Experts call for greater scrutiny around conflicts of interest between radiologists, AI developers
Researchers are calling for greater transparency around ties between radiologists and artificial intelligence developers.
The U.S. Food and Drug Administration has cleared hundreds of AI-enabled devices over the last decade, with many in medical imaging. However, rapid adoption raises concern about potential conflicts of interest, with many commenters on one recent AI-related FDA request for feedback failing to disclose their industry ties.
Researchers recently aimed to assess these potential conflicts, utilizing information from the Open Payments Database, created in 2013 to shine light on docs who accept money from industry sources. They found that manufacturers of over 90% of FDA-listed AI devices publicly disclosed zero payments to healthcare professionals nor teaching hospitals between 2017 and 2023.
“This is likely a consequence of gaps in disclosure requirements, inconsistent compliance and ambiguities in the regulatory classification of AI tools,” Dr. David-Dan Nguyen, MPH, a urology resident with the University of Toronto, and co-authors wrote Sept. 22 in JAMA. “For example, manufacturers of radiology devices may be less likely to disclose payments because many fall outside current reporting requirements due to differing clearance pathways, reimbursement models and uncertain regulatory classification.”
For the study, Nguyen et al. used the FDA’s AI/ML (Machine Learning)-Enabled Medical Device List to pinpoint products that satisfied premarket requirements. They then analyzed payments to healthcare professionals and teaching hospitals via the Centers for Medicare & Medicaid Services’ database. Of 846 FDA-listed AI devices, only 9% (or 79) were linked to payments to hospitals or providers between 2017 and 2023. Payments totaled $120.2 million, with about 49% relating to research and the rest going for “general” purposes, such as consulting work or gifts.
Total yearly industry payments climbed from $17.3 million in 2017 up to $24.6 million by the end of the study period. Meanwhile, the number of AI devices for which manufacturers reported industry payments increased from 32 up to 53 by 2023. Nearly half (about $59.4 million, or 49%) were associated with cardiovascular care, followed by radiology ($39.7 million or 33%). The median seven-year total payment per AI product was $142,538 in the “general” category and $238,362 for research, with both figures staying stable over time. In radiology, about $14 million went toward “general” payments while $25.6 million covered research.
A total of over 53,137 healthcare professionals and 883 teaching hospitals received payments during the study period. The number increased from 7,911 physicians and 116 hospitals in 2017 up to 14,066 physicians, 4,091 non-docs (CMS started tracking their industry ties in 2021) and 153 teaching hospitals by the end of the study period.
Research payments stayed stable while general allotments doubled, “suggesting a growth in marketing with limited parallel investment in research.” This rate of growth contrasts with the overall trend in payments from pharmaceutical and traditional device manufacturers, “which have declined over time.” Only a subset of FDA-cleared AI devices was available in Open Payments, indicating these industry conflicts are likely underestimated.
“Reporting requirements should be modernized to comprehensively capture AI/ML applications and ensure adequate transparency and oversight,” the authors urged. “These findings highlight the need for greater transparency as AI/ML technologies are adopted into clinical practice,” they added later.
Marty Stempniak has covered healthcare since 2012, with his byline appearing in the American Hospital Association's member magazine, Modern Healthcare and McKnight's. Prior to that, he wrote about village government and local business for his hometown newspaper in Oak Park, Illinois. He won a Peter Lisagor and Gold EXCEL awards in 2017 for his coverage of the opioid epidemic.