Chief information officers and other IT leaders at hospitals and health systems are being bombarded with a variety of AI technologies and are trying to navigate a growing market with a lot of hype.
In fact, advice on buying AI took up much of the morning at last week’s HIMSS 2024 Healthcare AI Forum, where digital health leaders compared notes on how to separate the signal from the noise.
Healthcare organizations must weigh the pros and cons of various platforms and tools — not all of which can always deliver on their lofty promises — and identify strategic partners that can help meet the challenges of integrating new AI technologies into their existing networks and workflows.
Tips for approaching suppliers
What digital transformation leaders want most from AI vendors is for them to be honest about their ingredients, according to Lee Schwamm, director of digital health at Yale New Haven Health System.
“First, you’re an AI company, and second, you’re a platform — it’s okay if you’re not either of those things,” he said at the Taming the Wild West of AI in Healthcare conference on Friday.
“You could say, I’m a company built on X platform and we’re starting to integrate some AI into our product – that makes sense. That’s probably what most of you are thinking.”
“You need to understand how you fit into the workflow because that’s the problem with point solutions that we have today,” adds Eve Cunningham, co-presenter and director of virtual care and digital health at Providence.
“In fact, integrating that into the workflow that we have now with the technology and infrastructure that we have is an incredibly complex process,” she added. “So you can have the best point solution in the world, but if they can’t integrate, and there’s no path to integrate, and you don’t know how to speak the language to understand that, then there’s just a very low level of interest in engaging on that front.”
Dr. David Newman, medical director of virtual health at Sanford Health, notes that it’s important to target pitches appropriately and find the right decision maker. “Trying to appeal to seven people is really counterproductive.”
He said he was talking to his 15-year-old daughter about his role in communicating with suppliers “and about the best way for them to contact us.”
“She said, ‘It’s like they’re sending me direct messages,’ which is true.”
The key, says Newman, is for technology vendors to know what the vendor’s mission is and what problem they’re trying to solve before reaching out.
“That way, instead of ignoring your email, I’ll respond to you,” he said.
“It’s not just about relationships and knowing someone,” Cunningham added, but also about being objective about the technology.
Many AI innovations are improving doctors’ lives and workflows, but testing them in partnership with vendors is causing “fatigue” for providers, she said.
Panelists encouraged providers to understand from the provider’s perspective – is it an improvement in physician productivity or is it something revolutionary that gives the health system something it never had before?
“Have you ever actually sat in a doctor’s office and looked at how many times they click? There’s no room for one more click,” Cunningham said.
“Sometimes they’re so good that we’re willing to break our workflow to accommodate them because it’s a superpower,” Schwamm added.
“You have a product that’s mature enough that we’re not going to build it for you,” Cunningham said. “We’re not your development shop.”
Supplier visibility, costs and backlash
There’s no simple answer to managing expensive product channels, says Schwamm: “Maybe I need someone who’s really partnering with me and building their product roadmap around my vision.
“At the present time, you are not qualified to [saying] “This will be my system forever,” and knowing that a technology module is “removable and replaceable” can be an advantage.
Cunningham admits that “there isn’t much need to remove and replace, but sometimes we have to,” such as with ambient listening technologies.
Digital healthcare leaders need to consider, “What will it look like in three years, five years, seven years?”
In her vision of what the tech-enabled doctor’s office will look like in the next five to seven years, a few things will happen simultaneously.
“There might be a big screen in the room, and there’s no keyboard,” she describes. “I’m talking to the patient. My notes are being compiled. All the things we’re talking about, ‘Hey, you have COPD,’ and all the data from the patient’s chart that’s relevant to COPD is coming up.
“‘Hey, we need an order to buy more PFT for you,’ and that order came in. ‘Hey, here’s a little bit of information on the next best course of action,’” she continued.
“Everything happens in the room, and when I walk out of the room, everything is done.”
To assess where vendors are at in their AI adoption, a recent HIMSS Market Insights survey looked at how healthcare organizations are using AI to deliver positive impact and explored the challenges they face when integrating AI into existing workflows and technologies.
Nicole Ramage, senior director of market intelligence at HIMSS, joined Schwamm to explore questions surrounding AI and insights from the report.
While nearly half of the organizations surveyed last spring were large organizations with 7,500 or more employees, the data showed, unsurprisingly, that “smaller organizations are less likely to be further along in their AI adoption journey,” Ramage said.
“I think your data makes it very clear that this is a capital-intensive process and also involves the ability to think about the workflows that you’re going to implement,” as well as the leadership structure that it requires, Schwamm said — noting that hospitals are “severely impacted by the devastating impact of COVID and the changes in age and population” and ROI is declining.
“It’s a constant downward trend, and the costs are a constant upward trend,” he said. “So it’s not a great formula.”
Ramage asked Schwamm what he sees as the biggest opportunity to transform AI in patient care and improve operational efficiency over the next three to five years.
“The easiest things to pursue from a transformation perspective are operational workflows or office work because they don’t involve patients,” he says. “They’re very low risk and relatively unregulated.”
While the biggest financial opportunities right now lie in office operations, there may also be growing employee backlash to the human-replacing impact of AI, Mr Schwamm said.
He notes that there are four ways to achieve ROI.
“Either you renegotiate the contract for a lower price or you cancel the contract,” he said. “You make it easier to do the same thing and do it at a lower cost, or you reduce your labor.”
AI will reshape the healthcare industry and its workforce.
“Whether you like it or not, this is going to be your next member,” said Sunil Dadlani, chief information and digital officer at Atlantic Health System.
Meanwhile, his co-chair, Charles Jaffe, CEO of HL7, said he was concerned about the politicization of the process. “The promise of AI is not a threat to anyone,” he said. “It’s a challenge to make their jobs easier.”
Still, Schwamm said small organizations are in a very vulnerable position. They can’t afford to be left behind.
Ramage asked Schwamm what approaches he recommends for smaller organizations to effectively drive AI adoption while maintaining employee engagement.
“If I were in a small organization, I would be capital constrained, and everything would affect me,” he said.
Without the in-house expertise and financial resources to hire consultants, he advises smaller organizations to partner with organizations they don’t compete with, then break away to conquer and validate the technology as a collaborative solution.
Schwamm suggests: “You have a group of five or six health systems, all of similar size to you, in different parts of the country, with no competition between you, and you say, ‘Hey, Jones Regional Hospital, how about you take care of this back office problem, and we’ll take care of the patient education problem?’”
“And then maybe even collective bargaining, right? Think about group buying opportunities.”
Data footprints and the law of exclusion
From a leadership perspective, the biggest challenge is who owns the AI data, Schwamm said.
“Before, if I worked with you and provided data so you could process transactions for me, at the end of the contract you would destroy my data or return all of it to me,” he said.
When data is used to train an AI model, “giving me back the data does not give me back the intellectual property you extracted from my data.”
When integrating AI into specific areas, “most mature healthcare systems will eventually have a dedicated change management function for AI because it is such a large part of change management projects and has a lot of labor implications.”
Dadlani noted during the ethics panel that healthcare accounts for one-third of the world’s data — and that data is growing, doubling every 46 days.
“So you need to apply more and more principles of technology and interoperability to really make smarter decisions about patient outcomes, patient safety and moving to the next generation,” he said.
This is where the rules of addition and subtraction come into play.
“Anytime you try to add more technology, you need to think and take a platform approach. Where can we simplify the process?” Dadlani asked.
As an organization continues to add technology, it increases complexity, he said, which can mean higher management costs, more data errors and more data breaches. If it’s a point solution and can’t integrate with other technologies, “it’s useless,” he said.
Andrea Fox is senior editor of Healthcare IT News.
Email: [email protected]
Healthcare IT News is a publication of HIMSS Media.
