North Country Healthcare is a collection of 13 small, rural healthcare centers in Northern Arizona trying to implement artificial intelligence. It’s an uphill battle.
The system’s struggle cracks open the issue of how to equitably deploy AI in healthcare when some clinics are at a distinct economic and geographic disadvantage—and if the speed of innovation is leaving rural America behind.
North Country's clinics range across Arizona's I-40 corridor, a highway that runs from North Carolina to Los Angeles. Its clinics stretch from the eastern edge of the state, through Flagstaff and the Grand Canyon, to the western edge. Many clinics border the Navajo Nation and other pockets of Native-owned land in the state.
This rural and desertous part of the country is defined as medically underserved. Forty percent of North Country’s patients are Medicaid recipients, and other patients don’t have the ability to pay for services at all.
The system provides a sprawling range of services from primary care to obstetric, to HIV prevention, to mental health and residency programs. It is not affiliated with an academic medical center, which puts it on the lower-resourced end of the spectrum among rural health clinics.
Like other healthcare delivery organizations across the country, it is hoping to take part in the promise of AI to help relieve burdens on its staff and to provide better care for its patients.
Many clinicians at the practice were tapping into the technology using free versions of AI, including ChatGPT and Freed AI, while following online tutorials. One physician at the practice, Jennifer Cortes, M.D., set out to find a solution for the system.
“It's pretty hard to get people to come work in some of these areas, and then once they get out, they really want to make it easy for them to stay—so that was a main motivation,” Cortes explained of her decision to pursue an organized approach for the system. “From a safety and liability perspective, people are going to be using this on their own, we have got to figure out how to get some guardrails around it and make sure there's a little more organizational awareness and just pieces in place to make it work well and make sure it's safe for our patients.”
Cortes connected with Duke University’s Health AI Partnership (HAIP), which seeks to provide technical assistance to healthcare facilities without the resources to implement AI. By attending free webinars hosted by HAIP, Cortes met Mark Sendak, co-lead at Health AI Partnership and population health and data science lead at the Duke Institute for Health Innovation, who encouraged her to apply to its newly-created practice network.
HAIP chose five candidates from a competitive application process to participate in the practice network, which offers a 12-month program to assist sites with implementing its eight key decision points for AI procurement and governance. HAIP offers a wealth of resources, not only the expertise of Duke technologists, but also monthly office hours with Mayo Clinic and DLA Piper and mentor check-ins with Kaiser Permanente.
“[Our] mission is to empower healthcare professionals to use AI safely, effectively and equitably, and we do that through community-informed, contemporary … best practices,” Sendak said in an interview with Fierce Healthcare.
Together, North Country Healthcare and HAIP have been working to pilot an AI scribe at one of North Country’s facilities. Cortes said the system plans to start in May, after a slight delay. It will soon be completing pre-pilot surveys of patients and clinicians.
Getting to the start line of the pilot has been rocky. Even with the technical expertise of HAIP, the system is struggling with underlying issues that afflict many rural health clinics: weak internet connectivity, outdated technology and limited funding to improve the constantly arising issues.
“I wish we could be more optimistic about everything … I'm excited when I hear all of the innovation, and this new idea here and this new thing there, but I'm honestly starting to get a little bit like, ugh!” Cortes groaned in exasperation, “You know that they're trying to make these advances, and we can't even adopt any of this stuff. What is the point?”
North Country Healthcare’s struggle with AI implementation is similar to others in the HAIP practice network. But right now, HAIP is only helping five clinics, and the country has thousands.
“The work we're doing in the practice network now is what actually keeps me up at night,” Sendak said. “I mean, we're working with five organizations. There's 1,600 community health centers. There's over 6,000 hospitals, and the vast majority, I would say 90-95% plus of those organizations, do not have expertise, do not have infrastructure, do not have access to the best-in-class solutions.”
The issues with AI implementation in rural clinics
In the remote parts of Northern Arizona where North Country has its clinics, broadband infrastructure is spotty. Access to high-speed Internet powers the clinics’ access to digitally stored patient data, the electronic health record (EHR) system and telehealth visits.
While all the clinics theoretically have Internet access and WiFi networks, Cortes said her clinics still have issues connecting to the internet almost daily. “When you think about that, who cares about AI? If you can't keep this thing connected, it's not going to make people happy,” she said.
The system has recently completed WiFi equipment upgrades, and it is working with an external group to improve its overall connectivity.
A few months ago, the system learned that the version of the EHR system that it uses will no longer be supported in North America. Now, the system is in the middle of contracting with a new EHR provider and must migrate to a new system, expected to take a year, Cortes said.
“In a lot of ways, [the new EHR] is going to be helpful for us,” she said. The newer EHR might make more AI applications available to the clinics. “We're all really excited about that possibility, but ... it's just another funding hurdle.”
Sendak has seen this problem repeated across other sites HAIP is working with to implement AI. Even if they don’t have to replace their EHR, Sendak said some EHRs restrict AI integrations. “It's a form of vendor lock-in that is limiting people's access to the best products,” he said.
He continued: “These small organizations don't have the purchasing power, the leverage of a Duke or a Mayo. They don't have data scientists on staff. They don't have compute environments to store data to conduct analysis.”
Since North Country Healthcare began working with HAIP, its chief information officer and compliance director have left the company.
“That’s part of why we're super grateful for the practice network, because they'll help and review some of our things,” Cortes said. “We do contract with a lawyer, ... but, oh man, it would be nice to at least have a compliance person.”
As the system is working through HAIP’s AI implementation framework, it has also started working on implementing a data governance framework. Doing both at once is challenging, especially given the system’s other limitations in staff and funding.
“It takes so long for a place like ours just to be able to focus and move this along in a timely manner," Cortes said. "And so as we started to focus on our AI governance, we just realized our general data governance was not well set up. So now we're trying to improve our data governance structure as a kind of a base, and then are planning to add the AI governance pieces to it along the way.”
Overshadowing the conversation was the issue of finding money to make technological improvements to the system. It was already hard to convince North Country’s leadership to participate in the HAIP program because of the cost of the AI scribe. But the system has hit some financial trouble since beginning the partnership, and the costs to improve their technological infrastructure are stacking up.
The dawn of the second Trump administration has also caused financial worry. The system receives grants from the federal government, which it is uncertain will continue, and the system fears that Congress will cut Medicaid funding, which accounts for almost half of the system’s revenue.
The uncertainty of the Trump administration is also causing organizations to be more cautious about spending, she said.
“This is after getting through the pandemic, having the whole Optum debacle—which resulted in a lot of people not getting paid and having to take cash advances, and those types of things are stressing organizations—then you have a funding freeze,” Cortes said, clarifying that she was speaking for herself and not for her organization. “I have been pretty frustrated, I would say.”
Sendak hopes that states could step in to help fund technical assistance programs for healthcare facilities implementing AI. HAIP has a grant to plan what a technical assistance program in California could look like.
The group is working out how to build out a national hub-and-spoke model for AI technical assistance and wants to partner with states to do so. The model would leverage local institutions with AI resources to assist those without.
The resource hubs could be funded by states or by the federal government similar to the country’s 16 telehealth resource centers, which have helped local facilities implement telemedicine and comply with state and federal regulations.
The vision will be laid out in a new paper co-authored by Sendak and nearly 50 others that is undergoing peer-review.
“There's been coordinated efforts at the state and federal level,” Sendak said, speaking of past initiatives. “We did it with EHRs, with regional extension centers. We did it with telemedicine, with telehealth resource centers. And I'm confident that we can do it for AI.”