Dive Brief:
- The Advanced Research Projects Agency for Health, or ARPA-H, announced the first contracts Wednesday for a project to build an agentic artificial intelligence system to support heart disease treatment.
- Called Agentic AI-EnableD CardioVascular CAre TransfOrmation (ADVOCATE), the program will run for four years, with ARPA-H committing a total of $62.7 million. In the first year, the agency plans to fund up to $33.7 million. Companies including Tempus AI and Updoc are tasked with building an agentic system that can autonomously support patients with heart failure in between visits.
- The program will work closely with the Food and Drug Administration to develop the regulatory framework for a new type of patient-facing clinical AI. The program was announced just weeks after the FDA called for feedback on approaches to regulating generative AI in medical devices.
Dive Insight:
ARPA-H’s goal is to create a clinical agentic AI system that serves as a digital member of the care team. Agentic AI describes a technology that can perceive, act and make decisions.
The hope is to develop an agent that can provide certain care autonomously and engage a patient’s clinical team when needed, Haider Warriach, program manager of ADVOCATE and a cardiologist, said in a statement.
The program aims to reduce the impacts of cardiovascular disease and support people who lack access to specialist care. Nearly half of U.S. counties have no cardiologist, according to ARPA-H, which is an agency within the Department of Health and Human Services.
The program is divided into three teams. The first team, which includes Tempus AI, Updoc and Atman Health, will work to make a patient-facing agentic system that can support patients with heart failure between visits and escalate to their care team. This group must submit a package to the FDA for authorization within two years.
The second team, led by Stanford University, will be responsible for building a supervisory AI system that monitors the patient-facing agents after deployment. The system will be disease agnostic and should detect unsafe recommendations and out of distribution behavior, which refers to how a model acts when it receives data that is significantly different than what it was trained on.
The third group, which includes Duke University and Kaiser Permanente, will work on deploying the agents in the real world, with a focus on rural and underserved communities. Duke will test the technology across five health systems and rural sites on both Epic and Cerner/Oracle electronic health records, and Kaiser will deploy it across 21 medical centers and more than 260 clinics, embedding ADVOCATE agents into Epic software workflows for heart failure patients.
The selected companies will also work with ARPA-H to establish shared evaluation standards, interoperability requirements and reimbursement pathways. Johns Hopkins University’s Applied Physics Laboratory will provide independent oversight of the program, assessing its technical performance and clinical outcomes.
The program comes as the FDA is taking a closer look at how to regulate new AI technologies. Recently, the agency sought feedback on regulating medical devices with a generative AI component.
Rick Abramson, director of the FDA’s Digital Health Center of Excellence, said in ARPA-H’s announcement that these kinds of efforts “will be critical as we think about continuous ongoing monitoring of autonomous clinical AI for both safety and effectiveness.”
