ARPA-H commits $63M to autonomous AI for heart failure care

The Advanced Research Projects Agency for Health has committed $62.7 million to develop autonomous AI systems that can monitor heart failure patients between clinical visits, adjust medications, and escalate cases to human physicians when necessary.

The initiative, named ADVOCATE (Agentic AI-Enabled Cardiovascular Care Transformation), represents a coordinated effort to establish both the technology and regulatory framework for a new class of clinical AI. ARPA-H selected three companies—Atman Health, Tempus AI, and UpDoc—to build patient-facing agents, with awards ranging from $7.7 million to $9.5 million each.

Why it matters

This marks the first major federal effort to create regulatory pathways for generative AI in high-risk clinical settings. Until now, the FDA has only approved predictive AI technologies. The program directly addresses a workforce crisis: nearly half of U.S. counties lack a cardiologist, while cardiovascular disease costs the nation nearly $500 billion annually and kills more than 200,000 Americans each year from preventable causes.

Building the regulatory framework

A defining feature of ADVOCATE is its coordination with the FDA from the outset. The teams developing patient-facing agents are expected to submit FDA authorization packages within two years. Stanford University will receive up to $15 million to build a supervisory AI system that monitors the clinical agents for unsafe recommendations.

"What we are trying to achieve with ADVOCATE is a technology that can serve as a clinician-extender: an autonomous agent smart enough to understand a patient's treatment needs, capable of providing certain care autonomously, and equally capable of engaging the clinical team when needed," said Haider Warraich, the program's manager and a practicing cardiologist.

ARPA-H chief medical officer Rafid Fadul emphasized the agency's focus on real-world deployment rather than research that remains theoretical. The program includes a 39-month pathway from prototype to clinical use, incorporating independent validation studies, performance benchmarking against cardiologists, and FDA-supervised clinical trials.

Real-world testing at scale

Duke University will lead validation testing across five health systems using both Epic and Oracle EHR systems, with an award of up to $15.5 million. Kaiser Permanente received up to $16.3 million to design implementation strategies and conduct a randomized trial enrolling approximately 2,500 patients across 21 medical centers.

The Kaiser Permanente study will proceed in phases: first analyzing existing EHR data to identify and test agentic AI solutions, then integrating successful systems into care workflows, and finally launching pilot studies followed by randomized trials.

Johns Hopkins University Applied Physics Laboratory will serve as an independent evaluator, validating the program's performance and safety claims.

The technology approaches

Atman Health will build its system on an evidence-based clinical decision engine combined with large language models, featuring a voice-first interface that adapts diagnostic questions based on patient responses. Tempus AI plans to extend its existing Olivia patient health app with continuous monitoring capabilities. UpDoc, whose CEO described the program as the "Manhattan Project of clinical AI," will create an agentic system operating within its proprietary Clinical Intelligence System designed to maintain physician oversight.

Sharif Vakili, UpDoc's CEO and a practicing primary care physician at Stanford Medicine, noted that heart failure represents "the moonshot indication for clinical AI" due to its complexity and cost burden.

ARPA-H projects the program could generate $28 billion in annual cost savings across the heart failure population if successful, while reducing preventable hospitalizations.

These details were first reported by Fierce Healthcare.