Dive Brief:
- The HHS provided an update Thursday on its plans to help speed the adoption of artificial intelligence in clinical care, part of a bid to improve Americans’ health and bring down healthcare costs.
- During a webinar, department leaders laid out takeaways from a request for information released late last year asking the sector how the HHS could use its regulatory, research and reimbursement powers to accelerate the use of AI in healthcare.
- The industry wants coordination across HHS agencies, support in AI implementation and governance, and assistance in determining what makes a good AI tool, Arman Sharma, the HHS’ deputy chief AI officer, said. “We believe that starting with these three things and acting on constant engagement from this community is what’s needed to establish trust,” he said. “And trust in this technology is the only thing that will lead to responsible, but also effective, adoption.”
Dive Insight:
AI is a promising technology for the healthcare sector, amid hopes it could handle a range of tasks like documenting care, analyzing health data, helping providers find clinical research, and drafting messages and other care guidance for patients.
The technology is critical to expand the healthcare sector’s capacity to care for an aging population with growing health needs, HHS leaders said during the presentation Thursday. Adoption could also tamp down administrative costs and give patients tools to help manage their health, potentially decreasing the nation’s climbing health spending.
But it isn’t always easy for healthcare organizations to roll out AI products. There are plenty of risks to deploying AI, including inaccurate outputs, biased data, model degradation over time, and privacy and security concerns, experts say.
Despite calls for stronger federal oversight, the Trump administration has largely pursued a deregulatory approach to AI, concerned about slowing development of the technology.
Still, the HHS says it has done a lot of work to determine what the sector needs from regulators around AI. To that end, the department rolled out the RFI in December seeking more input from the industry. And now, the HHS is detailing takeaways from the hundreds of comments it received from providers, researchers and professional groups.
Healthcare organizations want practical guidance from the HHS on implementing and creating governance structures for AI, Sharma said.
They’re also looking for evaluation and benchmarking tools that could help them determine which AI products work well. Finally, the sector wants the HHS to coordinate its AI strategy across agencies, he said.
“Too often in government, the right hand doesn’t talk to the left hand,” Sharma said. “And the opportunity for healthcare AI that we have is simply too important for us to get lost in ourselves, so to speak.”
HHS agency leaders also discussed what they’re doing to speed AI adoption. For example, the Administration for Community Living launched a competition earlier this year for technology developers creating AI tools that could help caregivers assisting older Americans or people with disabilities.
The Advanced Research Projects Agency for Health is also working on a project to develop AI agents that can autonomously manage cardiovascular disease care.
The Food and Drug Administration is actively working on policy development, so the agency can’t provide details on its approach, said Dr. Rick Abramson, the director of the FDA’s Digital Health Center of Excellence.
But the FDA is focusing on several major themes when it comes to AI oversight, including providing greater clarity on what the agency regulates and what it requires of developers, as well as developing regulation that is proportionate to the AI tool’s risk, he said.
The agency is also considering how to regulate across an AI product’s lifecycle — like balancing the need for oversight before and after a tool goes to market — and coordinating its policies with other government agencies, professional groups and international regulators.
“It’s been said that technology evolves on a scale of weeks to months, while regulation evolves on a scale of months to years,” Abramson said. “There is some truth to that, and the world is looking to the FDA for leadership in how to approach advanced clinical AI tools.”
