ARPA-H launches $63M effort to build FDA-authorized AI agents for heart failure care

The Advanced Research Projects Agency for Health (ARPA-H) launched a $62.7 million effort to develop a patient-facing agentic clinical AI system for cardiovascular care, specifically to treat heart failure.

The program, called Agentic AI-Enabled Cardiovascular Care Transformation (ADVOCATE), aims to create an autonomous digital member of the care team that can continuously monitor heart failure patients, provide certain aspects of care between visits and escalate issues to clinicians when needed. The selected teams will build patient-facing AI, a supervisory agentic AI and a scalable deployment plan to produce a clinical AI system for cardiovascular care.

A key aspect of the initiative is close collaboration with the FDA to establish a regulatory framework for a new class of autonomous clinical AI. The teams working on patient-facing agents are expected to submit an FDA authorization package within two years, ARPA-H said in a press release.

Teams participating in the program will work with ARPA-H to establish shared evaluation standards, interoperability requirements and reimbursement pathways. 

Johns Hopkins University Applied Physics Laboratory (JHU/APL) was tapped as an external evaluation partner. JHU/APL will independently validate the program's performance and safety.

When announcing the initiative in January, ARPA-H said it wanted to set novel regulatory precedent with the FDA for the use of generative AI in high-risk settings. To date, only predictive AI technologies have been approved by the agency. President Donald Trump’s ARPA-H is taking a big swing at creating and deploying generative AI to extend the healthcare workforce, former staff writer Emma Beavins reported at the time.

ARPA-H’s commitment is up to $33.7 million in the first year of the four-year, $62.7 million program. "If successful, ADVOCATE will save lives, reduce preventable hospitalizations, and generate an estimated $28 billion in annual cost savings across the heart failure population alone," agency officials said in the press release.

“Every day, Americans are dying from cardiovascular disease that we know how to prevent and treat—and too often the barrier isn’t the medicine; it’s access to a clinician who can guide the care,” Haider Warraich, M.D., ADVOCATE Program Manager, and a practicing cardiologist, said in a statement. “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. This work will change how we practice medicine—and it would not happen without ARPA-H driving the vision for the future.” 

The program aims to address major gaps in cardiovascular care, particularly in rural and underserved areas where nearly half of U.S. counties lack a cardiologist. 

The U.S. spends nearly half a trillion dollars on heart disease annually and still experiences far worse cardiovascular outcomes than peer nations. More than 200,000 Americans die each year from preventable impacts of cardiovascular disease (CVD), which is the leading cause of death in the country. 

ARPA-H funds high-risk, high-reward health research to advance innovation that will reach the patients' bedside, said Rafid Fadul, M.D., chief medical officer at ARPA-H, during a Consumer Technology Association event this week.

"If we come up with some cool science that just sits on the shelf at the end of that three to five years, we regard that as a failure. That's a science fair project. We're not interested in science fair projects," Fadul said.

By coordinating with the FDA and the Centers for Medicare and Medicaid Services, the ADVOCATE program is building the regulatory and reimbursement pathways for clinical AI, he noted.

The program offers a 39-month pathway that helps technologies move from prototype to real-world clinical use. That pathway includes independent verification and validation studies, benchmarking agent performance against cardiologists and clinical trials conducted under FDA Investigational Device Exemption. It also includes large-scale scalability studies deployed within partner health systems, with electronic health record integration, comparing agent-enabled care to usual care. 

ARPA-H selected three health tech companies—Atman Health, Tempus AI and UpDoc—to develop patient-facing clinical AI agents. The AI agents will serve as virtual care team members, monitoring patients between appointments, making medication adjustments and escalating cases to human clinicians when necessary, ARPA-H said.

Atman Health received an award of up to $7.7 million to create a clinical AI agentic system built on an evidence-based clinical decision engine. The system will also feature large language models (LLMs) with a voice-first interface that patients talk to and the system can adjust on the fly to ask each patient the diagnostic questions most likely to be useful, according to the company.

Tempus AI was awarded up to $9.5 million to build clinical AI agents that extend the company's existing Olivia patient health app, adding continuous monitoring with deeper clinical analysis when it detects meaningful changes in a patient’s health.

Clinical AI company UpDoc will receive up to $9.2 million as a participant in the ADVOCATE program. The company aims to create an agentic AI system that can continuously analyze patient data, reason through clinical decisions and recommend diagnostic, treatment and care management actions. Those recommendations would operate within UpDoc's proprietary Clinical Intelligence System, which is designed to maintain physician oversight and enforce clinical safeguards. The goal is to combine the reasoning capabilities of large language models with the safety, governance and auditability required for real-world healthcare use, according to the company in a press release.

Sharif Vakili, M.D., CEO and co-founder of UpDoc, described the ADVOCATE program as "ambitious and historic," noting Warraich's comments that it represents the "Manhattan Project of clinical AI."

"It's one of those things that the full gravity and significance of it may not be entirely appreciated until hindsight," Vakili, a practicing primary care physician at Stanford Medicine, told Fierce Healthcare. "To do clinical AI right, you need to coordinate with the entire ecosystem," he noted, which entails electronic health record integration, clear regulatory pathways, payment reform, physician engagement, health system governance and responsible data-sharing frameworks. 

UpDoc's current FDA-cleared insulin management system combines medical device functionality with conversational AI to support care within physician-prescribed treatment plans. The ADVOCATE program marks the company's move into heart failure care.

"Heart failure is the moonshot indication for clinical AI," Vakili said, noting that it is one of healthcare's most costly, complex and resource-intensive conditions.

"Our platform already supports continuous, personalized care under physician oversight at some of the nation's premier health systems. This program gives us the opportunity, and the rigorous validation pathway, to prove that autonomous clinical AI can be trusted with the patients who need it most," Vakili said.

ARPA-H selected a team at Stanford University to build a supervisory AI system designed to monitor the clinical agents for unsafe recommendations and other risks. The Stanford team will receive up to $15 million.

Duke University and Kaiser Permanente will lead real-world implementation and testing across multiple health systems.

Duke University, slated to receive up to $15.5 million, will lead a consortium providing a multi-site validation platform testing ADVOCATE agents across five health systems and rural sites on both Epic and Oracle (Cerner) EHR systems.

Researchers at Kaiser Permanente were awarded up to $16.3 million to design and execute a real-world implementation strategy. The first phase of the study will focus on identifying, testing and evaluating agentic AI solutions using existing EHR data from Kaiser Permanente Northern California, according to the health system in a press release. In the second phase, researchers will integrate agentic AI solutions found to accurately suggest possible treatment options for patients with heart failure into KPNC care workflows.

Kaiser Permanente researchers will then launch a pilot study, followed by a randomized trial, that will enroll about 2,500 patients across 21 medical centers to test the incremental value of agentic AI-supported care.

The investigators will also develop a national playbook on how to rigorously implement and safely evaluate AI in healthcare settings.

"This research will allow us to investigate whether agentic AI solutions, which can be trained to evaluate a situation and make complex decisions, can close care gaps for patients with heart failure in ways that are supported by the best available evidence," said Ankeet Bhatt, M.D., a research scientist at the Kaiser Permanente Division of Research and a cardiologist with The Permanente Medical Group, who is leading the project at the health system.