One concern about the rapid deployment of artificial intelligence in healthcare is that it could lead to a two-tiered system where smaller and rural community hospitals fall behind because they don’t have the same kind of tech resources. In a recent interview with Healthcare Innovation, Mark Howerter, M.D., chief medical officer for Columbus Community Hospital in Columbus, Neb., and Mark Pratt, M.D., chief medical officer for EHR vendor Altera Digital Health, spoke about how emerging AI tools can help rural providers reclaim precious time.

Healthcare Innovation: Dr. Howerter, could you first describe Columbus Community Hospital?

Howerter: We’re a 50-bed hospital in a community in the eastern half of Nebraska. Our population is about 25,000, and our catchment area is probably about 38,000. We have larger hospitals in our area, but they aren't right next door. Omaha has several tertiary care hospitals about 90 miles away. Lincoln has a couple of tertiary care hospitals, also 90 miles away. 

HCI: Do you have a chief information officer who helps makes decisions about EHRs and other technology?

Howerter: We do not. I think that's being bandied around right now. The vice president overseeing our IT department is our CFO, who has a great deal of interest in informatics, but we don't have a CIO.

HCI: Which EHR platform do you use?

Howerter: We’re using Paragon, hoping to go to Denali soon. 

HCI: Dr. Pratt, could you talk about some of the reasons why you think keeping up with these burgeoning developments in the AI world is as important for rural hospitals as it is for academic medical centers and larger health systems?

Pratt: I think a lot of these community hospitals, from a staffing perspective, their healthcare is just as complex. They could all really use a CMO and a CIO and all of those things. But you've got a lot of people who are being shared resources. Sometimes you'll see the CFO acting like a CIO, or other people wearing multiple hats. I think it becomes more important that you use technology to your advantage, to offload some of those more mundane  tasks as much as possible, so they can expand their own role. I think it just opens up capacity for those community hospitals that are already resource-constrained to do more.

HCI: Is Paragon Denali EHR specifically designed for rural and critical access and community hospitals?

Pratt: I would say that's our specialty area. Our average bed size is 85. We do have some larger customers, but for the most part, our clients are serving in a community hospital environment or rural healthcare area. So our EHR was designed to be comprehensive enough to be the EHR for a community hospital — from admission all the way to discharge and aftercare.

HCI: Is the cloud-based nature of it valuable to hospitals in those settings?

Pratt: We did this whole strategy with the cloud, because a lot of the community hospitals don't have the staff to maintain a data center, for example. If you're running our platform in an on-premise environment, that takes a good bit of hardware and hardware maintenance people to make sure it's up and running at 2 in the morning. That's a pretty large resource commitment. So if we can offload some of that burden from the customer, it’s just a bonus for them. In addition, the way we designed the system is based upon making sure it had a very robust security infrastructure and that it can be geo-redundant as well. 

HCI: Dr. Howerter, have you been reading about or seeing some of these AI features, like the ambient listening and note generation solutions, and is that something that's appealing to your clinicians?

Howerter: It is. We know that there's going to be an ambient listening component offered with Denali, but we have clinicians who are clamoring for ambient listening already, so we're using Nuance’s DAX for some of our clinicians right now. It's kind of making them subject matter experts on ambient listening, and we are going to be looking for their feedback when go with Denali. And it seems like there are certain clinics that really lend themselves to it, but in general, it seems like it's a winner. 

HCI: I’ve been interviewing people who are working on AI solutions that offer chart review. So let's say you're a hospitalist, and you get a patient coming from the ER, it pulls together and summarizes the basics of the patient's situation. 

Howerter: There are multiple scenarios where that would be helpful. Also, real-time abstracting. One of the things we drive ourselves nuts with is things like trying to keep up with sepsis measures. We get our charts abstracted after the fact, and real-time abstracting would be so much more valuable, because it doesn't do us any good to find out six weeks later that we missed a single measure and therefore missed the entire measure. Also, why in the world couldn't AI go through a medical record and generate a discharge summary? All the information is there, and all it needs to be taught is what salient features need to be in that discharge summary, right?

Pratt: Ambient listening is what everyone's been talking about first. But then that becomes your launch point for all kinds of other things you can do. With Denali, we start with that ambient conversation and take it a step further. We can parse out the different elements of the note — your review systems, your physical exam, your plan of care, and just based upon that conversation, it puts it in the correct portions of the note to make it look clinician-friendly. Now that you have access to the entire medical record, then why can't you use those insights into making suggestions — like, “hey, you mentioned this patient has a blood sugar of 250. Do you want to add diabetes with this ICD-10 code to your problem list? And it can start making smart recommendations about problems. Or it can say you mentioned you wanted a chest X-ray. Do you want to cue that up as your order?

Howerter: And what Dr. Pratt mentions is even more important in a community hospital than in a larger tertiary center. When my hospitalists are taking care of a patient who might need some expertise with nephrology, pulmonology, gastroenterology or hematology, they can't just write for the consult. Wouldn't it be nice to have, for instance, hematology-based recommendations to which they could either respond, yes that's still within my wheelhouse. I'm going to accept that. Sometimes that consult just needs to answer one or two really important questions, and that's it.

Or maybe I do need to transfer this patient out, but at least that would be some assistance that a community hospital can make much better use of than a tertiary center where you have a specialist readily available.

HCI: We are told that AI solutions are better when they have more complete data. Dr. Howerter, does your hospital face challenges with accessing or sharing patient data that's coming from external sources?

Howerter: Probably one of our biggest ongoing headaches is the fact that there are a lot of systems out there, and there's not a lot of interoperability. Typically everybody in Lincoln and Omaha's on the Epic system. If we had Epic, then we could dig right into their record, but we can’t, so that is very limiting. Even in our community, there are multiple  EMRs. 

HCI: Dr. Pratt, you described this direction about adding some of these AI features with Denali. Is some of that already happening with other customers?

Pratt: The ambient scribe platform we're in the process of implementing across our first round of installs, and that's where we're putting in these AI features. We've got multiple other AI initiatives going on around different topics. For example, Dr. Howerter mentioned dynamic abstracting and we are looking for opportunities there because once you have access to the data, why not say, hey, we noticed that you did a bedside procedure, but you didn't charge for it. It's documented in the chart, so you need to add this billing code before you submit the claim. The longer it takes to get the claim correct, the longer it takes you to get paid, which is so critical to these community hospitals. 

HCI: Dr. Howerter, do you think that deploying AI could be seen as helpful with recruiting and making the hospital more attractive to younger clinicians?

Howerter: I had a very interesting conversation with the CIO at Nebraska Medicine. We partner with Nebraska Medicine on telestroke and some other things, so we have reasons to talk to them about those kind of initiatives. We discussed that we're noticing an AI Wild West happening right now, where there are a lot of products out there that aren't necessarily attached to anything, but claim they can integrate into your EHR. And maybe they can, maybe they can't, but everybody's wanting to sell the AI tool they have. 

I think you have to be really circumspect about which products you use. Obviously, most of us would like to see our AI products integrated into our EHR, pre-vetted for security reasons, and also for ease-of-use reasons and for integration reasons.

HCI: Dr. Pratt, health systems are trying to figure out how to work with these AI startups, but even as an EHR development company, do you have to decide how much work to do internally vs. partnering?

Pratt: That's a longstanding issue. When do we develop vs. partner? But AI is helping us internally open up capacity. We’re actually able to use AI to more quickly get code done and QA code. So that's helping us deliver features faster to the customers. There's that whole side of AI as well. If you've never seen AI do code assisting, it's mind-boggling.