AdventHealth’s AI chief outlines careful approach to tech integration in health care (Courtesy of the Jacksonville Business Journal) — Rob Purinton sees the potential for artificial intelligence to improve human interaction in health care.
Named the first chief artificial intelligence officer for AdventHealth in January, he helps oversee the rollout of the growing technology across the Altamonte Springs-based nonprofit health system’s national footprint.
That includes more than 81 current applications, from tools which help guide where to place patients to get the best radiology images to tools to help with note-taking.
One thing Purington has observed is, when he first started, paper charts were used for patients and this created jokes about doctors’ handwriting. New use of ambient audio and video tech is not only making it easier for those employees, but also allows patients to get more direct communication without the doctor having to turn away.
“The ambient AI technologies are helping that happen more often, where the technology fades into the background, but the conversation and engagement between physician and patient is still there and even better than it was, say 10 years ago.”
The system has hundreds of IT employees who help with the technology’s rollout and the amount of AI-enabled technology is also growing each month. The growth also comes as the U.S. Food and Drug Administration has approved more than 900 AI tools.
Here, Purinton spoke with Orlando Business Journal about what the implementation of the technology looks like at AdventHealth, the challenges when it comes to the rollout of AI and more.
How would you describe your role to folks who may not know how you apply AI?
It’s a new role and a new field. The technology is not all that new, because there have been lots of applications for it. For example, [for] our radiology systems it helps pinpoint where to get the best images or how to position the patient to get the best images of a particular organ like the heart. There’s a lot of smaller applications for doing specific measurements like how much does your heart move when it beats or how much blood moves through per second. What’s more new is for everyone else there’s been Chat GPT, which came out in fall of 2022 and changed the conversation around the uses of generative AI. That’s really had lots of imaginations go into overdrive on what sorts of problems in healthcare can be solved with more modern AI tools.
The role for me is to help figure out what are some of those appropriate uses and how do we integrate them in a safe and responsible way? What are the things that will improve the value proposition for our patients — mainly, what will help them get better outcomes in a better experience for a better value?
What has the ramp up for AI technology looked like at AdventHealth?
It’s really been about going slow so we can do it safely. Making sure we really evaluate every new type of technology we introduce for patient care. What has really ramped up more quickly are some of those administrative uses, the things that help with the back office parts of healthcare you don’t necessarily see as a patient. Those have been some of the uses the industry is eager to get into. Things we throw a lot of energy and manpower at like the billing process — there’s a bunch of people involved in putting together a person’s bill and communicating with insurers and the back and forth there.
Some of those administrative functions have gotten more attention — not because they are administrative, but because they are further away from direct patient care, so there are less things to worry about with how reliable a particular AI tool is. There are over 900 FDA-approved AI tools, [and] I think north of half of those are radiology and imaging space alone. Where we really expect to see the ramp up in the next months and years is areas like digital pathology. In the same way AI tools have helped radiology and radiologists take better images and give better diagnosis, we expect the same of digital pathology … [to] notice more cancer earlier, so it can be treated with less intensity and more patients can benefit and be saved. We have rigorous processes to test AI use cases and tools. We’ve partnered with a national coalition called the Coalition for Health AI (CHAI), so there’s a number of other large health systems that are working on this problem of “how do we evaluate tools that really give us confidence this will be safe and effective for patients?”
What major hurdles have come up?
We are really careful on how we do implementation and rollout. One of the main factors that is part of the rollout is the people part. It’s the change management and education and making sure doctors, nurses and other caregivers understand the tools we are putting in front of them, how to use them safely and in what scenarios and how to talk to patients about the kinds of AI that are being integrated into the care. We don’t see those as challenges or hurdles — they are just part of the work of scaling AI. That’s been part of our playbook for rolling out AI from the get-go.
What is the training like for this?
It depends on the tool and how much it integrates into the workflow. We really look for things that are fairly seamless in how they interact with, let’s say, a radiologist’s workflow — that it is really obvious how it adds value, that it is obvious what kinds of predictions are being made with AI. I would say when we have a broader rollout of AI, we go incrementally and slow. For example, there’s a tool that helps our physicians with note-taking called ambient documentation. There’s a similar type of technology being rolled out at hospitals and medical practices around the country. But we are going slow, one practice at a time, making sure the doctors get a lot of training and can test and tinker with it before they use it with real patients.
The key on some things like that, especially when they really change how our physicians and nurses work, is to go slow and incrementally. It isn’t always the excitement that you hear from AI vendors or companies. They sometimes talk about big, flashy rollouts that go quickly. While that is exciting to put on slides or to hear at a keynote, the reality is the change is really deliberate and incremental.
What does the long-term roadmap for AI look like in the future at AdventHealth?
I’m a technophile, so I have a lot of optimism about where AI and technology will go in health care — otherwise, I would not have taken a job in this. Really, we have to stick to our principles around our consumer-centered, whole-patient care first and make sure we are being methodical about evaluating each use case as it comes across. Where it is going in the future, we are seeing new data inputs going into it.
Things like facial recognition [and] how does that help us better care for patients and understand what may be going on with them. We’re seeing new scanning technology, like for example at an eye doctor you may get a retinal scan to get your prescription, but it turns out those scans might be able to diagnose long-term chronic illness — so is that something that adds to the value of primary care? For new use cases, sometimes it’s finding existing technology that can be used in new ways and sometimes it is developing new technology that can advance the overall goals of outcomes and consumer experience.
We’re fortunate to be in this era where so much new technology is coming at us, but also in an organization that can go slow and really be deliberate on how it chooses how things will make it to physicians and patients.
English
Haitian Creole
Spanish