It’s been seven months since artificial intelligence company Doctronic began its partnership with the state of Utah on automated prescription refills.
The program allows 30-, 60-, or 90-day renewals for medications already prescribed by licensed providers, according to the mitigation agreement (PDF) between the Office of AI Policy and Doctronic. The company limited the types of medications to 191 commonly prescribed drugs, including psychiatric medications and birth control.
Matt Pavelle, co-founder and co-CEO of Doctronic, told Fierce Healthcare that medication non-adherence causes "125,000 preventable deaths" annually in the U.S., and costs the healthcare system "$100 billion."
“Statistically, about 30% of this problem is attributed to friction in the system and getting your prescription,” Pavelle said. “It could be you don't have any refills left, it could be that you can't afford the medicine, lots of reasons…We said we can at least solve one pain point in that system with AI.”
So, Doctronic approached the state of Utah with its proposal and—after working closely with the Utah Office of AI Policy—the pilot was born under the state’s regulatory relief program.
Adam Oskowitz, M.D., co-CEO and co-founder of Doctronic, told Fierce Healthcare Utah was selected as the state partner for several reasons, including being home to the “most mature AI sandbox.”
“They have shown an aptitude for forward thinking and technology-friendly applications,” Oskowitz said. “…They were also really data-driven, which we really liked, and the fact that they were data-driven aligned with our own goals, our own vision for how we would roll this out, and they were collaborative as well.”
Utah Office of AI Policy Director Zach Boyd, Ph.D, told Fierce Healthcare the program was created two years ago to aid the state in responding in “an agile manner as AI disrupts a lot of different industries.”
Boyd said the law permits the office to accept applications from companies in any sector, including healthcare, that want to deploy AI products that do not have a “good regulatory pathway” under current Utah law.
“That may be because of ambiguity, or it could be because it is actually against a current law or regulation,” Boyd said. “But they would contend that the nature of the technology means that, probably, the legislative intent would not have been to block that kind of deployment.”
Through the program, approved companies are granted a “temporary custom regulated environment” for one to three years, according to Boyd.
Early results
Currently in its first phase, results of the first five months of the program were published (PDF) in May. Data is based on Doctronic physicians reviewing interactions, though the state’s Office of AI Policy has initiated an independent review.
“Right now it's still humans in the loop, doctors in the loop, but the AI is still doing all the work,” Pavelle said.
The AI system recommended prescriptions for renewal in 72% of cases, and reviewing physicians agreed with those recommendations 91% of the time. In instances where renewals were not approved, reviewing physicians requested additional information—which was reviewed by a second physician. In 97% of cases, either the first or second reviewing physician found the renewal authorization appropriate based on the information the AI had gathered. This reflects expected inter-reviewer disagreement.
The datasheet notes “in most cases” the second physician disagreed with the initial request for more information.
When renewal was not deemed appropriate by the AI—in 28% of cases—69% of physicians agreed the denial was appropriate in clinical escalations. Physicians in the remaining 31% of cases said the escalation was “overly cautious.”
Oskowitz said the company can “tune the system” to approve more medications following the results of overly cautious decisions.
“It was a really great way for us to really understand what the capabilities of the system [are], how well it can discern between sort of subtle differences in diseases, and how we can update the system to be more reliable and provide more access to healthcare,” Oskowitz said.
As the program continues, Pavelle said repeat usage of the system is “off the charts.”
“In fact, lots of people are coming back and almost immediately doing second prescriptions, sometimes third prescriptions,” Pavelle said. “Lots of Americans, especially older Americans, have multiple chronic conditions and need to be on multiple medications. The system is really good at handling that because it's checking all the drug-drug interactions on every renewal.”
In the second phase, the company will sample 10% of cases to check the AI’s decision. The third phase will feature random sampling of cases.
Initial provider pushback and the roadmap ahead
The program was met with skepticism and vocal criticism from local health officials.
In April, the Utah Medical Licensing Board called for the program’s immediate suspension in a letter (PDF) to the Utah Department of Commerce. The Board said it was made aware of the partnership after its implementation—once the system was already live.
Utah officials responded to the Board (PDF), noting companies involved in the program must submit a "detailed deployment and safety plan, which is thoroughly vetted."
Boyd said the office has since established a joint working group with the Board.
“We don't want a sense that there's any kind of infighting,” Boyd said. “…[It] will give us the ability to do more and better work.”
Pavelle said the company was “very conservative” with the initial formulary, and there are plans to add new medications once it is proven safe to do so.
“For example, we took all injectables off,” Pavelle said. “This was a conscious decision for safety's sake, but there are diabetic medications and GLP-1s that people can likely benefit from, and so we would certainly hope that after the system is proven safe and effective, we would increase that formulary.”
Doctronic has talked to regulators in Arizona, Pavelle said, and noted other states—like Texas and Wyoming—have similar regulatory relief programs to Utah.
Moreover, the U.S. Department of Health and Human Services’ (HHS) Advanced Research Projects Agency for Health (ARPA-H) launched a program—dubbed Advocate—to build agentic AI for cardiovascular disease management, Pavelle said.
“There's a lot of movement from a lot of different governments on trying to see what can happen with AI and how we can solve problems and healthcare with it,” Pavelle said.