The Advanced Research Projects Agency for Health (ARPA-H) announced the first contracts for a program aimed at building an agentic artificial intelligence system to support the treatment of heart disease, with total funding of up to $62.7 million over four years. The agency intends to provide up to $33.7 million during the first year.
The program is called Agentic AI-EnableD CardioVascular CAre TransfOrmation (ADVOCATE) and focuses in its initial phase on assisting heart failure patients between medical visits. Companies and institutions participating in the system's implementation include Tempus AI, Updoc, and Atman Health, along with Stanford University, Duke University, and Kaiser Permanente.
Three Tracks to Build and Test the System
ADVOCATE divides the work among three interconnected teams. The first team will build a patient-facing system capable of providing continuous support to heart failure patients and escalating cases to the medical team when necessary. This team must submit a package to the U.S. Food and Drug Administration (FDA) to obtain authorization within two years.
The second team, led by Stanford University, will develop an oversight system that monitors patient-facing systems after they are deployed. This system will be disease-agnostic, with the task of detecting unsafe recommendations and behaviors outside the scope of the data on which the model was trained.
The third team, which includes Duke University and Kaiser Permanente, will deploy the systems in real-world care settings, with a particular focus on rural and underserved communities. Duke will test the technology across five health systems and rural sites using Epic and Cerner/Oracle electronic records, while Kaiser will deploy the systems at 21 medical centers and more than 260 clinics by integrating ADVOCATE agents into Epic's workflow for heart failure patients.
Why Does This Program Matter?
The project is not limited to creating a medical chatbot; its stated goal is to develop an agent capable of perceiving a patient's condition and taking certain actions independently, then involving the clinical team when the situation requires it. ARPA-H believes this approach could help reduce the impact of heart disease and support patients who do not have easy access to care from cardiologists. The agency says that roughly half of U.S. counties do not have a cardiologist.
However, the program's practical value will depend on how the safety of this autonomy is demonstrated, not solely on the model's capabilities. Accordingly, participants will work with ARPA-H to establish shared evaluation standards, interoperability requirements, and financial reimbursement pathways. The applied physics laboratories at Johns Hopkins University will also provide independent oversight to assess technical performance and clinical outcomes.
Regulation and Open Questions
ADVOCATE comes as the FDA's interest grows in how to regulate new artificial intelligence technologies, including medical devices that incorporate generative artificial intelligence components. The program will work with the agency to develop a regulatory framework for a new type of clinical artificial intelligence directed specifically at patients. The announcement does not yet establish that the system has received regulatory approval or is ready for widespread use; instead, it sets a two-year deadline for submitting the authorization package and makes continuous monitoring of safety and effectiveness a fundamental part of the project.