CMS Maps AI Payment Models as FDA Rethinks Device Approval

Federal agencies accelerate clinical AI strategy

The Centers for Medicare & Medicaid Services is developing new payment frameworks for AI-enabled healthcare technologies while coordinating with the FDA on regulatory approaches, according to officials speaking at a Consumer Technology Association event in Washington this week.

Stephanie Carlton, CMS deputy administrator and the agency's first chief clinical AI officer, outlined a four-pillar strategy covering public trust, data interoperability, market access pathways, and reimbursement models. The approach signals how the Trump administration plans to integrate AI into clinical care at scale.

Carlton indicated CMS will soon expand its ACCESS model—Advancing Chronic Care with Effective, Scalable Solutions—a 10-year value-based program launched in July with more than 150 participating healthcare organizations. The model provides outcomes-based payments for technology treating diabetes, hypertension, chronic kidney disease, obesity, depression, and anxiety.

"We have said that we want to announce more tracks and more opportunities to engage in that, so the team has been working very carefully on those additional tracks," Carlton said, though she provided no timeline.

Reimbursement questions take center stage

CMS faces fundamental questions about when and how to pay for AI tools as they move beyond wellness apps into medical functions. Carlton framed the challenge: distinguishing between capabilities available cheaply through consumer AI models versus technologies performing medical functions that meet CMS standards for reasonable and necessary care.

The ACCESS model represents an early test case. Carlton projected that by December 2028, the program could demonstrate technology's "massively deflationary impact on healthcare costs." She described paying technology companies directly—rather than only clinicians and facilities—as a paradigm shift the model is designed to evaluate.

CMS is also determining which health outcomes to monitor and what reporting requirements make sense without creating excessive administrative burden. "We don't want to build a whole quality industrial complex around reporting," Carlton said.

FDA acknowledges regulatory mismatch

The FDA's current review framework doesn't suit generative AI, according to Rick Abramson, associate director for digital health at the agency's Center for Devices and Radiological Health. "I'm absolutely confident that the evidentiary standard for FDA authorization of generative AI tools will change because it's square peg and round hole," he said.

The FDA issued a discussion paper in August exploring regulatory considerations for GenAI-enabled medical devices, with public comment open until October 19. The paper proposes evaluating devices based on their autonomy level and potential harm from incorrect outputs, using what the agency calls a "competency-based" framework combining benchmarking with real-world clinical validation.

Why it matters

The federal government is moving from exploratory AI pilots to concrete payment and regulatory frameworks that will determine which technologies reach patients and how providers get reimbursed. The ACCESS model's structure—paying for outcomes rather than services, and compensating technology companies directly—could reshape healthcare economics if it demonstrates cost reduction without compromising quality. Meanwhile, the FDA's acknowledgment that existing device approval processes don't fit generative AI suggests significant regulatory changes ahead, creating both uncertainty and opportunity for developers.

Physicians debate autonomy versus oversight

Healthcare leaders at the event clashed over how much human involvement AI systems require. Jesse Ehrenfeld, past president of the American Medical Association and now global chief medical officer at Aidoc, argued that even routine tasks like prescription refills need physician judgment to catch context changes that automated systems might miss.

Marc Paradis of SIYOM Consulting countered that AI will soon detect clinical signals beyond human perception, from voice changes to hyperspectral imaging patterns during surgery. Laura Adams, senior advisor at the National Academy of Medicine, challenged the assumption that clinicians must always review AI decisions, suggesting that requirement could create unnecessary delays where AI already performs well.

John Whyte, CEO of the American Medical Association, emphasized that AI should meet the same evidence standards as other healthcare interventions before expanding its role.

These details were first reported by Fierce Healthcare.