Panellists summarize the state of health AI regulations, standards and challenges | Hogan Lovells
Melissa Bianchi, partner in Hogan Lovells’ healthcare practice and leader of the firm’s digital healthcare initiative, speculated on the potential of artificial intelligence to revolutionize the healthcare industry to kick off the summit’s discussion of the state of regulatory affairs for AI in healthcare and its ability to eliminate waste to reduce, streamline payments and better diagnose patients. However, she noted that there are still questions about how the industry can best capitalize on this potential and what are the barriers preventing it from realizing its full potential.
What is “artificial intelligence”?
Panellists first discussed how AI should be defined with Kathleen Blake, MD, MPH, Senior Advisor to the American Medical Association (AMA), first emphasizing how AI product sponsors are required to demonstrate that the population in which the evidence is developed has been representing the communities around the world where the technology is being used and that population-based evidence is needed to safely expand the use of AI to new communities. dr Blake emphasized that AI must show that it improves equity for all and that it implies meaningful patient outcomes.
Following the remarks by Dr. Blake mentioned Kerri Haresign, director of technology and standards at the Consumer Technology Association (CTA), her organization’s two published AI standards that address definitions of AI and the importance of trustworthiness in AI. Ms Haresign warned that “we’re stuck trying to define AI broadly” by distinguishing between “assisted intelligence” and “autonomous intelligence,” the latter category requiring no human intervention. To define AI, said Dr. Blake that the AMA views AI as “augmented” intelligence and recommended that the focus should be on incremental gains through new technology.
Highlighting the importance of setting standards in AI, Ms Bianchi described that part of the goal of standards is to “build something that can be regulated and encourage more efficient, faster approvals.” dr Blake urged industry stakeholders to involve patients early in study planning.
Overcoming data challenges in AI development
Ms. Bianchi steered the discussion to the challenges associated with sourcing the large, high-quality datasets needed to build AI, noting that HIPAA was designed so long before many of today’s innovations that there are now challenges for provides access to datasets. dr Blake echoed that concern, describing HIPAA’s “nearly chimerical” competing goals, which aims to encourage broader patient access to their data while maintaining privacy. dr Blake promoted more automatic collection of data, patient-entered data, and better access for patients to review their data and correct errors.
At the CTA, Ms. Haresign said the industry recognized the importance of properly handling healthcare data and published industry practices to address it. Noting that there are challenges for healthcare providers to trust the data used in AI algorithms, Ms. Bianchi asked the panel what problems she thinks are emerging with accountability and bias. dr Blake responded that “trust is like explainability” and explained how clear labeling can help solve this dilemma. Ms. Haresign noted that the level of trust required of an AI product is commensurate with the risk associated with that drug or device.