ARPA-H Commits $62.7M to Build World's First FDA-Authorized Cardiovascular AI Agent
Background: A Structural Crisis in Cardiac Care
The United States faces a profound structural challenge in cardiovascular care. According to the American Heart Association (AHA), nearly half of all U.S. counties lack a single practicing cardiologist, while heart failure alone generates over $30 billion in annual healthcare costs. Against this backdrop, the Advanced Research Projects Agency for Health (ARPA-H) announced in September 2026 the launch of the ADVOCATE program — a four-year, $62.7 million initiative to develop the world's first FDA-authorized autonomous clinical AI agent system specifically designed for cardiovascular disease management.
ARPA-H committed up to $33.7 million for the program's first year alone, signaling the agency's high-priority commitment to this initiative. ADVOCATE — short for "Agentic AI-Enabled Cardiovascular Care Transformation" — aims to deploy AI agents as digital members of clinical care teams, providing 24/7 specialty-level cardiovascular care for heart failure and post-heart attack patients.
Three Technical Areas: From Patient Interaction to Scalable Deployment
The ADVOCATE program is structured around three interconnected Technical Areas (TAs), each led by different leading institutions:
TA1: Patient-Facing Clinical AI Agent
This area focuses on developing autonomous systems that interact directly with patients and escalate care to human medical teams when necessary.
- Atman Health: Developing a voice-first, LLM-based system that dynamically adjusts diagnostic questioning in real-time based on patient responses, enabling personalized symptom assessment.
- Tempus AI: Extending its "Olivia" patient health app to include continuous monitoring and deep clinical analysis, allowing patients to receive professional-grade cardiac health tracking at home.
- Updoc: Using a rules-based system to validate conversational AI actions against clinical protocols, ensuring AI recommendations meet medical standards.
TA2: Supervisory AI Agent
Safety is a core concern of the ADVOCATE program. Stanford University is developing a disease-agnostic, three-stage supervisory system:
- Outlier Filtering: Identifying and flagging unusual AI behavior patterns
- Rule-Based Screening: Validating AI decisions against established clinical protocols
- Deep-Research Auditing: Providing detailed rationales for clinical claims
This supervisory layer ensures that every significant decision made by TA1 agents undergoes independent review, substantially reducing the risk of AI errors causing patient harm.
TA3: Scalable Implementation
- Duke University: Leading a multi-site validation platform across five health systems, integrating with Epic and Cerner/Oracle EHR systems in collaboration with the American Heart Association.
- Kaiser Permanente: Executing enterprise-scale deployment across 21 medical centers and over 260 clinics, utilizing shadow-mode deployments and pragmatic clinical trials to ensure safety and effectiveness in real-world environments.
Clinical Capabilities: Beyond Traditional Telemedicine
The AI agent systems in the ADVOCATE program are designed to perform a range of tasks traditionally requiring cardiologists:
- Symptom Assessment: Real-time evaluation of cardiac symptoms through voice or text dialogue
- Appointment Scheduling: Automatically arranging appropriate medical appointments based on symptom severity
- Diet and Physical Therapy Recommendations: Providing personalized lifestyle intervention guidance
- Prescription Modification: Recommending or adjusting medication prescriptions within a supervisory framework
- Laboratory Test Ordering: Automatically ordering relevant blood or imaging tests based on clinical needs
These capabilities enable AI agents to provide continuous specialty care in areas with cardiologist shortages, filling critical gaps in the existing healthcare system.
Regulatory Framework: An Historic FDA Challenge
One of ADVOCATE's greatest challenges is the absence of an existing regulatory framework. Current FDA regulations lack specific provisions for patient-facing clinical agentic AI systems capable of performing physician-level functions.
To address this, ADVOCATE is working closely with the FDA to establish an entirely new regulatory framework. TA1 performers are required to submit an FDA-authorization package within 24 months of their contract award, meaning the regulatory pathway must be developed in parallel with implementation.
The Johns Hopkins University Applied Physics Laboratory (JHU/APL) serves as the independent external evaluation partner, assessing technical performance and clinical outcomes to provide objective data for FDA review.
Expected Impact: $28 Billion in Annual Savings Potential
ARPA-H estimates that the ADVOCATE program could generate up to $28 billion in annual cost savings within the heart failure patient population alone, primarily through:
- Reducing Preventable Hospitalizations: Continuous monitoring and early intervention to lower emergency admission rates due to condition deterioration
- Improving Medication Adherence: AI agents' continuous reminders and education helping patients better follow medication regimens
- Optimizing Resource Allocation: Concentrating limited cardiologist resources on complex cases requiring human intervention
Implications for Asia-Pacific: Shared Challenges in Cardiac Care
The ADVOCATE program holds significant relevance for the Asia-Pacific region. According to the World Health Organization, Asia-Pacific bears one of the world's heaviest cardiovascular disease burdens, with China, India, Japan, and Southeast Asian nations all facing cardiologist shortages and aging population pressures.
Healthcare institutions in Hong Kong, Singapore, and Australia have already begun exploring similar AI-assisted cardiac care models. ADVOCATE's success — particularly the establishment of its FDA authorization framework — will provide valuable reference for Asia-Pacific regulators, accelerating the compliance pathway for AI medical agents in the region.
Conclusion: A Milestone for Agentic AI in Healthcare
ARPA-H's ADVOCATE program represents an important milestone in AI's application to healthcare. It is not merely a funding initiative but a systematic attempt to establish a complete framework for agentic AI to operate safely and effectively in high-stakes clinical environments — from technical development to regulatory authorization, from single-site trials to enterprise-scale deployment.
As the program progresses, ADVOCATE's outcomes will provide critical empirical evidence for global healthcare AI development, driving agentic AI from laboratory settings into genuine clinical practice.
