently and Nabla?

Misurac: It is from both. The Arch Collaborative collects comments from people, and both Nabla and Evidently were mentioned by our users, particularly the clinicians, as reasons for improvement. The nursing team also had a lot to say about how they liked our implementation of secure chat. But from a clinician point of view, it was the new AI tools that they mentioned most often, and we're attributing the change in our net EHR experience score to that.

HCI: Anything else you want to work on with these tools?

Misurac: Getting back to your earlier question about billing, there is a need to have the documentation reflect the complexity of our patients’ illness. I think it's pretty common that our documentation falls short in that regard, sometimes in small ways that have a smaller impact and sometimes in bigger ways. So if there's an AI tool that by reading the chart will be able to give you a more comprehensive note that doesn't miss any of the elements of the complexity of your patient, that certainly would have an impact on billing and the financial impacts there.

Also, we haven’t deployed it this way yet, but something we're working on would be using Evidently to comb through charts and pick out patients who would be good research candidates.