Federal health agencies are reshaping how artificial intelligence enters American healthcare, with the Centers for Medicare & Medicaid Services testing an unprecedented payment approach and the Food and Drug Administration working to develop frameworks better suited to evaluating generative AI medical devices.

At a Consumer Technology Association healthcare event, officials outlined ambitious timelines and reforms that could accelerate AI adoption across chronic disease management while raising questions about oversight adequacy and clinical safety standards.

CMS Tests Paradigm Shift in Technology Payment

The centerpiece of the federal AI push is CMS's ACCESS model (Advancing Chronic Care with Effective, Scalable Solutions), a 10-year outcomes-based program launched in July 2025 with more than 150 participating healthcare organizations. The initiative targets six conditions: diabetes, hypertension, chronic kidney disease, obesity, depression, and anxiety.

Stephanie Carlton, CMS Deputy Administrator and the newly created Chief Clinical AI Officer under the Trump administration, outlined a four-pillar strategy addressing public trust, data sharing and interoperability, market access pathways, and reimbursement frameworks. The ACCESS model is testing what Carlton described as a "paradigm shift: paying tech companies versus just paying clinicians and healthcare facilities."

CMS has set a December 2028 goal to demonstrate that "technology can have a massively deflationary impact on healthcare costs" through the ACCESS model. Carlton's presentation suggested the agency views AI-enabled tools as essential to achieving value-based care objectives at scale.

The model represents an early test of whether AI tools can improve measurable health outcomes while managing total cost of care, potentially providing an evidence base for broader adoption.

FDA Seeks Framework Better Suited to Generative AI

While CMS explores new market access pathways, FDA officials acknowledged their current review framework doesn't fit generative AI capabilities well. Dr. Rick Abramson, FDA Associate Director for Digital Health, characterized the existing framework as a "square peg and round hole" problem when applied to generative AI medical devices.

"I'm absolutely confident that the evidentiary standard for FDA authorization of generative AI tools will change because it's [a] square peg and round hole," Abramson said, seeking industry input on how to improve the agency's approach.

The agency issued a discussion paper in August 2024 addressing generative AI-enabled medical devices, with a public comment period that closed in October 2024. FDA is proposing a two-factor evaluation framework based on device autonomy level—ranging from information provision to autonomous action—and potential harm from incorrect outputs.

The agency is considering a "competency-based" evaluation approach that would combine benchmarking with real-world clinical validation, potentially creating pathways for AI tools that demonstrate consistent performance across diverse clinical scenarios.

Clinicians Debate Appropriate AI Autonomy Levels

Despite federal enthusiasm, the CTA event revealed persistent clinical concerns about autonomous AI decision-making versus human-in-the-loop approaches. Clinicians debated appropriate levels of AI autonomy in clinical workflows, liability considerations, and the adequacy of validation standards.

Dr. Jesse Ehrenfeld, past president of the American Medical Association and now global chief medical officer at Aidoc, cited AI-enabled prescription refills being tested in Utah as an example where human physician judgment remains necessary. "I've had patients walk in the office, and you know something's not right. The clinical context has changed. There's something that says that a refill is not safe today," Ehrenfeld noted, emphasizing the need to balance scaling capacity with appropriate clinical escalation.

The tension between CMS's aggressive deployment timeline and FDA's acknowledgment that its framework needs to change highlights the challenge facing healthcare organizations: adopting AI tools to participate in value-based models while navigating evolving safety and efficacy standards that regulators are still developing.

For health technology companies, the ACCESS model creates opportunities to participate directly as Medicare partners in chronic disease management, but success depends on meeting outcomes targets by 2028 while satisfying emerging regulatory requirements that remain in flux.

Sources

  1. At CTA event, federal officials outline AI ambitions as clinicians debate risks — Fierce Healthcare