AI in healthcare is already with us, as evidenced by more than 500 AI/machine learning (ML)-enabled medical devices, primarily FDA-approved for use in radiology.
Radiologists, who are currently under increasing pressure to interpret more and more images, use AI-based algorithms to facilitate detection of pathology and reduce diagnostic errors. In turn, AI helps to reduce downstream costs to healthcare institutions. And, most importantly, patients can potentially benefit from faster turnaround times and early identification of at-risk factors in preventive care1.
However, the widespread and appropriate use of AI in healthcare faces challenges, including accommodations for reimbursement. With that in mind, it is essential to understand existing reimbursement models before proceeding with specific recommendations.
The largest healthcare payer in the United States is the Centers for Medicare and Medicaid Services (CMS). Similar to other federal agencies, the CMS is a complex organization with numerous systems and moving parts. Key terms to consider within this context are Current Procedural Terminology (CPT®) and Diagnosis Related Groups (DRGs).
CPT is the standardized language for codes that represent healthcare services provided to patients by doctors and healthcare professionals in radiology practice office settings. Similarly, DRGs serve the same purpose as CPTs except to categorize inpatient care for hospital reimbursement.
So, how does reimbursement accommodate AI? Certain CMS criteria must be met before reimbursement is allowed2. In hospitals, New Technology Add-on Payments (NTAPs) are an additional payment mechanism on top of DRGs for inpatient settings. This is meant to incentivize the use of novel technologies.
Reimbursement for using these exciting advancements would likely encourage greater use. While it’s too early to predict if per-use reimbursement would necessarily lead to AI overuse, the CMS aims for most beneficiaries to select value-based care arrangements by 20303.
Meanwhile, healthcare stakeholders could focus on reimbursement strategies that prioritize value and outcomes. For instance, higher reimbursement might reward AI algorithms proven to enhance patient outcomes in post-marketing studies. Financial incentives could also address AI shortcomings.
Rare diagnoses or traditionally underserved communities may not receive the same benefit from AI due to lack of representation in initial development4. AI devices that demonstrate interoperability across broader groups could be reimbursed more to encourage fair delivery to the people who may benefit the most.
Lastly, AI may derive immense value through opportunistic screening of frequently obtained imaging. An example would be AI programs assessing for osteoporosis or coronary artery calcification while the radiologist focuses on the clinical indication behind the imaging. The exact cost-effectiveness of such a practice is unknown, but modeled simulations suggest that this approach is potentially efficacious5.
Overall, reimbursement for AI in radiology currently operates on a per-use basis with CPTs and NTAPs. However, the future lies in value-based care, prioritizing patient outcomes and diversity in healthcare delivery.
References
- Kasireddy, M. and Lee, R.K. “The Economics of Artificial Intelligence: Focusing on the Metrics,” Applied Radiology, 2022: 13–15. Available at: https://appliedradiology.com/articles/the-economics-of-artificial-intelligence-focusing-on-the-metrics. Accessed Nov. 6, 2023.
- Lobig, F., Subramanian, D., Blankenburg, M., et al. “To Pay or Not to Pay for Artificial Intelligence Applications in Radiology,” npj Digital Medicine, 2023: 6(1). Available at: https://doi.org/10.1038/s41746-023-00861-4. Accessed Nov. 6, 2023.
- The Center for Medicare and Medicaid Innovation (CMS Innovation Center). Strategy Refresh White Paper (n.d.). Available at: https://www.cms.gov/priorities/innovation/strategic-direction-whitepaper. Accessed Nov. 6, 2023.
- Obermeyer, Z., Powers, B., Vogeli, C., et al. “Dissecting Racial Bias in an Algorithm Used to Manage the Health of Populations,” Science, 2019: 366(6464), 447–453. Available at: https://doi.org/10.1126/science.aax2342. Accessed Nov. 6, 2023.
- Pickhardt, P.J., Correale, L. and Hassan, C. “AI-Based Opportunistic CT Screening of Incidental Cardiovascular Disease, Osteoporosis, and Sarcopenia: Cost-Effectiveness Analysis,” Abdominal Radiology, 2023. Available at: https://doi.org/10.1007/s00261-023-03800-9. Accessed Nov. 6, 2023.
Additional ACR Topic References
- Hudnall, C.E. “Choosing AI,” ACR Bulletin, April 2023. Accessed Nov. 6, 2023.
- Nicola, G.N. “A New Era,” ACR Bulletin, November 2020. Accessed Nov. 6, 2023.
