AI and healthcare: Who’s footing the bill?
Key Takeaways
Medical technologies that utilize AI are beginning to flood the market.
Currently, the US Centers for Medicare & Medicaid Services (CMS) is covering AI on a per-use basis, which may result in overutilization.
Alternative payment frameworks may be required to remunerate providers for AI.
The age of AI is upon us, and the technology is already changing many fields, including healthcare. AI evaluates many data points to recommend actions in real time, thus resulting in a faster, less expensive, and more accurate diagnosis/prognosis vs human assessments.
The rise of AI in medicine is apparent at the FDA, where 222 AI devices were greenlit between 2015 and 2020, according to a report in npj Digital Medicine.[]
Questions arise, however, about who foots the bill for AI. In this capacity, the US Centers for Medicare & Medicaid Services (CMS) takes the lead.
Inroads at CMS
In October 2019, President Trump issued the Executive Order on Protecting and Improving Medicare for Our Nation’s Seniors (EO 13890) to finalize a new pathway for CMS coverage of AI called the Medicare Coverage of Innovative Technology (MCIT). The CMS Fact Sheet on MCIT explains that MCIT covers FDA-designated breakthrough medical devices. Coverage lasts 4 years and is available as early as the same day as the FDA authorizes breakthrough devices.[]
According to CMS, “We are aware that there are promising technologies being designated by the FDA for expedited development and priority review through the FDA Breakthrough Devices Program."
"However, with the current Medicare coverage options, we have had challenges keeping national Medicare coverage on pace with these innovations because of the availability of clinical evidence required to approve technologies through the NCD [National Coverage Determination] process."
— US Centers for Medicare & Medicaid Services
In response to this issue, and in accordance with EO 13890, CMS is finalizing MCIT to create a more efficient pathway to coverage for breakthrough devices.
The FDA suggests that the 4-year period will permit manufacturers to develop an evidence basis for their treatments, as well as clinical studies involving patients enrolled in a CMS program in a context of broad access to innovation.
“When MCIT coverage sunsets, manufacturers will have all current coverage options available such as a National Coverage Determination (NCD), one or more Local Coverage Determinations (LCD), and claim by claim decisions,” CMS wrote.
Per-use payments
As a first step, note the authors of an article in npj Digital Medicine, CMS has enacted per-use AI payments by covering either AI-specific CPT codes created by the AMA or a New Technology Add-On Payment (NTAP).
NTAPs are part of Medicare’s Inpatient Prospective Payment System (IPPS), and offer additional reimbursements to hospitals above the standard Medicare Severity Diagnosis-Related Group (MS-DRG) payment amount. The NTAP-covered technology must offer “substantial clinical benefit,” and the payment is intended for novel technologies that result in a higher cost for a provider—in excess of current episode-based payment. CMS has also created new CPT codes for AI devices used in outpatient settings.
Such reimbursement for CPT codes are based on Relative Value Units (RVUs) designated by the Medicare Physician Fee Schedule. RVUs reflect physician work, practice expenses, and liability insurance and are modified based on geographically associated medical costs and wage differentials.
Problems with pay-per-use approaches
One potential issue with pay-per-use strategies, say the authors of the npj Digital Medicine article, is overuse. CMS recognizes that although coverage of AI could result in greater access for services such as diabetic retinopathy screening, waste could be a byproduct.
"AI poses a greater risk of overutilization than most devices and offers greater potential for fraud, waste, and abuse."
— Authors, npj Digital Medicine
“Per-use reimbursement for AI may encourage overuse when the payment per-use exceeds the monetary value of the medical net benefit that the use confers to patients and of the productivity gain that the use confers to clinicians,” they wrote. “As with most medical technologies, overuse of AI may lead to overdiagnosis and trigger healthcare spending of little medical value, with a value measured as the improvement in patient outcomes expressed in monetary units.”
The authors draw a parallel, citing the case of stroke, for which decreases in fee-for-service reimbursements have yielded a decrease in the unnecessary utilization of radiology diagnostics.
Alternative forms of reimbursement
To combat the prospect of overutilization, the npj Digital Medicine authors suggest alternative ways to remunerate providers for the cost of AI. These include the following:
Rewarding health systems for patient-centered or process-related outcomes resulting from AI vs paying for AI directly
For AI interventions that confer substantial benefit to the patient and comply with regulatory standards, regulatory bodies can offer advance market commitments garnered from specific healthcare delivery challenges (like the X Prize)
Offering greater financial incentives to health systems for AI that boosts interoperability across multiple settings, and mitigates bias
Employing time-limited add-on payments for AI akin to the transitional drug add-on payment (TDAP) that covers the cost of new pharmaceuticals
Not paying for AI at all, and, instead, health systems only financially benefit from the cost savings inherent to AI
In reality, it will take time to iron out the kinks with regard to AI reimbursement, according to the authors of an article published in Radiology Artificial Intelligence.[]
“Payment for AI in the current fee-for-service environment may be challenging, and sustained adoption of AI may not occur within the framework of the IPPS [Inpatient Prospective Payment System] and PFS [Physician Fee Schedule],” the authors wrote.
“However, as payment systems evolve toward more mature value-based payment models where measuring improvement in quality becomes increasingly important at decreased costs, AI becomes a valuable tool for radiologists and healthcare systems. The entity that receives the most benefit likely will pay for AI and ultimately may consider this payment simply the cost of doing business,” they concluded.
What this means for you
Like AI, remuneration for this technology is emerging. Currently, CMS pays for AI intervention per use. Concerns abound that this approach may result in overutilization. Current CMS reimbursement policies are in place for 4 years, during which further data will be gathered. As the pay-per-use strategy may fall short of reaching healthcare goals, experts predict that payment frameworks will change.