In his role as chief digital and information officer at ECG Management Consultants, Nathan McCarthy has a close-up view on the differences between what large academic medical centers and small community hospitals are able to do with AI innovation. He discussed with Healthcare Innovation how he helps small to mid-size health system customers govern AI technologies, redesign workflows and identify where AI may not be worth implementing.

Healthcare Innovation: Let’s talk about some of the differences you are seeing between community hospital systems and large academic medical centers when it comes to AI deployment. What stands out the most as far as how different the approach is?

McCarthy: The smaller groups certainly have smaller budgets and fewer resources, so they've got to be a little bit tighter in terms of what they're they're going after. The groups I'm working with in that space are trying to be that first follower with certain things, but they have a pretty heavy focus on workforce productivity and and things that they know will have a meaningful ROI to them in in the short term, as opposed to diving into a lot more innovation. When they are innovating or doing things that don't necessarily have a playbook already, they’re making sure they have a strong business case behind it.

HCI: Do you see them doing a lot of things with either revenue cycle or patient engagement?

McCarthy: Yes, there’s definitely a lot of that. There’s also a workforce component. There's some general administrative stuff that they were able to knock off. They say, we know we can get the ROI in those spots, but how do we do it more efficiently?

HCI: Are the community health systems more or less willing to work with startups? And do they tend to be more or less reliant on their EHR vendor in terms of the pace of AI development?

McCarthy: It's a mixture. I think there are certainly a lot leaning into their EHR vendor — especially those that are on Epic. I think I've seen that a little bit with the groups that are on Oracle. Meditech is probably a more prominent EHR in the small hospitals, so it certainly is different there. Its customers are probably more willing to talk to some of the startups, especially when the startups are willing to go a little bit more at risk or have a different creative cost structure because they know they've got more of a playground with this community health system than they're likely going to get with an academic health system to try to get that proof point, if you will.

Some groups go talk to these startups and less mature IT vendors, and then they're immediately turning around and going in to talk to Epic. Do you have this? Where is this? It then turns into a roadmap decision. One thing I'm definitely seeing is a stronger push to have much shorter contracts or use-based contracts, and the vendors are having to adjust. That’s also part of the SaaS world in general. But what used to be three to five years as a normal agreement, groups are wanting to have a two-year contract or a one-year contract, or purely off of usage.

HCI: We see a lot of the large health systems appointing somebody with a title like chief AI officer, but probably not so much in the community health systems. So who are the decision-makers on AI? Is it the CIO or the CMIO in those cases?

McCarthy: It is. Many have talked about a CAIO role, but for most that still lives within the CIO/CDIO/CMIO landscape. When they hit certain business areas, you will have an AI lead in revenue cycle or finance, or their patient access center has a liaison to the technical team.

HCI: Coming in as a consultant, do you help them figure out how to redesign workflows or how to identify where AI might not be worth implementing?

McCarthy: We do. There’s both the build vs. buy decision framework as well as where to invest and where not to invest. We try to anchor them: What’s the business need we're trying to solve for? And are the AI solutions out there mature enough? With the community hospitals, some of that goes into how mature your group is, which is a different question than the Cleveland Clinics or the Johns Hopkins of the world face, right? 

They might have freed up some productivity in their IT team and ask how can I redeploy some of this to start taking on some of that innovation? Or they're seeing rises in certain vendor costs. Might they be able to do something themselves because that vendor isn't meeting the need?

HCI: Do you see examples where community hospitals have invested in AI pilots without a clear strategy and maybe it has involved some kind of financial risk and now they're trying to figure out what to do about it? 

McCarthy: Most certainly. I think we've also been called in when folks were expecting AI to lead to a pretty significant cost savings or labor change, and that hasn't transpired. We are asked to help them either actually meet that goal or reconfigure the AI or get out of that solution altogether.

HCI: Are you seeing some small health systems sitting on the sidelines so far regarding AI?

McCarthy: There are a few of them. We definitely have a handful of clients that are taking a much more cautious approach or they're spending a lot of time trying to make sure they've got governance down before they move forward.

HCI: We’ve been talking about whether there's ROI on some of these projects. Are health systems pretty good at setting up the right kind of metrics to tell whether they actually are getting a return on their investment here?

McCarthy: There are certainly groups that are measuring that very well and tracking it. There are other groups that — at least when we're involved — are looking at that up front, but I don't have a whole lot of faith that they're going to be monitoring that to a point of saying, "Hey, we need to cut bait or pivot or optimize from here because we're not seeing the return we were expecting to.”

HCI: Without naming names, can you think of an anecdotal example of a community health system that's done some really interesting things in the AI space?

McCarthy: There’s a group in the Northwest that I think has done a really good job of picking their spots with some of the basic return-on-employee components. They’ve really started to dig into measuring what they're doing. Their focus is on rolling out a lot more agents in their call center, and they are watching the metrics closely to see if they are getting what they expected in terms of productivity and slots being filled. They also study the patient satisfaction. How often are they getting out of the AI parts of this and getting to a human again? I think they've had some good governance established around AI.