BANGOR, Maine — Rural health care providers, which have struggled for years, are facing nearly $1 trillion in Medicaid cuts over the coming decade. The Trump administration is pointing to a cutting-edge, if unproven way to save them: artificial intelligence.

“The best way to help” rural communities, is “AI-based avatars,” Mehmet Oz, who leads the Centers for Medicare and Medicaid Services, said at an event on mental health this year.

His comments are among the grand promises administration officials have made about AI. They’ve also touted the prospect of concierge care for every American and AI nurses as good as any doctor, for example. Their argument is that AI, fueled by the $50 billion rural health transformation fund that came along with the Medicaid cuts, could remake rural care in the U.S. 

The technology could lower the cost of delivering care, Oz has said, and could be a deflationary force across the health care economy.

But providers serving rural communities aren’t so sure.

“I don’t think AI is going to make health care costs go down in the near term at all,” said Lori Dwyer, president and CEO of Penobscot Community Health Care, a system that stretches across Maine. “I don’t think it’s going to save us.”

Her system has already started using ambient scribes, an AI technology that listens to a clinician’s conversation with a patient and then helps document it in the electronic health record. While that’s had plenty of benefits, saving money isn’t one of them, Dwyer said.

“It’s allowed us to off-load work, administrative burdens from our providers and our care teams so that they can focus on patients and patient care and so that they don’t burn out,” she said, adding that documentation time is down, and communications are improved. “But that doesn’t reduce costs.”

Other ways the system is using AI, such as for patient communications or to aid in remote patient monitoring, have been helpful, she said — but only create marginal economic efficiencies at a time when years of work to stabilize the system’s finances is essentially being wiped out by the looming cuts.

A troubling reality on AI

Across dozens of interviews with rural health providers, hospital system leaders, and health AI experts, a troubling reality was repeated: Investing in AI technology would be costly, especially for already at-risk systems, and its savings potential hasn’t yet been proved. Despite the tools holding promise, they’re expensive — and buyers are facing a wave of new financial pressures.

Rural health system leaders are excited about the technology’s potential in the distant future and feel the need to invest in it in order to keep up with the nationwide advancement. 

But the large-scale, system-altering changes that promise huge savings and a fundamental shift in health care delivery seem well out of reach for many rural health systems. 

Health AI leaders and rural health providers alike are concerned those gains will go largely to the big systems with money to invest in AI-native infrastructure — a costly, time-consuming endeavor. In the end, the technology could increase health inequities in the U.S. instead of shrinking them, they worry.

Despite the challenges, some rural health leaders see adopting AI technology as their only way forward in an exceedingly difficult environment. 

A few system leaders told STAT they’re accelerating their adoption of AI tools — because of the cuts — even beyond what they were doing at the start of the ChatGPT age. Finding new efficiencies has become a mandatory task, and more than 70% of hospitals are estimated to be using predictive AI, according to one survey, and companies that create electronic health records software are grappling with how to make the most of the technology.

“We’re accelerating it probably faster than I thought we would,” said Trampas Hutches, the Mountain Region president at MaineHealth, who noted the technology was one of the top priorities in MaineHealth’s three-year plan. “The only way to meet the demand is technology and AI.”

Hutches said the system is in the process of making AI a “care team member,” as well as using the technology for ambient notetaking and in administrative work.

“We’re really trying to push the envelope on acceptance of AI as a care team,” he said, though he added that the cost of the technology is a concern. “We’re doing everything we can to make sure our cost structure does not go any higher than it is today.”

AI’s rural limits

Medical technology doesn’t have to be on the cutting edge to come up against challenges in rural America. Telehealth, for instance, which often uses tech from a decade ago or more, still has trouble getting traction in remote areas at times, because internet infrastructure can be weak or missing — or because the regulatory and payment structures are still unsettled.

AI faces those problems — and more. 

Low access to broadband in Maine, for example, could keep internet-dependent systems from being reliable, said hospital leaders in the state. Others may not have the data infrastructure needed to take advantage of new systems: Forms that are faxed or handwritten aren’t always easily entered into digital systems.

Even some digitized systems aren’t easily or reliably digestible by AI systems. Even advanced hospital systems have had to undergo lengthy, tedious reorganizations to prepare for the technology.

“If you’re employed by a big health system, you’re much more likely to have that opportunity,” James Jarvis, president of the Maine Medical Association, said of the new technology. “If you’re at a private practice, you’re much less likely to have that advantage.”

Many rural providers — including those trying to accelerate their adoption of the technology — are finding it difficult to figure out which AI products are worthwhile.

“There’s a lot of garbage out there,” said James McHugh, a managing director at Berkeley Research Group who focuses on health technology and automation and helps clients implement the technology. Many AI implementations in health care haven’t gone particularly well so far, and adoption has taken substantial resources, he said.

Many system leaders look to invest in AI in the hopes that it will lower their labor costs, but that “usually doesn’t happen,” he said. There’s potential for it in the future, he added, but it’s not as simple as the grand claims being touted by AI vendors and administration officials.

The calculus for hospitals goes beyond just whether the tech that vendors are selling will actually improve their bottom line. It also includes whether the tools are safe for patient data and care.

“There are so many out there,” Hutches, who works at a relatively well-resourced system, said of the tools being sold, comparing it to the dot-com bubble. “You don’t know where to go.”

Small practices and clinics, especially those in the middle of a struggle to stay afloat, may have an even harder time landing on the right adoption strategy or finding the cash to invest.

“It still creates inequities for sure, as does the education and ability to use the tools themselves,” Dwyer said.

Even so, some AI vendors have reported expansion in rural areas. Aidoc, a radiology AI company, has seen “definitive growth” in more remote areas, Jesse Ehrenfeld, chief medical officer at the company and past president of the American Medical Association, told STAT.

“As you’d expect, a lot of those places are not well resourced: They don’t have AI governance in place, they don’t have the bandwidth to do deployment at scale,” he said. “It looks very different.”

Ehrenfeld said he sees a future where Aidoc’s technology can make sense of patient imaging to help under-resourced systems figure out how to deploy clinicians and medical transportation. But for now, rural leaders say fundamental issues of financial sustainability will have to be solved before an AI-powered medical system can come to many Americans.

“If you’re technology challenged,” Jarvis said, “it doesn’t help that someone comes up with a technology-based solution.”

STAT’s coverage of the health AI digital divide is supported by a grant from the Commonwealth Fund. Our financial supporters are not involved in any decisions about our journalism.