Rev Cycle Edge
Fee-for-service payment approach creates obstacles to AI adoption by physicians
A new analysis argues that fee-for-service reimbursement can penalize physicians for using clinical AI, raising questions about how payment models should evolve to support better outcomes and workflow efficiency.
Published
1 hour ago
AI is poised to transform how healthcare is provided and paid for, but the commonly used fee-for-service (FFS) model of reimbursement could be a problem in getting physicians to adopt clinical AI tools, a group of researchers has found.
The sticking point is that when a physician uses AI clinical technology or software in a fee-for-service model, most of the potential gains in efficiency come at the expense of lowered reimbursement for the attending physician. If a doctor normally bills based on time, and the average visit shrinks to 15 minutes from 30 minutes — because of clinical AI — the physician takes a big hit in pay.
Sharif Vakili, MD, MS, MBA, the lead author of a report published in the March issue of NEJM Catalyst, said that in researching possible uses of AI, the researchers concluded that the incentives are not set up properly to encourage use by physicians.
“If you prescribe this thing, it takes away your easy visits, and you’re sort of disintermediated from the economic equation, too,” said Vakili, a clinical assistant professor in the department of medicine at Stanford University School of Medicine.
Clinical AI is regulated as a software medical device and would normally be paid for by a payer using Current Procedural Terminology (CPT) codes.
“You get a CPT code, you get coverage for it, physician prescribes it, patient gets it, payer covers it,” Vakili said.
But when that software medical device is a clinical AI solution that’s delivering care that physicians would normally do, that device starts taking away the physician’s business, he said.
A way around the problem
One proposed solution, as hinted at by the title of the NEJM study “Artificial intelligence in the clinic: Don’t pay for the tool, pay for the care,” is to pay for the treatment, not the device.[1] To do that, a billing option could be added based on medical device management (MDM) tied to the complexity of care , so the physician’s payment isn’t affected if the visit is completed in half the time compared with before the assistance of AI.
The proposed reimbursement model has some advantages. Katie Gilfillan, director of value-based care and clinical integration for HFMA, said the approach proposed by Vakili, could work better than Medicare’s Advancing Chronic Care with Effective, Scalable Solutions (ACCESS) model because it keeps care management focused on the physician.
“While ACCESS introduces outcome-aligned payments, it risks further fragmenting care by shifting responsibility for chronic condition management to third-party, tech-enabled providers rather than the patient’s PCP,” Gilfillan said.
Although PCPs in ACCESS can bill for care co-management, the financial and operational center of gravity still sits with the technology provider, she said.
Plus, by allowing primary care providers to bill fee-for-service for technology-supported services, Vakili’s approach creates a more direct incentive for physicians to adopt tools that enhance care delivery within the existing patient relationship, Gilfillan said.
Primary care physician, quality key
That alignment matters—because when the PCP remains the hub of care, there is greater continuity, accountability, and integration across conditions, she said.
Importantly, the model ties the use of AI tools to improved outcomes, not just adoption. That distinction helps ensure technology is used to augment clinical decision-making and patient management, rather than simply adding another layer of services or costs.
[1] Vakili, S., et al., “AI in the clinic: Don’t pay for the tool, pay for the care,” NEJM Catalyst, March, 2026
