New CPT codes are expanding reimbursement for AI-powered CCTA tools

"What Category I CPT codes give you is sort of a national standard of care and established service," Santillo explained to Cardiovascular Business. "What Category 1 codes give you is physician valuation. So the physicians now have RVUs attached to those codes."

Clinical evidence drives reimbursement

The U.S. Food and Drug Administration (FDA) has cleared more than 1,614 clinical AI algorithms, with about 80% of these in medical imaging. All of these are assigned a Category III tracking code to attach with billing so the use and proliferation of the algorithms can be tracked. Of these, Santillo said only three have earned Category I codes, with one being in ophthalmology.

She said attaining a Category 1 code for payments requires more than demonstrating that an algorithm can analyze medical images. The American Medical Association's (AMA) CPT panel must consider whether a technology represents a new medical service, whether substantial clinical evidence supports its value and whether clinicians are widely using the service.

FFR-CT had a Category III code for several years before moving to Category I in 2024. AI plaque analysis made the transition after only four years, which Santillo said was very fast. She pointed to the relatively rapid adoption of plaque analysis and strong clinical evidence as the primary reasons it advanced so quickly through that process.

Both technologies provide information intended to help clinicians make treatment decisions for defined patient populations, including patients with intermediate-risk chest pain. A large amount of clinical study data demonstrated that these AI results resulted in changes in patient care.

Simply using AI to flag a potential abnormality may not justify separate payment if identifying that finding is already part of a physician's existing responsibilities. The stronger case for reimbursement comes when an algorithm provides additional information such as plaque quantification or an assessment of associated patient risk.

Defining AI services and physician work

The AMA has also introduced terminology to distinguish different types of AI services. The framework describes augmentative, autonomous and assistive AI and their specific roles. Augmentative AI provides physicians with a new dataset or report that requires independent interpretation. Autonomous AI operates without a physician in the loop, and assistive AI identifies or flags information for consideration.

Santillo said augmentative services in the CCTA pathway involve additional physician work, helping define them as distinct services for reimbursement purposes. She added that the AI is performing an analysis that could not be done in the real world by physicians because the time involved would be too labor intensive.

Accurately valuing physician work is another important step used to establish work RVUs (wRVUs). Santillo urged clinicians to complete professional society surveys to estimate the time, effort and intensity required to perform services. These AMA/Specialty Society RVS Update Committee (RUC) assessments are used to inform RVU valuations.

If physicians fail to respond or provide inaccurate information, Santillo said, reimbursement may not adequately reflect the work involved.

"It's a big deal. It really is. I wish these emails went out with blinking lights so that the physicians would say, 'Yes, this is really important and key for me to fill out and fill," she added.

Although AI adoption is expanding, Santillo said the reimbursement processes remain deliberately gradual. Cardiology has gained an early advantage because FFR-CT and plaque analysis have demonstrated measurable clinical value, while some AI applications in radiology primarily flag findings physicians are already expected to identify. She said the studies needed to show clinical value and how it impacts care decisions is often lacking when it comes to radiology algorithms.