Show me the money: Who will foot the AI bill? And how?
Artificial intelligence (AI) has the potential to revolutionize patient-care and serve as a valuable tool for radiologists. But with all its promise, a recent editorialist asked: Has anyone thought about how it will all be paid for? And by whom?
Kurt Schoppe, MD, with Radiology Associations of North Texas in Fort Worth analyzed this question in the Journal of the American College of Radiology (JACR). He addressed the issue through the three most common ways medical procedures are currently reimbursed: Current Procedural Terminology (CPT) codes and relative value units (RVUs), hospital outpatient prospective payment schedule (HOPPS) and business expenses.
CPT Codes and RVUs
Nearly every procedure undertaken by a radiologist requires a CPT code. Schoppe argued that it will be difficult for AI software to jump the hurdles required to gain such a code. He writes, one: AI will have trouble navigating the peer-review process to demonstrate efficacy and safety; and two: it will be hard to demonstrate the need for the procedure to be performed by a large number of U.S. physicians.
Hypothetically clearing those hurdles, the tool would then be examined by the Relative Value Scale Update Committee (RUC) to receive the necessary relative value unit (RVU) code. The problem here, he says, is that the criteria are largely based on the time it takes to perform a service, for which many AI tools he has seen take “minimal to no physician work,” he wrote.
Practices break down expenses into either direct or indirect. Indirect expenses require a complicated Medicare formula along with direct expense data. Schoppe fears AI may encounter the same expensing problem that has plagued PACS.
Radiologists may view many costs associated with PACS as direct expenses, but Medicare thinks otherwise, he argued. The only expenses that fall into this category are computer monitors and the CPU tower to run the software. Hardware and software inputs are considered indirect.
“Bureaucratically mischaracterized like PACS is likely the fate of many AI tools,” he wrote. “If there is no direct attributable expense on a per-patient basis for the AI software, then it, too, will be an indirect practice expense. That means no direct payments for your accountants to point to for a return on investment.”
HOPPS
Medicare uses self-reported hospital cost data to reimburse institutions for technical components of radiology services rendered on outpatients in the hospital setting via HOPPS. However, Schoppe points out this number is minimized when the impact of incremental costs for other supplies or hardware are spread across thousands of hospitals.
Many hospitals do not even report these smalls costs, “so if hospitals do not report their costs for using AI software correctly, Medicare will not reimburse them,” he wrote.
“Unfortunately, fixing these issues turns into a massive collective action problem; the technical reimbursements will only go up if almost all of the hospitals correct their cost reporting together, Schoppe wrote. “And this is before adding AI as another component of costs.”
Business Expense
As long as neither the government nor private payers steps up to reimburse hospitals for AI products, Schiff believes paying for these tools will become another cost for hospitals.
“[W]ithout the promise of governmental largesse or large inflows of reimbursements from private payers, vendors may take a pass on investing resources in radiology or healthcare-specific applications for AI,” he concluded. “And that would be a shame, because we need so desperately to improve, both for our own sake as physicians and for our patients.”
Matt joined Chicago’s TriMed team in 2018 covering all areas of health imaging after two years reporting on the hospital field. He holds a bachelor’s in English from UIC, and enjoys a good cup of coffee and an interesting documentary.