Rural Arkansas hospitals facing closure and staff shortages could find cost-saving solutions in artificial intelligence, but the technology is often the least accessible in small communities.
Not-metro-adjacent hospitals are less likely to adopt AI, widening a digital divide that leads to health care disparities, according to a study from the Federal Reserve Bank of St. Louis.
Nicole Summers-Gabr, a senior researcher at the St. Louis Fed who conducted the study, said in an interview that this pattern appears nationwide. She chose to specifically look at not-metro-adjacent hospitals because research doesn’t typically address metro-adjacency — only metro or not-metro.
“If it’s a small community next to a metro area, that means they’re probably going to have access to the resources that are there, but if it’s a community that is not by a metro area, that means that they’re going to be a lot more isolated,” Summers-Gabr said.
That isolation matters as Arkansas faces significant rural health care challenges. A 2025 report from the Center for Healthcare Quality & Payment Reform found that 30 out of 47 rural hospitals in Arkansas are at risk of closing, with 11 at immediate risk.
AI applications could prove useful at addressing those challenges. New AI tools can diagnose conditions like lung cancer and multiple sclerosis without a specialist present, Summers-Gabr said.
“That doesn’t treat them,” she said. “But what that does is give them an answer right away on what their issues are, so they can get the treatment that they do need a lot sooner.”
Sophisticated AI applications are already being piloted in the state. Mercy announced last week that Mercy Hospital Fort Smith has been helping develop an ambient AI tool that documents nursing observations from conversations between a patient and caregiver, automatically feeding information into that patient’s electronic health records. Mercy Northwest Arkansas is also expected to adopt the tool soon.
On the administrative side, hospitals can use AI for revenue cycle management, including billing and claims processing, supply chain optimization, patient scheduling, and forecasting to predict staffing needs during busy periods like flu season.
But Summers-Gabr’s research found that, at least as of 2023, the latest year for which data is available, rural hospitals were adopting AI at dramatically lower rates than metro hospitals. Among Arkansas’ metro hospitals, 20% reported some type of AI usage. For metro-adjacent hospitals, that percentage drops to 8.3%, and not-metro-adjacent hospitals reported just 5.6% usage. Summers-Gabr said more detailed data with updated 2024 adoption rates and clearer use-case definitions will be released next month.
That disparity raises more concerns than just a lack of usage. If training models do not receive enough information about rural patients, there won’t be sufficient data to optimize the needs of small communities, leading to data bias in diagnostic and other tools.
“Its ability to diagnose is only going to be as good as the data that’s read into it, and that data needs to be very representative of the population,” Summers-Gabr said. “So if all of the data comes from [patients] 65 and up, and you’re trying to diagnose a condition in a 12-year-old, that’s going to be tough.”
She said AI still needs human verification, but it can be leveraged by understaffed hospitals. Facilities also have to understand the requirements of these tools, determine staffing needs, address regulatory compliance, and assess patient acceptance of AI-assisted care.
“Where there is need, innovation always comes, and there is a need in health care right now,” Summers-Gabr said. “But it’s not going to be great unless it’s done in a way that is compassionate and is considerate of all different patients that are seen by doctors.”
Sign up for Daily E-news
