The U.S. Department of Health and Human Services announced new funding Thursday to improve the use and maintenance of artificial intelligence medical devices.
Research has shown that machine learning models used in clinical settings can deteriorate over time. The new HHS money is earmarked for the Advanced Research Projects Agency for Health, or ARPA-H, which will use it to innovate its efforts to make AI tools more reliable for doctors and more useful for patients.
That initiative is known as the Performance and Reliability Evaluation for Continuous Modifications and Useability of Artificial Intelligence, or PRECISE-AI.
WHY IT IS IMPORTANT
According to ARPA-H, more than 950 medical devices with integrated AI capabilities have been approved by the U.S. Food and Drug Administration, a tenfold increase from 2018.
While these new AI tools can transform physicians’ ability to deliver care, ML can deteriorate over time due to changes in clinical activity inputs, data collection, patient populations, or IT infrastructure.
“The promise of AI-driven healthcare tools is only as great as the relevant real-world data they are based on,” Berkman Sahiner, program manager of PRECISE-AI, said in a statement.
Current technologies challenge users to monitor and maintain the performance of AI-driven devices in real clinical environments.
“The inability to automatically monitor and maintain the performance of a medical device with AI based on real-world actions poses risks to physicians and patients,” Sahiner said.
Through PRECISE-AI, HHS aims to establish an open-source repository of tools that provide AI-driven clinical decision support by automatically identifying and correcting performance degradation without the need for human oversight.
Root cause analysis tools should also provide clear and actionable information about the sources of degradation, so that healthcare users can better interpret the uncertainty in the model.
By “recommending the optimal approach to detect and mitigate underlying performance degradation of AI models,” clinicians can provide better patient care, Sahiner said.
The program’s upcoming final procurement will focus on five technical areas. Teams will begin by developing tools to make the most accurate estimate of a patient’s diagnosis, given the available evidence, or “ground truth,” the agency said.
The teams then develop autonomous ways to monitor an AI tool’s performance against these ground truths, determine the root causes of any degradation, and make necessary corrections.
They also create mechanisms and the underlying data infrastructure that alert clinicians, AI tool developers, hospital administrators, and regulators when performance degradation occurs.
The response date for the request is currently January 15, 2025.
THE BIGGER TREND
As an independent entity of the National Institutes of Health, ARPA-H strives to invest in ways to build stronger, healthier, and more resilient health systems.
The focus is on creating scalable platforms and strategies, from advanced mobile hospitals delivering acute care in rural areas to securing open source software used in critical infrastructure.
As part of the Digital Health Security Initiative, DIGIHEALS, ARPA-H collaborated with the Defense Advanced Research Projects Agency on the Artificial Intelligence Cyber Challenge. In May, HHS also announced that $50 million would be made available under ARPA-H to help providers patch ransomware vulnerabilities in networks and medical devices.
ON THE RECORD
“PRECISE-AI addresses a growing gap in ensuring that AI tools used in clinical decision-making are accurate, safe, and robust in real-world settings,” ARPA-H Executive Director Renee Wegrzyn said in the procurement announcement. “In doing so, ARPA-H is creating a foundation of trust between clinicians and these AI tools, further expanding the potential of AI to improve health outcomes for all Americans.”
Andrea Fox is Editor-in-Chief of Healthcare IT News.
Email address: [email protected]
Healthcare IT News is a publication of HIMSS Media.
The HIMSS AI in Healthcare Forum will take place September 5-6 in Boston. More information and registration.
