Uneven access to radiology AI could deepen healthcare disparities

Uneven patient access to radiology artificial intelligence solutions could be deepening America’s healthcare disparities, according to new Neiman Health Policy Institute research.  

Medicare’s New Technology Add-On Payment, or NTAP, is meant to incentivize the use of innovative technologies—giving hospitals extra money for AI implementation. However, despite the boost, the economics of adoption may make it more difficult for low-resource sites to deploy imaging AI, experts detailed June 24 in the American Journal of Neuroradiology.

To better understand this potential divide, researchers analyzed nationally representative Medicare claims data for the three years following establishment of an add-on payment for a stroke artificial intelligence solution.

“AI tools have the potential to improve speed and accuracy in stroke detection, but our findings show that access to these technologies depends more on where a patient is treated than on their clinical needs,” lead author Casey Pelzl, MPH, principal research scientist at the Neiman Institute, said in a statement June 26. “That has important implications for variation and disparities in stroke care.”

The study used a 5% sample of Medicare claims, extrapolating the results across the U.S. for the final tallies. Its sample spanned 2020 to 2023, covering three years after CMS granted its New Technology Add-On Payment for AI that detects large vessel occlusion in an acute ischemic stroke. Quickly pinpointing this problem—when a major artery supplying blood to the brain is blocked—is crucial so that patients can receive treatment promptly. The add-on payment is meant to bridge the financial gap for the first few years, where standard reimbursement may not fully cover the high costs of these cutting-edge treatments. 

Among over 2,100 acute ischemic stroke episodes across 1,100 healthcare facilities during the study period, add-on payment use appeared to increase gradually. This peaked at around 21% of such stroke cases as of 2022, the authors found. Add-on-payment-backed AI use was most common in healthcare episodes that involved CT imaging, comprehensive stroke centers, or hospitals with over 1,000 beds. No disparities in AI use were observed across patient demographics nor stroke-severity measures. 

However, deeper analysis showed that add-on-payment-backed AI use was about 6 times higher in 2022, and 2 times higher among beneficiaries in America’s “Stroke Belt” across the Southeast. It also was approximately 1.5 times higher at comprehensive stroke centers, the authors reported. Add-on-payment-supported AI use declined in 2023 as the temporary code began to sunset. Overall, AI was used in fewer than 15% of analyzed stroke cases. Hospitals servicing more socioeconomically deprived areas also were “significantly” less likely to use add-on-payment-supported AI. This potentially raises concern that “patients in underserved communities may have reduced access to AI-enhanced stroke evaluation,” the Neiman institute noted.

“AI tools may support faster stroke evaluation, but operational readiness, infrastructure, and clinical workflows play a major role in determining whether these tools are actually used in practice,” co-author Maria X. Sanmartin, PhD, an assistant professor at the Zucker School of Medicine at Hofstra/Northwell, said in the announcement. “When adoption is concentrated in facilities that already excel in stroke care, it misses the opportunity to improve care in less-resourced settings where potential gains are the greatest.” 

Documented barriers to implementing AI tools can include challenges integrating them into existing workflows, provider distrust, and staff learning curves. The latter can be especially impactful at facilities not designated as comprehensive stroke centers, Sanmartin and colleagues wrote. Further, economics can be more challenging at smaller sites needing more resources to support deployment. 

“However, shared AI-as-a-service hubs could allow smaller hospitals to use AI without investing in full AI solutions,” the authors note. “Centralized service models have proven effective in helping under-resourced facilities move to more advanced, efficient and effective care.” 

“Selective access to AI at hospitals with greater financial, technological and training resources may further increase disparities in stroke evaluation and management in the inpatient setting,” the authors added later. “Future research should evaluate the impact of NTAP code approval on diffusion and uptake of AI use in stroke care and improvement of patient outcomes among [acute ischemic stroke] inpatients,” they concluded.

Read more, including potential study limitations, in AJNR, the official journal of the American Society of Neuroradiology. The study also was supported by a grant from the American Heart Association.