Chief information officers and other IT leaders at hospitals and health systems are being bombarded with a variety of AI technologies and trying to navigate a fast-growing market amid a lot of hype.
In fact, at the 2024 HIMSS AI in Healthcare Forum last week, artificial intelligence procurement advice took up much of the morning, as digital health leaders exchanged ideas on how to separate the signal from the noise.
Healthcare organizations must weigh the pros and cons of different platforms and tools, not all of which necessarily meet their lofty expectations, while simultaneously identifying strategic partners that can meet the challenges of integrating new AI technologies into existing networks and workflows.
Tips for approaching providers
According to Lee Schwam, chief digital health officer at Yale New Haven Health System, what digital transformation leaders want most from AI vendors is honesty about what it consists of.
“Number one, you’re an AI company, and number two, you’re a platform. It’s okay if you’re not one of those,” he said on a panel on Friday about taming the AI wilderness in healthcare.
“It makes perfect sense, as it probably does for most of you, to say, ‘We’re an X-based company and we’re starting to build AI into our products.'”
“The problem with current point solutions is you really need to understand how they fit into your workflow,” added co-panelist Eve Cunningham, head of virtual care and digital health at Providence.
“In fact, integrating that into workflows with your current tech stack and infrastructure is a very complicated process,” she added. “So even if you have the best point solutions in the world, if you can’t integrate it, if you don’t have a path to integration, if you don’t know how to speak the language to understand it, there’s going to be very little interest in working in that space.”
It’s also important to properly target your sales pitch and find the right decision makers, says Dr. David Newman, chief medical officer of virtual care at Sanford Health: “If you’re trying to get seven customers, that actually backfires.”
He said he had spoken to his 15-year-old daughter about his role in speaking with vendors and “about the best way for them to contact us.”
“She said, ‘It’s like he’s sending me a direct message,’ and she’s right.”
The key, Newman said, is that technology vendors need to know what a provider’s mission is and what problem they’re trying to solve before reaching out.
“That way, I can respond to your emails instead of just skimming through them,” he said.
“It’s not just about relationships and who you know, but rather it’s important to objectively evaluate the technology,” Cunningham added.
Many of the AI enhancements address physician viability and workflow, but partnering with vendors to conduct testing has caused “pilot fatigue” among health care workers, she said.
Panelists urged vendors to understand it from the provider’s perspective: Is this about improving physician productivity or is it something innovative that brings something to the health system that hasn’t been done before?
“Have you ever actually sat in a doctor’s office and seen how many times they click? There’s no room for one more click,” Cunningham said.
“Sometimes, their capabilities are so good that we’re willing to interrupt our workflow to accommodate them, because they’re super-powered,” Schwam added.
“Your product is mature enough that we’re not going to build it for you,” Cunningham said. “We’re not your development company.”
Provider Vision, Costs, and Reaction
There’s no easy answer to managing a costly product pipeline, Schwam said: “Maybe I want someone who’s really going to collaborate with me and create a product roadmap that aligns with my vision.”
“Right now, I don’t have the luxury of saying, ‘This is going to be my system forever,'” so knowing that technology modules can be “removed and replaced” could be an advantage.
“There’s not much appetite to rip and replace, but sometimes with things like ambient listening technology you have to do that,” Cunningham acknowledged.
Digital healthcare leaders need to think, “What will happen in three, five, seven years?”
In her vision of a technology-driven clinic five to seven years from today, several things are happening at once.
“There may be a big screen in the room, but there’s no keyboard,” she explained. “I’m having a conversation with the patient. My notes are being drafted. Everything we’re talking about — ‘You have COPD,’ and all the data related to COPD pops up from the patient’s chart.”
“I say, ‘We need to order more PFTs for you,’ and the order comes in. ‘Here’s a little bit of information on the next best course of action,'” she continued.
“All this happens in the room, and when I leave, it’s all done.”
To determine where providers are currently at in their AI adoption journey, a recent HIMSS Market Insights survey explored AI usage in healthcare organizations to uncover its positive impacts and the challenges they face in integrating AI into workflows and existing technologies.
HIMSS Senior Market Insights Manager Nicole Ramage joined Schwamm to discuss questions surrounding AI and insights from the report.
Nearly half of the organizations surveyed this spring were large organizations with more than 7,500 employees, but as expected, the data showed that “smaller organizations are less advanced in their AI adoption efforts,” Ramage said.
“I think the data really shows that this is a capital-intensive process and also the ability to think about the workflow that you pursue and the leadership structure that it requires,” Schwam said, noting that hospitals are “severely cash-strapped by the devastating impacts of COVID and age and demographic changes” and that ROI is worsening.
“This is a continuing downward trend and costs are continuing upward trend,” he said. “So this is not a good equation.”
Ramage asked Schwam what he sees as the biggest opportunities for AI transformation in patient care and operational efficiency over the next three to five years.
“The best places to go from a transformation perspective are non-patient workflows and back-office operations,” he said. “The risks are very low and the regulations are relatively light.”
Schwam said the biggest economic opportunity right now is in back-office work, but there is also growing employee backlash against the impact of AI replacing humans.
He noted that there are four ways to get ROI.
“Either we renegotiate the contract at a lower price, or we cancel the contract,” he said. “Either we make it easier and cheaper to do the same thing, or we cut our workforce.”
AI will reshape healthcare and its workforce.
“Whether you agree with it or not, that’s going to be your next member,” said Sunil Dadlani, chief information and digital officer at Atlantic Health System.
Meanwhile, Charles Jaffe, CEO of HL7 and a co-author of the paper, said he was concerned about the politicization of the process. “The promise of AI isn’t a threat to anyone,” he said. “It’s about making their jobs easier.”
But smaller organizations are in a very vulnerable position, Schwam said: They can’t afford to fall behind.
Ramage asked Schwam what approach he would recommend for smaller organizations to effectively accelerate AI adoption while maintaining staff commitment.
“If I’m in a small organization, we have limited funding and the impact is big,” he said.
Lacking in-house expertise and the funds to hire consultants, he advised smaller organizations to partner with non-competitors, divide and conquer, and jointly vet technologies.
“You could have five or six health systems your size in different parts of the country, not competing with each other, and say, ‘Jones Regional Hospital, you take on the back-office issues, and we’ll take on the patient navigation issues,'” Schwam suggested.
“And maybe collective bargaining will become possible. Think about group purchasing opportunities.”
Data Footprint and the Law of Subtraction
From a leadership standpoint, Schwam said the biggest challenge is who owns the AI data.
“Previously, if I worked with you and provided you with data so that you could process transactions for me, at the end of the contract you would destroy my data or give it all back to me,” he said.
If the data is used to train AI models, “returning the data does not mean returning any intellectual property extracted from the data.”
When it comes to integrating AI into certain areas, “most mature health systems will have some sort of change management function that’s dedicated to AI, because AI is a big part of the change management project and has a big impact on the workforce.”
Dadlani told the ethics committee that healthcare accounts for a third of the world’s data and is growing and doubling every 46 days.
“So we need to bring in more technology and interoperability principles to make smarter decisions about patient outcomes and safety and move that into the next generation,” he said.
This is where the laws of addition and subtraction come into play.
“Any time you’re trying to add more and more technology, you have to look at and adopt a platform approach. Where can you simplify the process?,” Dadlani asked.
As organizations continue to add more technology, it adds complexity, which means higher management costs, more data failures and more data breaches, he said. If it’s a point solution and it can’t be integrated with other technologies, it’s “useless,” he said.
Andrea Fox is a senior editor at Healthcare IT News.
Email: afox@himss.org
Healthcare IT News is a publication of HIMSS Media.
