Imaging Pearls ❯ Deep Learning ❯ AI and Reimbursement
|
-- OR -- |
|
- From typists to truckers, people have long worried whether automation will complement or disrupt their work. This debate divides radiology into two camps. The augmentation camp believes AI will handle routine tasks and free up doctors to focus on complex interpretations, interventions, and ‘soft skills’. The replacement camp believes that AI will soon interpret diagnostic imaging to an above-human standard and make many radiologists redundant. The camps can coexist: AI may boost some radiologists and displace others. Yet, most commentators overlook an uncomfortable question: as AI becomes more capable, what happens to the value of radiologists’ work? In other words, cui bono – who benefits from the productivity gains?
Perspective: AI Productivity Will Not Benefit Employed Radiologists
Ruthven, Heathcote Agten, Christoph
Europ J of Radiology AI 3(2025) 1000033 - Yet for AI firms and investors, radiology remains too good an opportunity to ignore. It offers what they value most: a large market with high-quality datasets. As such, radiology is by far the top target for medical AI. By December 2024, 76 % of all AI-enabled medical applications cleared by the FDA target radiology. Meanwhile, ageing populations, rising chronic disease, and cheaper imaging have driven up imaging volumes. Too few radiologists exist to meet demand: the UK expects a 40 % shortage of consultants by 2028, and the US a gap of 122,000 by 2032. In 2025, Hinton appeared to repent, telling The New York Times he spoke “too broadly” and was referring only to image analysis. He now says AI will make radiologists more efficient and accurate – a measured position leaving him room to be right either way. Again, this sidesteps the question: what will these changes mean for radiologists’ careers?
- Many radiologists assume workforce shortages will protect them from automation. But history suggests the opposite: automation is most profitable when labour is scarce and expensive. High radiologist salaries increase the pressure to automate. Once adopted, automation tends to reduce labour’s share of the value it creates. As economists Acemoglu and Restrepo argue, automation “increases the size of the pie, but labour gets a smaller slice.” This displacement effect, where capital substitutes for human tasks, is why productivity gains come at labour’s expense. For instance, in Bessen’s study of the US textile industry he charted how a weaver in 1900 could produce 50 times more cloth than a one in 1800. Rising productivity made cloth cheaper and demand for it soared, as consumers discovered the joy of owning more than one outfit. This created four times more weaving jobs. However, people have a finite desire for cloth. Demand at some point can no longer keep up with productivity, and a tipping point arrives. After 1920, output per worker kept rising by about 3 % a year, but demand for cloth stagnated and employers cut jobs). Bessen’s inverted U-curve charts this boom-and-bust cycle for labour: on the way up, lower costs open markets and create jobs. On the way down, machines replace labour, and the job market shrinks.
- “Could radiology be approaching its own downward slope? Past innovations like CT, PACS, and voice recognition boosted productivity, yet never enough to exceed demand. Today, AI systems increasingly perform at or above radiologist level. Consider three examples from the past year: in breast cancer screening, a South Korean trial shows that AI-assisted readers detect 13.8 % more cancers than radiologists working alone. In MSK radiographs, a Finnish emergency-department study finds that two AI algorithms match expert MSK radiologists at 89 % accuracy. In chest X-ray reporting, a DeepMind study shows that 75 % of radiologists judge AI-generated reports preferable or equivalent to expert-written ones.”
Perspective: AI Productivity Will Not Benefit Employed Radiologists
Ruthven, Heathcote Agten, Christoph
Europ J of Radiology AI 3(2025) 1000033 - Labour is scarce today, but if AI tools can increase productivity by orders of magnitude, practices will employ fewer radiologists. As AI capabilities grow, radiologists will shift from reading to reviewing, from interpreting to confirming. Tasks such as biopsies and patient consultations will likely remain in human hands. Radiology will continue to exist, but the nature and value of the work will change. For some, this brings opportunities. For others, a loss of autonomy, status, or income.
- Radiologists cannot control the pace of AI, but they can prepare a clear ‘Plan B′. Some may seek equity in a practice, move into education or industry, or shift toward procedures and subspecialties that automation is less likely to affect. Younger radiologists should think not only about how to adapt, but how to build additional income streams. Those later in their careers may focus on securing roles that are harder to replace or on reducing clinical time on their own terms. Langlotz’s idea – that radiologists who use AI will replace those who do not – is partly true. But for employed radiologists, the more likely outcome is that their profession will be less in demand and less well paid. Those who remain in the field will do different work, with no guarantee that employers will value, reward, or respect it as they once did.
Perspective: AI Productivity Will Not Benefit Employed Radiologists
Ruthven, Heathcote Agten, Christoph
Europ J of Radiology AI 3(2025) 1000033 - “Employers, private equity firms, and AI vendors will capture most of the profit from AI-driven productivity. Employed radiologists are unlikely to receive these gains. Without structural change in how value is shared, increased productivity will come at the expense of those who still work.”
Perspective: AI Productivity Will Not Benefit Employed Radiologists
Ruthven, Heathcote Agten, Christoph
Europ J of Radiology AI 3(2025) 1000033
- “AI-enabled clinical services (see the Figure for examples) have the potential to simultaneously lower health care spending and improve health outcomes. They may reduce the time and effort devoted to diagnosing disease, which can be associated with improved health outcomes, increased productivity, and lowered labor costs. For example, Viz LVO (Viz.ai) reviews computed tomography angiography to triage patients with suspected stroke. All may also replace more invasive and expensive diagnostic tests, potentially reducing treatment complications, diagnostic spending, and subsequent treatment costs. For example, Heart Flow Analysis (HeartFlow) is advertised as a replacement for invasive and expensive tests to manage coronary artery disease. Moreover, AI-enabled services may allow less specialized physicians to make diagnoses that would otherwise require more specialized physicians; for example, Luminetics Core (Digital Diagnostics) can be used by primary care clinicians to test for diabetic retinopathy. Although cost offsets from AI-enabled clinical services are anticipated, they are not guaranteed, in part because AI-enabled services may increase use.”
How Should Medicare Pay for Artificial Intelligence?
Zink A et al.
JAMA Intern Med. 2024 May 28. doi: 10.1001/jamainternmed.2024.1648. Online ahead of print. - “Currently, CMS payment for AI-enabled services is determined by existing rules for reimbursing new technologie that rely on 1 of 3 payment pathways: (1) bundling the new technology with an existing service without an initial payment adjustment and adjusting the service price over time, (2) bundling with an existing service but including a transitional add-on payment for use of the new technology (until a new price for the service that reflects the AI-enabled component can be established), or (3) paying as a separately payable service. Existing methods for establishing new technology payment to clinicians, including for AI-enabled clinical services, rely on the price set by the firm (which likely reflects monopoly pricing power and CMS treatment of this price as a cost to the clinician) rather than the true cost to the firm of providing the service.”
How Should Medicare Pay for Artificial Intelligence?
Zink A et al.
JAMA Intern Med. 2024 May 28. doi: 10.1001/jamainternmed.2024.1648. Online ahead of print. - “An associated complication is that when the price charged by AI firms is high,CMS is likely to use the separately payable pathway, which is the most generous to clinicians.8This incentivizes AI firms to set ahigher price. In the extreme, if CMS fees cover the full price charged by AI firms, the equilibrium price charged to clinicians would have no limit. This is why transitional new technology reimbursement is typically a fraction of the price charged to clinicians. Competition can also keep prices down, but the market for US Food and Drug Administration– approved AI services within certain clinical settings has been uncompetitive to date despite AI firms not being subject to the same exclusivity restrictions as the pharmaceutical industry.”
How Should Medicare Pay for Artificial Intelligence?
Zink A et al.
JAMA Intern Med. 2024 May 28. doi: 10.1001/jamainternmed.2024.1648. Online ahead of print. - “Because AI has the potential to improve productivity and quality within the health care system,CMSmust promote its development and diffusion. However, it is important that high prices for AI-enabled clinical services are not locked in for perpetuity. Striking this balance is important and may vary on a case-by-case basis, but should be possible with close attention to how AI is associated with workflows and the quality of clinical services.”
How Should Medicare Pay for Artificial Intelligence?
Zink A et al.
JAMA Intern Med. 2024 May 28. doi: 10.1001/jamainternmed.2024.1648. Online ahead of print.
- “There has been a rapid increase in the number of artificial intelligence (AI) systems approved for clinical use by the U.S. Food and Drug Administration (FDA) over the past 20 years and a concomitant increase in the use of these AI systems in medicine. The field of AI raises more questions than answers at this point, one of which is, “Who owns or will own the data used to develop an AI system?” Data, even in health care, are a commodity, and the answer to this question will help guide the answer to the subsequent question, “Who will get paid for AI?” The aim of this article is to discuss data, data creation, and the potential financial stakeholders for the development of AI in medicine and to open the discussion about data ownership and patient consent in the era of AI in medicine.”
Who Will Get Paid for Artificial Intelligence in Medicine?
Colin Rowell, Ronnie Sebro
Radiology: Artificial Intelligence 2022; 4(5):e220054 - "Health care AI systems are not self-learning. Health care professionals first must attend to a patient, diagnose a disease or condition, request tests or imaging, interpret those tests, communicate findings to the patient, and enter data into a system where they can be later accessed. Therefore, during regular clinical care, health care professionals create an asset (data) that has value. It takes health care professionals with years of training, knowledge, and expertise to help train health care AI systems by annotating which patient data correlate to which disease, pathologic condition, or outcome of interest. Furthermore, health care professionals also actively create annotations or diagnoses to be used to train and validate AI systems. This process can be quite time-consuming for health care professional experts in their respective fields.”
Who Will Get Paid for Artificial Intelligence in Medicine?
Colin Rowell, Ronnie Sebro
Radiology: Artificial Intelligence 2022; 4(5):e220054 - "Currently, large databases being utilized for AI research have patients whose health care has been paid for by Medicaid and Medicare. These large databases are the result of millions of dollars of payments by U.S. taxpayers. The argument can be made that data and products derived from data arising from these databases should be made available to the U.S. public for free, as they were financed by U.S. taxpayers. This notion is not novel as, at this time, research studies funded by the National Institutes of Health are made available online to the public for free on PubMed (www.pubmed.gov).”
Who Will Get Paid for Artificial Intelligence in Medicine?
Colin Rowell, Ronnie Sebro
Radiology: Artificial Intelligence 2022; 4(5):e220054 - “Health care systems, in response to the Henrietta Lacks case, have made broad, sweeping, all-inclusive, blanket consent forms that essentially ensure that a patient waives his or her rights to anything derived from that patient. Disclosures and release statements saturate consent forms throughout modern medicine, but how often does the average patient read and understand these consent documents in their entirety? In some cases, the consent for receiving clinical care in a health care system is paired with the abdication of the patient’s rights to their data.”
Who Will Get Paid for Artificial Intelligence in Medicine?
Colin Rowell, Ronnie Sebro
Radiology: Artificial Intelligence 2022; 4(5):e220054 - “A patient’s data may be used to create an algorithm that will only benefit future patients. For example, AI-assisted imaging systems may indeed be beneficial for patients by providing improved sensitivity and diagnostic accuracy, but these benefits extend only to future patients, not to the patients whose data are being used to train the algorithms. At the time of consent, even the health care professional obtaining consent has no idea of all the potential ramifications of the patient giving up their rights to their data. Also, patients are often not given the opportunity to refuse the use of their data for research purposes. Patient data utilized for research studies with the goal of commercialization may require a separate consent document explaining to the patients that signing the consent renounces their potential future financial claims.”
Who Will Get Paid for Artificial Intelligence in Medicine?
Colin Rowell, Ronnie Sebro
Radiology: Artificial Intelligence 2022; 4(5):e220054 - “Some have proposed dynamic consent where patients should not only be notified when their data are used in research but also informed of the implications of the patient’s data being used in this research. However, such a scheme is extremely difficult because it requires continuous communication between patients and the individuals using their data. Dynamic consent would also not be possible for deceased patients. Meta consent is based on the idea that patients are given the opportunity to make choices on the basis of their preferences for how and when to provide consent. The advantages and disadvantages of meta consents are still being debated. Another challenging issue is that a patient may revoke his or her consent at any time.”
Who Will Get Paid for Artificial Intelligence in Medicine?
Colin Rowell, Ronnie Sebro
Radiology: Artificial Intelligence 2022; 4(5):e220054 - “The patient data are anonymized in a way that meets HIPAA regulations governing patient privacy rights and preserves privacy for health data sharing. Further research is required to evaluate the ethics surrounding patient consent in this setting. The local ethics review boards have tremendous importance in evaluating the ethics surrounding patient consent and should ensure patients genuinely understand what they are contributing their data to and the scope of any research that will be performed using the patient’s data. A recent lawsuit in the United Kingdom involves a patient plaintiff who was very concerned that his data were shared with Google’s AI firm DeepMind by the Royal Free London National Health Service Foundation Trust.”
Who Will Get Paid for Artificial Intelligence in Medicine?
Colin Rowell, Ronnie Sebro
Radiology: Artificial Intelligence 2022; 4(5):e220054 - "Health care systems may benefit through the sale of presumably anonymized patient data to AI companies. However, such actions raise the question of who owns the data within a health care system for extramural sharing of patient data. Are the data owned by the health care system, the department within the health care system that generated the data, the chief informatics officer, or the treating physician? Furthermore, consider a scenario where oncologists at an institution consult with an AI company and share the imaging data from their patients with that AI company. Are the data owned by the oncologists, the health care system, or the radiology department that stores the imaging data? Are the data also co-owned by others, such as the patient whose data are being shared? Further work is required to clearly understand the intramural ownership of data within health care systems. Further research is also required to understand good data sharing practices.”
Who Will Get Paid for Artificial Intelligence in Medicine?
Colin Rowell, Ronnie Sebro
Radiology: Artificial Intelligence 2022; 4(5):e220054 - "There are outstanding questions regarding reimbursement for the use of AI in health care. Given that data are now a commodity, the question of data ownership in this context is of paramount importance. There is no consensus on who owns medical data, or for how long. There are multiple stakeholders and multiple individuals who are essential when creating an AI system. Dissecting the individual contribution of each stakeholder and each individual to the development of an AI system is difficult and, in some cases, intractable. An urgent discussion is required in the scientific community to really understand data ownership as it pertains to medical AI and how its use will be reimbursed.”
Who Will Get Paid for Artificial Intelligence in Medicine?
Colin Rowell, Ronnie Sebro
Radiology: Artificial Intelligence 2022; 4(5):e220054 - The Patient and Consent for Data Use for AI?
- What if the patient says no?
- What if the patients changes their mind and wants to no longer grant permission?
- Does the payer (insurance or US Government) own the data they paid for ?
- Does ownership claims also mean potential liability claims?
- “If AI is an unreimbursed business expense, it changes the potential return on investment for all of the outside money that continues to pour into companies creating products using AI. When even the troglodytes of radiology see a future with AI benefiting both patients and the specialty, we should perhaps temper our enthusiasm because of these financial realities. The barriers to entry for new products and services in health care are high, and for good reason. But without the promise of governmental largesse or large inflows of reimbursements from private payers, vendors may take a pass on investing resources in radiology or health care-specific applications for AI.”
Artificial Intelligence: Who Pays and How? Schoppe, Kurt Journal of the American College of Radiology (in press) - “By now, it’s almost old news: big data will transform medicine. It’s essential to remember, however, that data by themselves are useless. To be useful, data must be analyzed, interpreted, and acted on. Thus, it is algorithms —
not data sets — that will prove transformative.”
Predicting the Future — Big Data, Machine Learning, and Clinical Medicine Obermeyer Z, Emanuel EJ N Engl J Med 375;13 September 29, 2016 - “Most computer-based algorithms in medicine are “expert systems” — rule sets encoding knowledge on a given topic, which are applied to draw conclusions about specific clinical scenarios, such as detecting drug interactions or judging the appropriateness of obtaining imaging. Expert systems work the way an ideal medical student would: they take general principles about medicine and apply them to new patients.”
Predicting the Future — Big Data, Machine Learning, and Clinical Medicine Obermeyer Z, Emanuel EJ N Engl J Med 375;13 September 29, 2016 - “Machine learning, conversely, approaches problems as a doctor progressing through residency might: by learning rules from data. Starting with patient-level observations, algorithms sift through vast numbers of variables, looking for combinations that reliably predict outcomes.”
Predicting the Future — Big Data, Machine Learning, and Clinical Medicine Obermeyer Z, Emanuel EJ N Engl J Med 375;13 September 29, 2016