SAN DIEGO – A common theme at the HIMSS AI in Healthcare Forum last week was that artificial intelligence represents a new paradigm shift for the way healthcare is delivered (and paid for) and that with such a rapidly emerging technology, we're all figuring it out together.
As the two-day event came to a close on Friday, three leaders from three different facets of the healthcare industry each offered their own perspectives and experiences to date on how to efficiently and effectively leverage (and scale) the promise of AI. .
In a discussion moderated by HIMSS Chief Research Officer Anne Snowdon, IT leaders from Providence, VMware and Humana offered some real-world lessons for deploying and expanding AI models.
Tariq Dastagir, AVP of medical informatics and clinical trends at Humana, set the stage.
“Margins are thin and the pressure on us to reduce healthcare costs is high every year,” he said. “And that obviously has to be done with better results. It has to be done efficiently. The hope is that we can use a lot of these new technologies to do that.”
A key question, he added, is “what is the right use case and business case to deploy this, and where can you get by with simpler solutions? Not everything needs a genAI model or a protective model. Sometimes it can just be your simple solution.” risk score that you can use to predict something.”
As more healthcare organizations embrace the promise of AI, they must be prepared to grapple with these questions and think critically about them.
“Look for differing opinions and not agreement, because that's where you really learn,” Dastagir advised. “We get excited about a lot of things. But the point is, what's really going to make the real difference and how does everyone else feel about it and are they on board, do they feel the same way? And if not, how do they do that? Do you bring them around ? Or how do you learn how to evolve your use cases to the point where it starts to make sense for everyone?”
Corey Lyons, senior staff solutions engineer at VMware, agreed – noting that it is easier to fit AI's next-gen capabilities into existing technology processes and workflows than to implement them.
“We talk about infrastructure, we talk about technical debt, we talk about the intersections of the technologies that will work alongside analytics, and these other proven business processes,” he said.
“We want to help our customers understand, 'Look, you've been running things in this very traditional, well-understood way,'” Lyons said. “We're in a transition to a more agile way, where the applications will iterate more often.
“What's exciting and challenging for us today is when you look at large language models, all these other processes that require a tremendous amount of horsepower to generate… when you try to substantiate that into, okay, how do we do today the day's business?” and how can our teams be successful? There's a big departure from, “We can actually do this repeatedly, successfully, safely, with the (necessary) degree of automation security.”
As VMware looks to the future, he says, “we're trying to help organizations say whether it's a private cloud that you're hosting, you're working with a hyperscaler or you're deploying these solutions to the edge – where I honestly think this is about a will have the biggest impact in a few years – collectively we will hopefully all be able to use these things,” said Lyons. “We are one of the few organizations that can help everyone every step of that journey with an eye toward, 'This is how the older applications and older processes meet the newer techniques and capabilities.'”
At Seattle-based Providence, a longtime IT innovation leader, AI-based tools are already being deployed in a variety of clinical and operational use cases.
“We have Nuance's DAX product, and over 1,500 providers use it,” says Eve Cunningham
head of virtual care and digital health at Providence. “We also have a digital assistant and a clinical content management product called MedPearl that we developed and incubated at Providence, which scales and has over 7,000 users. We also use generative AI to help with inbox management. “
Lessons have been learned across all three of these applications, she said
The first thing I would say, I know you were on it too, is that we need clinical sponsorship and executive sponsorship. It is absolutely crucial that the coordination is there.
You need to make sure you understand the problem you are trying to solve. You can define it and put it into words. If you can speak the CFO's love language, measure ROI and KPIs, and be very steadfast in how you're going to measure that as you start to scale things up.
I know some people say they don't believe in pilots, they just want to go straight to scale. You can't always do that. You want to do pilots – what you don't want to do is do perpetual pilots. So you have to be able to say, 'Hey, we're going to fail.' And we've done that before. We said, 'Hey, this isn't working. We are going to stop using this application or working with a supplier because things are going wrong.' You must be able to do that.
You also need to make sure that what you're trying to solve aligns with the key strategic priorities for the healthcare system or organization you're working with.
For example, at Providence, in my division, our three top priorities are:
-
Staff shortages and burnout
-
Hospital throughput and capacity, because we don't have hospital beds and so we have to figure out how to digitally enable hospital capacity, deploy it virtually or treat patients in rural hospitals that don't have to go to large hospitals, by virtually enabling rural hospitals to provide specialist care to provide
-
And fragmentation of care
So when we look at the different solutions that we're evaluating, we're thinking in the context of those three big pain points, which I know are not unique to us.
Then the other thing, once you realize, okay, this looks like this is a problem that we're solving, this is going to hit a huge pain point, it's strategically aligned.
Okay, what is the feasibility of the solutions out there? What is the maturity of the solutions? Is this something we have to build ourselves? Can we buy it? Can we work with someone to develop together?
And furthermore, how does it fit into the workflow? And is it possible to integrate and merge it into this workflow? Because it could be the best idea in the world, that would solve the world's biggest problem.
But if my doctors have to make 17 clicks to use it, it won't be adopted. And what is the question of the end users? Will you hear from the people to whom you will actually try to push this solution? What is the question? How does it fit into the workflow? And what's the lift for them to train on it, to adopt it? What is the change management aspect if you have to do that?
And then of course you discuss the risk, bias and safety. Do you have the right infrastructure?
I'll give you an example. We have a huge demand from radiology to bring some of these algorithms to our radiology department so that we can leverage the FDA approved algorithms that use AI to accelerate and optimize the physician workflow and to to be able to read images, but also about the quality.
But we have 27 different PACS servers. We do not have any infrastructure middleware that can connect Nuance's network to our PACS servers. So the technical debt and building the infrastructure is going to be necessary to actually be able to connect the AI to the workflow, and with the system there's a pretty significant increase.
It doesn't mean we won't do it, and it doesn't mean there isn't a desire. It's like, how do we connect the dots?
So those are all things that we're thinking about and we've recently put a governance structure in place. I co-chair a clinical AI working group. Describe kind of our guardrails and the evaluation process that we work through and setting up and creating the muscle
There is no shortage of good ideas, but the question is: how do you filter and set priorities? And what are the things that seem like common sense that you think you can make happen.