Presbyterian CMIO on Bringing Gen AI Copilots to Rural N.M. Practices
Lori Walker, chief medical information officer at nonprofit Presbyterian Healthcare Services in New Mexico, said her team had been working to build a custom care pathway tool for diabetes patients in Epic, but then switched gears and began piloting generative AI copilots being developed by a company called RhythmX AI to help with chronic care management. Walker recently spoke with Healthcare Innovation about the pilot effort under way at Presbyterian.
Healthcare Innovation: Before we talk about generative AI, could you tell us a bit about your background?
Walker: I am chief medical information officer, I've been with Presbyterian for about 20 years. I'm a nurse practitioner by background. Presbyterian is an integrated delivery system. We have our own health plan.
HCI: Does Presbyterian face some specific challenges operating in a lot of rural settings?
Walker: Yes, we’re a very rural state. We have our central delivery system, which is really in the Santa Fe, Rio Rancho, and Albuquerque area. Then we have some very small hospitals and clinics in small towns throughout the state.
HCI: Were there some issues that your clinicians were facing that led you to connect up to RhythmX AI?
Walker: Social determinants of health are a really big thing for our clinicians here in rural New Mexico. We’ve got a lot of poverty and diversity. We have to focus on that on top of all of the complex chronic conditions that our patients have.
Like a lot of other health care systems, we were seeing provider burnout, and we're seeing access as a struggle. We'll have a patient who may drive two hours to a primary care appointment, and they have a 30-minute slot. They may have seven problems on their problem list, and for each problem there may be two or three medications and two or three labs. We also want to stay ahead of quality screenings such as mammographies. Then we add a layer of the social determinants of health such as food insecurity, transportation insecurity, and financial issues. The providers have 15 of these patients throughout the day, and it's a lot of cognitive burden.
When we looked at the way our providers were spending time — doing clinical documentation, but also after hours getting ready for the next day in chart prep, chart search, chart review, identifying care gaps, making sure they have all those orders ready to go, that was all very time-consuming. So when I saw RhythmX I thought here's a nice summary that really highlights the things you need to focus on. And here are some recommendations around their medications. And you can see those labs trended very nicely. And then you have this AI assistant that you can ask things like formulary questions.
HCI: Was there an area of focus you were seeking to address first?
Walker: We were starting on a care pathway for diabetes. We have a robust Native American and Hispanic population here in the state, and see quite a lot of diabetes.
We tried to do a custom build in Epic and worked pretty rigorously on it, but just couldn't bring it to life. But then I saw RhythmX, and they were doing this for four chronic conditions — diabetes, hypertension, heart failure and COPD. It was really what we had tried to develop with an Epic build, but we didn't have the AI component. When I saw this from RhythmX, I was like, 'Oh my gosh, this is exactly what we've been trying to develop for one condition.’ It not only surfaces what medications you should prescribe; it also has the ability to look back at what you have tried in the past that maybe the patient didn't tolerate. Maybe they have an allergy or had severe side effects or something. So it is really trying to be hyper-personalized to the patient vs. just looking at rules or criteria.
HCI: And does their solution fully integrated into Epic?
Walker: Yes, that was the other really big piece of it. Our providers really want to be within Epic. Like most organizations, they don't want things to change their workflow. We all like the muscle memory and we have our habits in the way that we use the tools. This is very much within the workflow, which was really exciting as well.
HCI: Did you initially pilot with one clinic or hospital to see what would happen?
Walker: We are very much in our pilot now. We are truly an alpha partner with RhythmX. We are developing as we go, which is really exciting, and I think it's new for organization and for our pilot providers. We've done other pilots, but we've never been the alpha pilot site. Currently, we have it in five of our primary care clinics. I've kept it very small to start. We have it with nine providers, a mix of physicians and advanced practice providers.
HCI: What kind of initial feedback have you heard about having these recommendations surfaced? Does it have to be fine-tuned?
Walker: It definitely had to be fine-tuned. I’m not going out trying to find a problem, right? I want a clear, defined problem that we have, and then ask whether AI is the right solution. The other thing that is a big focus for us is that it's not about using AI to practice medicine. It’s there to support us. It's not there to make the decisions. We've got a mix of clinicians where some have really leaned into AI and others are late adopters, which is what you want. This is truly to reduce the cognitive burden.
For every suggestion that the AI brings up on obtaining additional labs or titrating medications up or down, or adding an additional medication, the clinicians can give a thumbs up or a thumbs down. That’s why being a pilot provider is a little bit time-consuming, but we have to have that feedback so that we can learn from it, and we can understand why it was appropriate or wasn't.
I know my primary care providers really well in my organization, so I went with some who are super-innovative early adopters and with a handful who are skeptical. They're going to question it, they're going to challenge it. No matter what the AI brings up, they’re going to take the time to research, to see if this is truly accurate.
HCI: This sounds like it makes perfect sense for primary care, but could it also work in specialist offices, too, or even in an inpatient setting?
Walker: We're definitely thinking about that as we're looking at developing the partnership with RhythmX AI. We clearly have the same access issue with our specialists. We're always trying to recruit. As I think about this tool, we might get some of those physicians who are fresh out of their residency programs. How do we support them? We need them to come in and take a panel of patients, and get up to speed quickly. Could RhythmX help support them? Also, could we help guide the appropriateness of referrals? If we already have access issue with our cardiologists, how do we get in front of the primary care providers to help them make appropriate referrals based on the evidence?
HCI: So what’s next on your agenda?
Walker: The hope is that over the next four to six weeks we will continue to expand out in the primary care space as we assess whether we're ready to go enterprise-wide with this pilot. We've got 15 primary care clinics in our central delivery system. At what point do we feel like we're ready to go big bang, at least in the central delivery space?
