Editor’s Note: Colorado’s healthcare system is entering a period of accelerating instability — a “tsunami,” as multiple leaders describe it — driven by collapsing hospital finances, shrinking Medicaid reimbursements, disappearing services and a political fight over who is responsible for this unraveling. Healthcare Freefall is a multipart series examining how Colorado arrived at this moment, what H.R. 1 will do next and why communities across the state are bracing for deeper shocks.

Rural hospitals in Colorado say they are being pulled into a technological fight they didn’t choose.

Hospitals said insurance companies have deployed artificial intelligence to deny claims faster and more efficiently, and they are scrambling to adopt their own AI tools to defend their revenue — even as they struggle with shrinking margins, aging populations and service cuts.

Other observers counter that insurance companies must adopt AI tools — not so they could harm patience and maximize their profits, but because algorithms are very well-suited to the task of sifting through massive data, such as claims, as well as make other processes more efficient, ultimately helping all parties.

Nick Colleran, CEO of Delta Health, described the dynamic bluntly.

“All these insurance companies, they’re using AI to deny stuff from us right away,” he said. “They can look through our chart and say, ‘You didn’t have this code in’ or ‘You didn’t cover this.’ They’re scrubbing, finding ways to deny us.”

The insurance association in Colorado did not immediately respond to questions.

The issue of insurance companies using AI or tools backed by AI algorithms to deny claims has exploded in recent years.

In one case, an insurance company said the plaintiffs, in fact, failed to prove their claims were denied by an AI algorithm or show that the company improperly denied a covered service. The insurance company also said while the plaintiffs claimed its use of the algorithm was a “secret scheme to defraud,” they “cannot identify any actual misrepresentations.”

While many are worried about AI algorithms making health decisions, researchers from Stanford argued that, like other tools, the technology offers both problems and opportunities.

“In the sunniest scenario, it will help insurers approve requests more efficiently, improve communications with providers and patients, and conserve reviewers’ time for hard decisions in cases where care requests might not be evidence based,” they said in article in Health Affairs.

“In the darkest, it will supercharge flawed processes, making prior authorization cheaper to administer and thereby lowering barriers to expanding its use. Insurers have financial incentives to move in both directions; AI can facilitate their best and worst impulses,” they added.

Specifically, the researchers said that the massive amounts of submitted claims necessitate automation — and that older, non-AI algorithms “have made large-scale errors.”

AI, the authors of the article said, could tackle this problem, such as by fully automating the approvals of prior authorizations and claims, a task that AI is “well-suited.”

“Despite concern about denials, most requests are approved,” they said, noting that Medicare Advantage plans approved 93% of prior authorization requests from 2019–2023.

AI, they said, could be configured to approve “clearly allowable requests,” which would reduce stress and delays, “while allowing insurers’ medical reviewers to focus on more complex requests.”

AI could also help in reducing denials because of incomplete or poorly explained information, they said. Human preparers of these requests could lack clinical training or struggle to find relevant information — AI tools can automatically pull from databases.

Finally, the researchers said, AI can “reduce barriers to appealing denials, helping rectify wrongful denials and strengthen insurers’ incentives to avoid them.”

For Colorado’s Delta Health, which ended 2024 with a negative 9% margin, those denials are not just administrative headaches — they are existential threats, Colleran said.

“Just to pay our people — 70% of healthcare costs are in labor,” Colleran said. “It’s become a real bleak outlook for rural healthcare.”

The hospital has already cut its labor and delivery unit, eliminating a service that delivered roughly 120 babies a year.

“We decided to close that unit because we were losing a million dollars plus,” Colleran said. Women now travel 35 minutes to Montrose or more than an hour from rural parts of Delta County for care.

And the financial pressure is intensifying. Delta Health will lose its 340B eligibility — a $3 million hit — because its patient mix has shifted toward older patients, with fewer Medicaid patients.

“We really live closer to a critical access hospital,” Colleran said. “But we don’t get all the same benefits.”

Hospitals fight back

U.S. Rep. Brittany Pettersen, who represents the 7th Congressional District, said AI is entering healthcare faster than regulations can keep up.

“Insurance companies are building fast algorithmic systems that can deny claims quicker,” she said.

Colleran described the dynamic as a technological tug-of-war.

“Insurance companies are using AI to deny stuff from us right away,” he said.

To survive, Delta Health is adopting AI tools designed specifically to counter insurers’ algorithms. The hospital recently purchased software that flags likely denials before claims are even submitted.

“The software will pick it up and say, ‘Put this in so they can’t deny you next time,’” Colleran said. “We can make sure our doctors are documenting properly, our billers are catching those things, our coders are coding those places.”

The tool also tracks patterns in insurer behavior. “They’ll say, ‘For this insurer, they’re denying it every time because of X, Y and Z,’” he said. “So, we’ll be able to get ahead of that.”

Delta Health’s board chair said the pace of coding changes makes manual compliance nearly impossible.

“Human beings can’t keep track of constant updates to codes,” she said. “This AI tool will help sift through all that stuff real time.”

Are hospitals ready?

Chris Looby, a professor of health professions at Metropolitan State University of Denver, said the AI arms race is happening faster than hospitals can adapt — especially rural ones.

He said insurers’ use of AI is only one side of the story. Hospitals themselves are beginning to use machine-learning models to predict patient needs, emergency department visits and high-risk events. But the technology requires clean, standardized data — something hospitals rarely have.

“Artificial intelligence requires good, clean data,” Looby said. “Hospitals have lots of data, but richness means high quality. And EMRs were built around billing and accounts receivable, not around the kind of data you need.”

Looby said most hospitals cannot replicate that high level of modeling because their data is inconsistent.

“There’s about 5,000 hospitals and all 5,000 do something different,” he said.

McKinsey & Company, which offers software solutions and data and analytic tools, sees a lot of upside in AI use by insurance companies. In a blog, the company said that generative AI, in particular, is particularly salient to insurers because it is capable of “levels of reasoning, judgment, creativity, and empathy that far exceed previous innovations’ capabilities.”

Insofar as AI claims, in particular, the company noted that one UK insurer launched more than 80 AI models to “improve outcomes in its claims domain, cutting liability assessment time for complex cases by 23 days, improving the accuracy of routing claims to the appropriate teams by 30 percent, and reducing customer complaints by 65 percent.”

The company added that AI can add an element of “empathy,” noting an carrier that uses AI to generate some 50,000 claims-related communications every day, “finding them clearer and more empathetic than those written by humans.”

Looby, the Metropolitan State professor, said the biggest obstacle to AI-driven care isn’t technology. It’s human behavior.

“Artificial intelligence will tell me all these great things that I can and should do, but I’m not going to do them,” he said. “Americans don’t want their lives ruled by Big Brother.”

He described Colorado as “ruggedly individualistic,” citing James Michener’s book about Colorado, Centennial, and his own experience moving to the state.

“They’re not going to raise their hands for anything,” he said.

Even clinicians resist change.

“It takes 10 to 15 years to get clinical practice to change,” Looby said. “People just don’t believe the new stuff, the new approach, the new way.”

Still, Looby said AI is already writing clinical notes in some systems, ensuring documentation aligns with insurer expectations and reduces denials.

“AI will produce your note as a provider,” he said. “They’re working with the medical staff to let AI write these notes so that the language is there, so they can bill and not have denials.”

He said Medicare pays clean claims in 15 days — a lifeline for hospitals — while commercial insurers often take 30 to 45 days, and Medicaid can take far longer.

“If you can get paid in 15 days in the healthcare business, that’s fabulous,” he said.

Looby said the next major technology shift will come from quantum computing, which he expects to accelerate pattern detection and drug development.

“We can test a thousand things at the same time at the snap of a finger,” he said. “We’ll find patterns faster.”

For rural hospitals like Delta Health, the hope is that AI will help them survive long enough to benefit from those advances. For now, Colleran said, the goal is simpler: stay ahead of insurers’ algorithms and keep the doors open.

“We’re trying just to break even or make 100 grand,” he said. “That’s where we’re at.”