The Economics of AI Adoption

In today’s value-based healthcare model, providers strive to meet the Centers for Medicare & Medicaid Services’ (CMS) goals to improve care for individuals and populations while reducing costs. Many radiologists have implemented artificial intelligence (AI) to help meet these goals. However, reimbursement models haven’t kept pace with the growing use of AI, and the CMS and radiologists are grappling with how to properly reimburse and bill for AI-aided image analysis.

To shed light on this quickly evolving topic, Applied Radiology hosted a recent roundtable discussion on the economics of AI. Sponsored by Guerbet, this was the second segment of AI Insights, a series of webinars focusing on the growing use of AI in healthcare.

The three-member panel was led by Sonia Gupta, MD, chief medical officer of Enterprise Imaging at Optum in Nashville, Tennessee; it also included Ryan Lee, MD, MBA, MRMD, a neuroradiologist and chair of the department of radiology at Einstein Healthcare Network in Philadelphia, Pennsylvania; and Melissa Chen, MD, a clinical neuroradiologist at The University of Texas MD Anderson Cancer Center in Houston, Texas.

Dr. Chen is also an advisor for the American Society of Neuroradiology to the American Medical Association (AMA) Relative Value Scale Update Committee. Drs Gupta and Lee also serve on the editorial advisory board of Applied Radiology.

Changing reimbursements and CPT codes

Dr Chen said healthcare reimbursement in the United States is complex, which has stalled a clear plan for reimbursement of AI-supported image interpretation. Despite this slowdown, progress is still being made.

Some AI tools fall under the New Technology Add-On Payment (NTAP) designation, which makes them eligible for reimbursement beyond the standard payment made to hospitals. Today, two radiology AI technologies are approved for NTAPs: stroke-detection software and a new algorithm for cardiac ultrasonography that helps clinicians not trained in sonography to obtain diagnostic-quality images.

“This payment shows they want to democratize healthcare and expand access to a lot of patients, such as in a rural hospital with a shortage of ultrasound technologists,” said Dr Chen.



Category I CPT codes guarantee reimbursement for procedures, and Category III CPT codes cover  emerging technologies, which aren’t always reimbursable. Recently, the CMS has proposed payment for some Category III codes under a new technology ambulatory procedure code.

“It's good news that we're starting to see more and more algorithms being adopted, and I think it's just a matter of ensuring that we have access across all different types of ways that patients may need to access AI,” said Dr Chen.

Dr Chen dives deeper into this topic in her 2021 paper titled, Who Will Pay for AI?.



Going beyond the dollar

Dr Lee said focusing solely on AI reimbursement can obscure the wide-ranging benefits of AI, including reducing burnout among radiologists.

“Some algorithms save a lot of mental effort, even if you can't prove it’s saving time. That is just as important in this environment where radiologists are burnt out. Anything we can do to alleviate burnout and improve wellness is something we should consider. While we're waiting to see how the reimbursement sorts itself out, we have to look at other ways that AI might help,” he said.

Dr Gupta agreed, observing that burnout can have significant financial implications in the future. “Even if we cannot demonstrate a very clear ROI with some of these algorithms, if you can prevent one radiologist from resigning, going part-time, or cutting their career short, the ROI on that has many multiples,” she said.

Dr Chen added that, “Looking at benefits other than a dollar-for-dollar reimbursement is important, while ensuring it doesn't decrease the efficiency of radiologists.” She speculated that the future of AI payment might lie in bundling AI solutions with imaging equipment purchases, effectively transforming AI from a reimbursable code to a business expense.



Pilot programs are a “win-win”

Dr Lee believes in the power of AI pilot programs, calling them a “win-win” for developers and healthcare institutions. “We’re willing to put in sweat equity to help try to develop an algorithm, and the AI vendor is able to try out the algorithm in our environment, which has a lot of utility,” he said.

Einstein Healthcare Network often collaborates with AI developers in this way, with some partnerships lasting up to five years and others converting to commercial contracts. Under these collaborations, Einstein can “test drive” a given technology in their own environment for six months or more to ensure it will work within its PACS and IT systems.

Dr Lee noted that the hospital is essentially co-developing the product with the company, which can cause its own set of challenges, versus an established tool  that can be test-driven before any purchasing decisions are made.



Advice for AI implementation

As practices and facilities assess which AI tools to invest in, Dr Lee said the first step is to ensure the PACS and IT teams are up to the challenge. “Being able to deploy is critical because these are not small products to implement,” he said.

The next step is to evaluate a potential vendor’s troubleshooting capabilities, which requires communication between an institution’s own team and the vendor’s team. “That’s often overlooked because things can fail on so many different levels, on the vendor side or on your side. Having that ability to troubleshoot and having both those teams interact is key,” he said.

Third, it’s critical to get buy-in from radiologists before implementing any new technology. At Einstein Healthcare Network, radiology department leaders introduce the technology to selected radiologists before presenting it to the entire department, which Dr Lee says has improved adoption rates. “You can have the best algorithm in the world, but if nobody’s using it, it’s for nothing,” he said.

Dr Gupta agreed. “A lot of departments and practices have their group of super users bridge that gap and try it out first, then teach others in the department or the practice how to use the system. That’s been a good recipe for success in some of these pilots,” she said.



Keeping up with the changing AI landscape

Dr Chen said reimbursement for AI is evolving, requiring organizations like the American College of Radiology and the AMA to provide radiologists with continuing support. “It’s encouraging to see that physician-led organizations are making sure radiologists are part of the foundation of how we create CPT codes, and potentially that could lead to reimbursement,” she said.

Ultimately, the radiologist’s role will become increasingly essential to choosing and deploying AI solutions. “It’s easy to see how these tools will get better over time to help the radiologist and ultimately the patient. That’s one reason why I’m interested in helping develop these, for the patients and for the radiologist,” said Dr Lee.

Hear the entire panel discussion here!

© Anderson Publishing, Ltd. 2024 All rights reserved. Reproduction in whole or part without express written permission Is strictly prohibited.