The state of Utah has embarked on the first test of artificial intelligence as an autonomous clinical decision-maker under a regulatory suspension paradigm with AI company Doctronic.
Rather than wait days or weeks to get an appointment with their doctor to get a routine medication refilled, Utahns can chat with an AI agent online to renew their prescriptions nearly instantly.
Doctronic, and the state, are betting that the issue is a significant enough pain point for Utah patients, pharmacists and doctors that they will be willing to engage with AI to fix the issue. The company says the refill automation will improve medication adherence and prevent downstream medical issues.
Some of the medications the AI may refill for Utah patients are statins for high cholesterol, medications for high blood pressure, psychiatric medications and birth control.
Utah legislators created an AI regulatory sandbox in the state in 2024. The Utah AI office can waive certain laws in order to test a novel program, often with private companies. The intent is to gather data to prove out the concept and eventually present the data to legislators.
In this case, the state is making exceptions to its laws on professional licensure, scope of practice, professional conduct and telehealth prescribing, among others.
“For most practitioners of medicine, the lines of who does what in medicine are kind of sacred and legally defined, and they have to be that way,” Zach Boyd, Ph.D., director of the Office of Artificial Intelligence Policy for the state of Utah, said. “We're in the Department of Commerce in the state, which is the entity that draws and redraws those lines all the time. So this is actually our core expertise, is to figure out those lines to redraw and how we tell when they move.”
In order to qualify for the regulatory relief program, Doctronic entered into an agreement with the state and designed a program with safety checks, ongoing reporting to the state and data privacy requirements.
In an interview with Doctronic executives, they explained that the pilot program they chose minimizes risk and creates a context where AI decision-making without a clinician is safe and automates routine clerical work for clinicians.
“I'm sure you've gone to a pharmacy at one point and asked for your renewal, and there was none available,” Adam Oskowitz, M.D., co-founder of Doctronic, said. “It's a great point for us to just kind of interject and help everyone out.”
The company has limited the types of medications it will refill to 191 commonly prescribed drugs. It excluded risky medications like narcotics, stimulants, injectables and antibiotics used for short-term illness. The formulary was reviewed by independent pharmacists and the state.
Moreover, because the AI only does refills, the prescription has been previously prescribed by a licensed practitioner, and the patient has taken the medication before.
The rollout includes multiple phases that incrementally give the AI application more autonomy.
The first 250 refills will be overseen by a human clinician. Then, Doctronic will sample 10% of cases to check the AI’s decision. The third phase will include random sampling of cases.
Monthly, Doctronic will report the number of users of the AI technology, impact analyses, technology assessments and information about user complaints.
It will also report the number of accepted and denied refills, the number of requests reviewed by a licensed clinician and any trends or observations made by clinicians.
“We hope to be sharing this broadly as we go,” Matt Pavelle, co-founder and co-CEO of Doctronic, said. “We want everyone to understand what's going well, what's not going well.”
Doctronic has talked to regulators in Arizona, and it has an upcoming conversation with the state of Texas, which passed an AI sandbox law last year that went into effect Jan. 1. It also has talked to regulators at the Department of Health and Human Services about the program.
“The goal is, after a certain period of time of deployment, we're going to have a robust evidentiary case for what the best practices are for this kind of thing, which, to be clear, right now would not be legally possible without the sandbox,” Boyd said. “This is traditionally a protected domain, and so we do need this kind of regulatory reset to press forward here.”