Current Approaches to AI Adoption in Radiology
Welcome to our newsletter on neuroradiology and AI where we have collected some of the recent news within the field. If you would like to be informed when the next Cercare Bulletin is out, please sign up here.
U.K. introduces guidance for integrating AI into radiology workflow: The U.K. Royal College of Radiologists (RCR) published standards for radiology departments to use when integrating artificial intelligence into their established PACS or RIS. They share five recommendations that include integrating AI into reporting workflows in ways that don’t add extra burdens to radiologists; ensuring that AI algorithm accuracy is clearly declared for patient management decision-making; communicating AI findings to the RIS using existing HL7 global technical standards; communicating AI findings to PACS with existing DICOM global technical standards; and ensuring a robust workflow so that AI analysis is complete and available on PACS before humans begin image interpretation.
How AI reimbursement will drive adoption: One barrier to radiologists adopting AI algorithms is the reimbursement models. In the U.S., relying on value-based care payment models is a potential solution, given the more difficult route of reimbursing via the fee-for-service model. In the U.S., 72% of healthcare-approved AI by the Food and Drug Administration (FDA) is in the radiology field, as of early 2020. Two AI tools are currently eligible for reimbursement for three years under a special program by the Centers for Medicare and Medicaid Services (CMS), but that doesn’t mean private payers will necessarily cover them.
Authors of a special report in Radiology: Artificial Intelligence suggest that organizations deriving benefit from AI algorithm use could consider paying for it themselves, as the cost of doing business. AI programs that improve patient health and reduce costs could be a worthwhile investment for radiology practices in, even if they are not directly reimbursed.
Reimbursement issues were also cited in a study from the American College of Radiology (ACR) Data Science Institute (DSI). The study showed that 30% of members are using AI in their clinical practice, though it’s more common for large practices to use it, compared to small ones. Most used it for image interpretation, and the top uses were for detecting intracranial hemorrhage, pulmonary emboli, and mammographic abnormalities. More than half of respondents asked ACR to evaluate how to improve reimbursement for AI clinical use, as that is a key factor in widespread adoption.
One way AI can reduce costs if practices pay for it themselves, is with clinical decision support (CDS). Medicare will mandate providers use CDS for certain imaging procedures starting in 2022. AI assisting with the CDS process can help reduce the financial risks of declined radiology reimbursements within health systems, for this Medicare mandate.
American College of Radiology’s new database of federally-cleared algorithms The ACR Data Science Institute developed a searchable catalog with 111 class 2 medical imaging AI algorithms that are FDA-cleared. Users can search the database using categories like body area, modality, subspecialty or clearance date. A 2020 ACR survey of its members showed that 30% of radiology practices use some type of AI in their practice.
Developing effective and useful AI algorithms is only one part of the equation. It’s important to consider the regulatory and payment processes as well, as these impact adoption. Vendors can create the best algorithms to help radiologists with their workflow and to improve patient care but without these algorithms getting implemented, they are not able to move the field forward so they can help the relevant players.
Would you like to stay up-to-date with the latest news in AI and neuroradiology? Stay informed with our Cercare Bulletin.
Read other news