Hacker-City
Hacker-City
Get the brief
Health|May 26, 2026|5 min read

What do early data from Utah's Doctronic AI pilot show?

Early data from Utah's five-month pilot program with AI startup Doctronic reveals the system renewed prescriptions 72% of the time without physician escalation, though independent review of the pilot's safety claims is still pending. The experiment tests whether AI can legally assist with prescription renewals while operating under the state's AI innovation regulatory sandbox.

#ai-prescriptions#doctronic#utah-ai-pilot#healthcare-ai#fda-regulation#prescription-renewal#medical-ai#health-tech#ai-safety#regulatory-sandbox
S

STAT

Contributor

What do early data from Utah's Doctronic AI pilot show?

You're reading the web edition of STAT's Health Tech newsletter, our guide to how technology is transforming the life sciences.

Good morning health tech readers!

What have we learned from Doctronic's AI experiment in Utah?

Recent data released by the state of Utah provides insights into its five-month pilot program with AI startup Doctronic. Earlier this year, the state granted authorization for the company to utilize an AI chatbot to manage prescription renewals for approximately 200 medications. This pilot operates under a regulatory framework that allows the state's Office of AI Policy to waive certain regulations for AI entities that agree to abide by oversight and safeguards. The intention behind this initiative is to enhance the understanding of effective regulations for AI applications in healthcare. However, the Utah Medical Licensing Board has expressed significant opposition to the pilot, and experts have previously cautioned that there are inherent risks, particularly regarding the legality of an AI system renewing prescriptions without Food and Drug Administration (FDA) approval.

Currently, the pilot is in its initial phase, during which all AI decision-making is subjected to clinician review. While the exact number of sessions conducted by Doctronic in Utah has not been disclosed, state officials have indicated that it is limited. So what insights have emerged from this initiative?

The AI successfully renews prescriptions 72% of the time without requiring escalation to a physician for issues such as missing lab work. Additionally, in 69% of instances, reviewers determined that escalations were warranted.

Discrepancy in AI and human agreement. In cases where the AI facilitated renewals, physicians concurred with those decisions 91% of the time. However, there was a significant discrepancy in cases of disagreement, where secondary clinicians frequently contradicted the initial reviewer’s assessment. Notably, in 3% of instances, neither clinician aligned with the AI's judgment.

Addressing bias. A critical limitation of this data is that all oversight and review are being conducted by Doctronic's internal team. Nonetheless, state officials plan to undertake an independent review. Zach Boyd, the director of Utah's AI office, stated that the state has acquired an anonymized sample of conversations for expert evaluation "to both confirm Doctronic's reports and provide qualitative feedback on potential improvements to the pilot."

Doctronic Co-CEO Adam Oskowitz remarked that the preliminary data highlights the system's safety and efficacy: "No adverse events or contraindicated prescriptions were identified in the early results of Phase 1," he communicated. "Phase 1's pre-approval gate functioned as intended with no inappropriate prescriptions identified because nothing got through that shouldn't have."

While the statistics present compelling information, Girish Nadkarni, chief AI officer at the Mount Sinai Health System in New York, cautioned: "We should not let early operational numbers lead to the conclusion that 'AI prescribing is safe.' A more prudent interpretation would suggest that a limited AI system may assist in processing select refill requests, yet we must still gather independent evidence concerning its safety, equity, and effectiveness compared to a well-structured physician-supervised workflow."

BigHat Biosciences CEO on what creates an edge in AI drug design

BigHat Biosciences has established collaborations with leading pharmaceutical companies, including Merck, Amgen, Abbvie, and Lilly, and has recently completed projects with Johnson & Johnson. The organization is also advancing two of its own candidates into clinical trials.

In a recent conversation, Brittany Trang engaged CEO Peyton Greenside in San Mateo to discuss the significant challenges faced in AI drug design. "If you want me to design you a protein right now in six hours, I'm happy to do it," she stated.

AI developers aren't taking advantage of FDA policy that allows speedy updates

AI developers have frequently voiced concerns that the rapid evolution of technology outpaces the slow pace of FDA reviews. Due to the requirement to lock software for studies and submissions, products might become outdated by the time they reach the market.

Recent research indicates that developers have not fully capitalized on a regulatory policy permitting pre-approval from the FDA for modifications. An analysis of public documents from 794 AI-enabled medical devices authorized between 2023 and 2025 revealed that only 42 had plans filed with the FDA.

The study noted a modest increase in submissions beginning in 2025, which may suggest a delay in the adoption of this policy. Nevertheless, the authors concluded that "low overall adoption suggests there are still unidentified barriers to the use of PCCP or that certain AI-enabled devices may not be suitable for prespecified design changes that maintain safety and effectiveness."

Oura inches toward public filing

Last week, smart ring manufacturer Oura disclosed that it had confidentially submitted a draft registration for a public offering. This swift action follows the company's announcement of raising $900 million last October.

According to analyses from Exits & Outcomes reviewing filings in other jurisdictions, Oura's revenues reportedly reached $924 million for its fiscal year ending in September 2025, a significant rise from the $396 million reported the previous year.

E&O author, Brian Dolan, observed: "One of the thresholds for filing for an IPO as an 'Emerging Growth Company' is total annual gross revenues under $1.235 billion. While it's uncertain whether this threshold will expedite Oura’s IPO plans for 2026, it appears that the company may exceed this threshold should it file next year. This maneuver would allow Oura to disclose only its performance for the past two fiscal years, among other requirements."

Dolan also cautioned that "accounting practices vary across countries, so these figures may differ in the filings for a public offering in the United States."

Share this story