People in Utah can now renew prescription medication online using an artificial intelligence system without any input or review from an actual doctor.
Using Doctronic — a generative A.I. model — Utahns only need to answer a few questions and pay $4 in order to obtain a new prescription.
Doctronic, according to co-founder Dr. Adam Oskowitz, uses a shared database connected to about 95% of pharmacies to verify that the person has previously received a prescription for the requested drug.
If the A.I. is unsure that the prescription should be renewed, it refers a patient to a human doctor, Oskowitz said. Doctronic has doctors licensed in Utah on call at all times, added Matt Pavelle, Doctronic’s other co-founder.
The first 250 prescriptions renewed by the A.I. will be reviewed by a doctor, according to Doctronic’s agreement with the state. The contract also states that Doctronic also can’t renew prescriptions outside of a list of 191 drugs, none of which are controlled substances.
But while Oskowitz and Pavelle said they hope the system increases accessibility to medicine while lowering the cost of prescription renewals, others familiar with Utah’s medical field are concerned of what could happen when a human doctor isn’t involved in prescription renewals.
Michelle McOmber, CEO of the Utah Medical Association, said she has several concerns about the new program and what it claims to offer patients. She said she wondered: How can A.I. make up for medical histories taken by human doctors? And how would it know if a prescription was meant to be a temporary trial?
“I think A.I. could be used in a way that would be helpful if you have a connection just between A.I. and the pharmacy,” McOmber said. “A.I. making the decision on the renewal is concerning.”
The pilot program started, according to Oskowitz and Pavelle, as the next step in progressing Doctronic’s online generative A.I. model, which was built to provide medical advice at no cost.
Oskowitz and Pavelle cite a study they conducted with other researchers, which they say found Doctronic matched the top diagnosis of an actual clinician in 81% of 500 urgent-care telehealth visits. The study found that treatment plans made by A.I. and a human clinician aligned 99.2% of the time. The study, published in July by medRxiv, has not been peer-reviewed, though Oskowitz said it is currently in that process.
Wanting to take Doctronic further in the medical industry, Oskowitz said he and Pavelle looked toward Utah — a state that offers “regulatory relief” to companies looking to pilot A.I. programs.
They approached the state with their idea roughly five months ago, according to Zach Boyd, the director of Utah’s Office of Artificial Intelligence Policy.
“They first approached with this idea that maybe they could go farther with their model than others had gone before,” Boyd said. “They had these statistics on how good their product had been. … They were looking at what is kind of the natural next advanced thing for it to try.”
Working with independent doctors and pharmacists, Boyd said the state was able to agree on a contract to allow Doctronic to renew Utahns’ prescriptions.
The pilot program is meant to test the A.I.’s safety while hoping it will reduce medication costs and increase medication accessibility, Boyd said.
“The state doesn’t take a position whether people should like this business,” Boyd added. “That’s not the state’s role.”
The contract allows a 12-month agreement that began in October, although Doctronic didn’t start renewing prescriptions until December, according to Melanie Hall, the Utah Department of Commerce’s communications director.
Even though the A.I. is designed to send patients to an actual doctor if it determines a potential issue with the prescription, McOmber doesn’t think the safeguards are enough.
The A.I. is meant to ask patients questions — such as what other drugs they are taking, if they are taking their medication, and if their medication is working, Oskowitz said.
This could be a problem, McOmber said, for any patients who do not know all of the drugs they take. The use of A.I. to substitute for a doctor, she said, could miss the nuance of different patients’ diagnosis and their individual needs.
“To actually say that A.I. is making a decision that should be made by a provider, a physician — that is very concerning to us,” she said.