Imaging AI tools and algorithms continue to be rapidly developed and deployed into clinics, but experts say there’s an elephant in the room that still needs to be addressed: reimbursement.
Specifically, AI supporters are focusing on the lack of current procedural terminology (CPT) codes representing imaging services performed with AI. CPT codes are needed for appropriate insurance reimbursement.
“It’s the wild west out there as far as how it’s being handled,” said Sandy Coffta, vice president of client services at Healthcare Administrative Partners.
New research comes out every week in radiology highlighting the performance and potential AI in assisting diagnosis of various pathologies. This also reflects the regulatory side of things. The U.S. Food and Drug Administration (FDA) cleared the 1,000th clinical AI algorithm to close out 2024, and radiology is represented in about 75% of FDA-approved AI devices.
However, reimbursement for use of these tools may get more complicated as new algorithms and devices are employed by imaging departments.
Currently, AI reimbursement costs are handled via one of a few ways. One is passing these costs on to patients in the form of a lump sum that’s included in their imaging exams. Another is hospitals absorbing these costs to differentiate themselves from competitors.
Coffta added that, from a compliance perspective, mammography CPT codes for example include computer-aided detection (CAD), which may be interpreted as AI not being able to be reimbursed separately under current CPT codes.
For breast imaging, AI reimbursement can be a challenge. Previous research suggests that incorporating AI into mammograms can increase cancer detection rates by up to 20% while mitigating false-positive cases.
While the Affordable Care Act requires health plans to cover 100% of the cost of an annual breast cancer screening exam, other breast exams like AI-assisted imaging or breast MRIs are not covered, leaving patients with out-of-pocket costs. This may be especially difficult for women with dense breasts, who need supplemental imaging along with their screening mammogram.
“What we’re seeing is a relatively patchwork implementation and integration of AI into breast screening clinics,” said Kelsey Hampton, PhD, director of mission communications and education at Susan G. Komen. “A lot of people may not know whether or not their breast imaging clinic has an option for an AI-assisted read for their normal mammograms.”
Hampton said that research and policy updates regarding AI’s place in the clinic are needed to ensure equitable access for all women.
“This is certainly a new frontier of science and technology, and while it holds a lot of promise, there are still a lot of unknowns,” she added.
While there are currently no direct reimbursements for AI use in the clinic, there have been some applications for new technology add-on payments (NTAPs) -- which make AI technologies available for reimbursement via diagnosis-related groups (DRGs), for example.
“[NTAPs have] been used in AI in the setting of stroke for temporary reimbursements,” said Mahmud Mossa-Basha, MD, from the University of Washington in Seattle.
Mossa-Basha, who specializes in neuroradiology, said that eventual direct reimbursement will be based on value through outcomes of using AI in imaging.
“I think reimbursement and support for AI is taking more of a conservative approach. I think initially there was a lot of excitement in our field for AI and the things that it could do. And now, I think people are taking a pause and saying, ‘Let’s see what it actually does.’ ”
He added that research and science is catching up to AI development, which will eventually lead to payors catching up, along with support by the Centers for Medicare and Medicaid Services (CMS).
Coffta said that patients in the meantime can ask whether imaging departments offer AI enhancement as a service and how the costs are dispersed.
“A lot of what I’m hearing is anywhere from $40 to $100 typically across different practices,” she said. “And since there are no procedure codes, patients should be aware that even if a facility offers to bill their insurance, it’s unlikely they will pay.”
Hampton said that for breast imaging, patients should also consult their clinicians about the pros and cons of having AI assistance for breast cancer screening.
“One caveat about some of these AI models is…a lot of the benefits of AI are only as good as the models they’re trained on,” she said. “Not all models are trained on a wide representation of people who need this breast imaging. Talking about where the data comes from, how the models are trained, and if you as a person are someone who might benefit from that AI read is a conversation we recommend having with your doctor.”
The experts said that validating AI in the clinic is needed to prove that such devices are billable services, which could lead to temporary, then permanent CPT coding. From there, payors will develop their own policies on what health services using AI are covered. This aligns with CMS’ goal that every Medicare beneficiary and most Medicaid beneficiaries will be in a value-based care arrangement by 2030.
Mossa-Basha said that hospitals and AI developers need to prove “clear, unequivocal value” of AI algorithms and devices and work toward being net-zero or positive when it comes to the costs of using AI. This includes proving that AI can lead to better efficiency and faster interpretation in imaging.
“But really, it’s up to the AI manufacturers to prove that increased efficiency and for that to be actuated into [imaging] departments,” he said, adding that he believes getting government support for direct reimbursement will come sooner than later.
“It’s going to really push forward the utilization and value of AI in medical imaging,” he said.
Whether you are a professional looking for a new job or a representative of an organization who needs workforce solutions - we are here to help.