According to a survey released on September 22, the AI/Machine Learning (ML) developers paid $39.7 million for radiological devices between 2017 and 2023. Jama.
The payments are part of the total $179 million awarded to medical professionals and hospitals within the time frame, secondly to $59.4 million paid for cardiovascular devices, which was paid during the study by David-Dan Nguyen, MD, University of Toronto, Ontario, Canada.
“Finance relationships with manufacturers can have a significant impact on decisions to adopt, implement or promote these technologies,” the group wrote.
Rapidly approved by the US Food and Drug Administration (FDA) over the past decade, researchers wrote that AI/ML medical devices have raised “well-known concerns” about conflicts of interest. Thus, this analysis examined the scope and nature of payments made to health professionals and teaching hospitals by manufacturers of FDA-certified AI/ML-enabled devices.
The group first identified 846 devices from the FDA List of AI/ML-compatible medical devices. Of these, 79 (9.3%) were related to payments to healthcare professionals and hospitals between 2017 and 2023, based on an analysis of the US Centers for Medicare and Medicaid Services' Open Payment Database.
In total, $1200.2 million was a general payment over the period, and $59.1 million in research payments were provided. Annual payments increased from $17.3 million in 2017 to $24.6 million in 2023, increasing primarily general payments, doubling from $6.6 million to $13.3 million.
Additionally, the number of devices that manufacturers reported payments increased from 32 in 2017 to 53 in 2023, with the median payments per product generally of $142,538 and $238,362 in the study.
Finally, the researchers noted that the number of recipients increased from 7,911 doctors and 116 teaching hospitals in 2017 to 14,066 doctors, 4,091 non-medical health professionals, and 153 teaching hospitals in 2023.
Ultimately, these reported payments say that these reports could be just the tip of the iceberg, as more than 90% of FDA-registered AI/ML-enabled devices did not publish such payments.
“For example, manufacturers of radiological devices may be less likely to disclose payments as many people fall outside current reporting requirements due to different clearance pathways, reimbursement models, and uncertain regulatory classifications,” they wrote.
In conclusion, the researchers proposed that reporting requirements need to be modernised to comprehensively capture AI/ML applications and ensure appropriate transparency and monitoring.
Complete research is available here.
