Radiology is by far the field with the highest penetration of artificial intelligence (AI) in medicine, but with minimal or no reimbursement, hospital administrators may find a return on investment (ROI) to justify spending on AI. ) may ask where it is. Radiation business We spoke to some key thought leaders to get their thoughts on the value of AI.
The U.S. Food and Drug Administration (FDA) has approved well over 500 market-approved AI medical algorithms as of January 2023. Most of these are related to medical imaging. It contains 396 radiation algorithms. Cardiology is his second most approvals with 58, many of which are also dedicated to cardiac imaging.
“We’ve seen an increasing number of approved algorithms available for commercial use from the US FDA,” said chief medical officer of the American Radiological Society (ACR) Data Sciences Institute and past president of the ACR. Bibb Allen, M.D., FACR, explains. But the adoption rate is very slow, he said. This is because hospitals want assurance that AI will work as intended, improving workflows, reducing costs, and having clear evidence that it will improve patient care. There is also no reimbursement to cover the costs of
“I think it’s important to find the right AI tools and the right value proposition for institutions,” says Allen. “On the payment and policy side, there are two reasonable arguments. On the other hand, payers say, “Wait a minute. Why should I offer an additional payment based on the price of the service if I have already paid for it?” And I think that will be a struggle for health policy makers. ”
Does the AI need to be reimbursed or is it just a practice fee?
The decision to adopt AI is not really up to the radiologist. Keith J. Dreyer, Ph.D., chief scientific officer, DO, PhD, in the American College of Radiological Sciences (ACR) Data Sciences Institute, says managers or entire group practices should first determine if AI adoption is costly. explains that you need to In many cases, that decision rests with hospital or health system executives who do not understand the day-to-day work of radiologists.
“Currently, there are not many use cases that have proven to payers that AI is valuable and willing to reimburse clinicians who use AI. Record (EHR), it’s just a practice fee,” says Dreyer. “We think of AI as a complete category of its own, but when we think of it as a technology, the question is whether this technology will be more efficient. So it’s worth it. Will you pay?”
Dreyer said the lack of reimbursement for AI could also deter health systems from spending more to buy the technology. However, health systems need to consider whether technology can help radiologists’ workflows be enhanced and more efficient, or whether AI can help improve patient outcomes.
“Does AI save radiologists time and increase their efficiency, making AI worth buying?” Dreyer said. “There is no published evidence to prove it 100% yes. I’ve heard anecdotes and seen some small studies that show this, but I don’t believe this to be the case.” But this is case-by-case, because it can be faster or slower depending on how it was implemented, as well as the use case regarding the accuracy of the algorithm.”
AI could help address the growing shortage of radiologists
Charles E. Kahn, Jr., MD MS, editor of the Radiological Society of North America (RSNA) journal, says the role of AI is becoming more important as the United States faces an increasing shortage of radiologists. explained. Radiology: Artificial Intelligence, and Professor and Vice Chair of Radiology at the University of Pennsylvania Perelman School of Medicine. He said the technology will help augment radiologists to improve their workflow and make them more efficient. As an example, he said his AI would take first-pass readings and triage tests into suspected normal and suspected disease. This allows radiologists to focus on reading suspected or more complex cases.
This augmentation to help offset the shortage of radiologists was echoed by Ed Nicol, MD, Consultant Cardiologist, Cardiac Imaging Machine and Emeritus Senior Clinical Lecturer at Kings College London. He is also president-elect of the Society for Cardiovascular CT (SCCT). He believes that simple but time-consuming tasks are performed by AI and helping prioritize exams based on complexity gets more of what really matters from human exam leaders. said it was well worth the cost of
“You don’t pay a radiologist or a cardiologist to draw the line on things. My 7 and 10 year olds would probably do just fine with that. Also, they’ve done their first readings and their first 10 There’s even an AI system that can determine 12 heart CT scans, and 12 of them are abnormal. We can. These technologies already exist, but we don’t. What we’re really paying radiologists and cardiologists is to put the findings in context.” explained Nicole.
What is the Radiation AI Value Proposition?
Allen outlined some key AI value propositions for hospitals to consider.
“AI models can find things that radiologists can’t find, or things that we radiologists can’t see,” Allen said. This includes using radiomics to elucidate brain tumor phenotypes so that appropriate therapies can be selected.
Population health is another area where AI can sift through vast amounts of image data to identify patients with important incidental findings such as pulmonary embolism, coronary artery disease, emphysema, and fatty liver.
“All these things nobody cares about when a patient is in the emergency department for diverticulitis. They’re about to get antibiotics. And that’s the end of their episode of care.” , the fact that they have fatty liver, or have coronary artery calcification, or that they are only 40 years old is buried in the report or is lost, even if it doesn’t matter.Follow Up List for. So this opportunistic AI screening of population health presents a huge opportunity for ROI.”
For AI that can act as a second eye for radiologists, that could be an area where radiologists themselves see the value of such algorithms. Even AI that helps detect non-acute cancers, lung nodules, or other conditions could help prevent scan misses or act as a second opinion. As the radiologist’s second eye, AI can double-check datasets to ensure nothing is overlooked or get a second opinion on questionable areas of an image. AI could also help radiologists detect rare conditions that they may encounter only a few times in their career.
“Having a second eye to help detect more breast cancers might be worth enough to give us extra reassurance. Well worth it.
But AI’s opportunity may also lie in the shift from a per-service pricing model to a value-based payment model.
“With a value-based payment system, you can imagine any tool that will be more efficient. If you are getting the same reward, you can pay a little more for the AI to be more efficient, and the bottom line will be It helps,” Allen explained. .
AI, which can automatically identify pneumothorax in mobile digital x-ray systems and alert clinical care teams to suspected strokes, pulmonary embolisms, and other emergencies, could also improve patient outcomes.
Many AI algorithms also perform complex reproducible measurements, image reformatting, anatomical labeling, anatomical contouring, and other time-consuming tasks. In such cases, AI can help reduce tedious manual processes, improve the information and accuracy of radiology reports, and reduce the cost of reading exams, especially as reimbursement for reading exams is declining. Save time.