ACR Bulletin

Covering topics relevant to the practice of radiology

Into the Unknown

Much remains to be seen regarding how AI will affect our practices, patients, and bottom lines.
Jump to Article

Until AI’s role in radiology is better understood, it is difficult to design a payment model; until payment exists for AI, its development and role will remain unclear.

—Bulletin Author
December 28, 2020

Nearly all radiologists have faced the unknown. Common scenarios include encroachment upon one’s turf, decisions on whether a new cardiaccapable CT scanner will be located in radiology or cardiology, and practice mergers changing competition in regional markets. Handling these situations requires acknowledging that change is coming without knowing what form it will take and realizing its potential effects on your practice’s operations and income.  

There’s no blueprint to handle the unknown. To set themselves up for success, radiologists can navigate these waters by gaining expertise, becoming involved in decision-making, and shaping the ultimate outcome. AI is another example of the unknown. We have heard predictions that it might replace radiologists someday. In reality, no one knows where AI is going or the course it will take.   Similarly, how AI will be reimbursed remains unknown. In fact, the lack of payment policy is a contributing factor to the unpredictability of AI integration into clinical radiology practice and sets up a “chicken and egg” conundrum. Until AI’s role in radiology is better understood, it is difficult to design a payment model; until payment exists for AI, its development and role will remain unclear.  

Many factors contribute to the lack of established payment policy for AI. First, AI technology development is changing faster than the existing payment policy process, which is oriented toward traditional technologies that evolve slowly. Second, the term “AI” means different things to different people and is often used in an all-encompassing manner. It is applied to algorithms that are autonomous and function independently of the radiologist, to technologies that identify imperceptible findings that actually increase complexity and interpretive tasks of radiologists — and to everything in between. A one-size-fits-all approach is limiting and problematic.   A successful AI payment policy needs to recognize and incorporate how AI is applied clinically and its effect on radiologists’ work. For example, a well-designed payment system would distinguish among AI technologies that complement radiologists, supplement radiologists, or function autonomously.  

Complementary AI technology improves a radiologist’s performance by acting as a separate set of eyes (conducting a task that human beings can accomplish). Computer-aided detection (CAD) in mammography is a long-standing example; the computer and radiologist both search for microcalcifications and other actionable findings and therefore complement each other. Interestingly, payment policy for CAD has existed for more than a decade and is incorporated into mammography billing codes.   Supplementary AI technology adds to what a radiologist can accomplish. Technologies exist that identify findings below the threshold of human perception; examples include detecting otherwise imperceptible early changes of emphysema or vertebral body compression fractures. This AI information adds work by requiring the radiologist to analyze and interpret those additional findings in the context of an individual patient. A well-designed and fair payment system would recognize and pay for the additional work resulting from incorporating AI into clinical care. Autonomous AI systems analyze imaging data and generate reports. Despite some predictions, this would most likely change the radiologist’s role rather than obviate it. In fact, radiologists might become more integral to patient care if these models involve integrating data from multiple clinical sources, such as amalgamating imaging findings with pathology results and oncologic notes. 

AI payment policy is in its infancy and its future is unclear. Where it ends up will undoubtedly affect radiologists’ operations and reimbursement. As these tools evolve, the ACR is helping radiologists do what they do best. The College is enabling us to become experts, involving us in the decision-making process, and positioning us to influence the course and outcome of AI payment policy.  

Specifically, the ACR’s Data Science Institute® works with industry leaders to shape AI technology in a radiologist-friendly manner. The ACR Commission on Economics’ Coding and Nomenclature Committee crafts CPT® codes and represents radiologists’ interests to the AMA’s CPT Editorial Panel that is charged with approving these codes. The Relative Value Scale Update Committee advocates for appropriate reimbursement for these approved codes. Finally, the Commission on Governmental Relations monitors regulations affecting utilization and reimbursement of new technology. 

AI is coming and will change how radiologists work and how we are reimbursed. At this time, the future of AI reimbursement is unknown. However, the ACR is positioned to influence the payment process and represent its members’ interests in this new frontier.

Author Timothy A. Crummy, MD, MHA, FACR,  Vice Chair, ACR Coding and Nomenclature Committee