UW Health and Epic say policymakers should provide “targeted investment and support” to ensure AI in health care benefits rural communities and other underserved areas.
The Madison health care system and Verona-based electronic health records business yesterday released a report on how AI should be regulated, drawing from a summit they held last month in Washington, D.C. The roundtable event had more than 25 participants including those representing health systems, the offices of federal lawmakers and the U.S. Department of Health and Human Services.
The report includes a number of recommendations for lawmakers and federal officials, including offering incentives to encourage broader adoption of AI technologies in “under-resourced” settings, creating registries for AI tools to track their use across different settings and offering open-source access to such tools.
Chero Goswami, chief information and digital officer for UW Health, notes access to technology is already “somewhat divided” between large academic medical centers and rural hospitals across the country.
“So how do we ensure that, while a one-size-fits-all may not be possible, we don’t further the digital divide, as the term goes, and ensure that the rural hospitals also have access and can afford the access to these technologies,” he said in remarks provided by the health care system.
As artificial intelligence becomes more widespread in health care, Goswami emphasizes the importance of transparency in how it’s being used.
“It is to protect the interests of patients,” he said, referring to the goal of issuing this report. “Most of us, even in the industry, don’t have a clear understanding right now where AI begins and where AI ends. And without that clarity, we often may up inadvertently causing panic and anxiety to our patients and their families.”
Pediatric care experts at last month’s discussion said child health data is often excluded from AI models, leading to “gaps in AI performance in pediatric radiology” as algorithms trained mainly on adult cases struggle to accurately diagnose and predict outcomes for children.
The report also calls for efforts to foster “local validation” of AI models, noting this will be important for tailoring specific applications to certain patient groups and health care operations. Report authors say this will be “especially critical for widespread adoption outside of major metropolitan areas” and top clinical care sites.
They also argue any new regulations should extend beyond the patient-provider relationship to encompass any interaction involving patient data. They say “regulatory frameworks must adapt” as the health care industry expands beyond the hospital or clinic into telehealth, at-home care and community organizations.
Goswami draws a line between national oversight of AI and local efforts to regulate it, noting patients and communities in Wisconsin are “very different” from those in southern California and New England, for example.
“So we have to manage that balance of local versus national when it comes to passing these regulations,” he said.
See the report.
See recent related stories on Gov. Tony Evers’ AI task force and an AI-focused startup accelerator program.
