The Advanced Medical Technology Association (AdvaMed) and CapView Strategies have collaborated to produce a series of recommendations for improving Medicare coverage of artificial intelligence (AI), endorsing the appointment of a chief AI officer at the Centers for Medicare & Medicaid Services. However, there are risks associated with the use of AI in medicine, risks that may not be clearly understood by the users and vendors of medical AI software.
The AdvaMed/CapView analysis states that the Medicare program already pays for some AI-related medical software, including software used to diagnose diabetic retinopathy as a physician service. The paper states that recent policy decisions by CMS have been helpful, but that the Medicare program does not provide a comprehensive and systematic methodology for assigning an appropriate value for these services. Another issue is that the Medicare program fails to adequately account for the costs associated with these services.
According to the paper, recent surveys indicate that patients maybe be uncomfortable with a physician’s reliance on AI, although many respondents to the surveys stated that they are aware of the benefits, such as the potential for a reduction in the rate of clinical errors. As of January 2022, the code set developed by the American Medical Association for the Current Procedural Terminology system provides a method for classifying AI and augmented intelligence tools. However, the Medicare program does not provide a clear set of guidelines for establishment of AI codes under the Healthcare Common Procedure Coding System (HCPCS). The report further highlights the need for adequate reimbursement levels as incentives for physician and hospital adoption of AI.
In addition to the appointment of a chief AI officer at CMS, the AdvaMed/CapView paper recommends that policies for AI payment under several prospective payment programs be adjusted to adequately cover the costs and benefits of AI. For the Medicare Inpatient Prospective Payment System (IPPS), the report recommends that the cost of AI software be fully considered for payments under the New Technology Add-on Payment (NTAP) program. Other recommendations are related to the prospective payment systems for hospital outpatient services and for physician services paid under Medicare Part B.
There is cause for concern with regard to the potential for liability, however, as the associated legal risks may be experienced by hospitals, by physicians and by the vendors of AI medical software. For physicians and hospitals, the concerns include, but are not limited to, the question of whether the AI platforms they use are compliant with the privacy provisions of the Health Insurance Portability and Accountability Act (HIPAA). However, a physician may not always recognize when medical software incorporates AI, which could prove to be a problem if patient consent is needed for the use of that AI software.
For the vendor of the software, the problem may become more complex if the AI incorporates machine-learning (ML) features that lead to an evolution of the software’s functions. There are several legal theories under which products liability litigation may be brought, such as strict liability and defective design.