Photo: Hero Images/Getty Images
Atropos Health's Alexandria Real World Evidence library, which now contains 33 million pieces of precision evidence-based findings, will soon be available to about one-third of U.S. physicians and about half of health systems through a growing network of clinical workflow partners, the company said this week.
New "First Edition Partners," including Meta, Heidi Health, Vim, Avo, Autonomize AI, Vye Health and Humata – and previous ones such as Microsoft and Arcadia – will integrate Atropos' RWE library to bolster clinical agent responses in healthcare workflows. The announcement follows the company's implementation of a multi-layered evidence review process.
WHY IT MATTERS
The integration of Alexandria's medical evidence into more ambient, prior authorization, value-based care and agentic healthcare operations platforms brings novel precision evidence to a significant number of existing healthcare workflows, Atropos said in an announcement on Tuesday.
Alexandria, which Atropos said is expected to grow to two billion pieces of precision evidence-based findings by the end of the year, can support clinical decision-making Q&A access from within existing workflows and supports partners' large language model (LLM) training.
"This initial release led to a two-to three-times increase over other LLMs in our published 'Answers with Evidence' eval across 5000 real physician questions," said Atropos CEO and cofounder Dr. Brigham Hyde, in a social media post after the announcement.
An evaluation of Alexandria's precision RWE summaries improved responses by more than 100 percentage points across a benchmark of 5,000 clinician questions and outperformed GPT 5.4, Claude 4.6 Opus, Llama 4 Maverick and Gemini 3.1 Pro Preview, the company said.
The ultimate goal is to make all medical evidence available to all clinicians within their workflows, Hyde said.
At this time, Microsoft, Arcadia, Meta, Heidi Health, Humata, Vim, Avo, Autonomize AI and Vye Health are integrating Alexandria RWE into their healthcare tools and platforms, Atropos said.
It's a significant step in scaling real-world health data, according to Dr. Freddy Abnousi, Meta's health technology vice president.
"Making this evidence accessible within existing workflows will empower AI systems to better understand what patients are experiencing in their daily lives and drive more meaningful personalized care," he said in a statement.
"Two problems have defined the limits of evidence-based medicine: creating evidence at the scale and specificity that real clinical questions demand, and delivering it inside the encounter where the decision is made," added Oron Afek, CEO and cofounder of Vim, a physician workflow platform vendor. "Atropos Health cracked the first."
"By integrating Atropos Health's real-world evidence, we can fill clinical data gaps that traditionally cause delays or denials," said prior auth company Humata CEO Jeremy Friese.
"We're helping our customers turn their workflows into living systems of evidence-driven intelligence, where decisions are continuously grounded in real-world data and held to the same rigor as published research," added Ganesh Padmanabhan, CEO of Autonomize AI, an agentic platform for healthcare.
This past week, Atropos said it implemented a four-step evidence review process to ensure the quality and accuracy of its evidence-driven data at scale.
The company collaborated with doctors and industry experts to ensure clinical review of AI-generated insights that align with professional medical research standards, and with Becaris, the publisher of The Journal of Comparative Effectiveness Research, to assemble an external expert evaluation service for peer-reviewed content.
As a result, each AI-generated summary's "Answered with Evidence" evaluation addresses the quality of the evidence used to answer the specific clinical question, the company said.
Any novel evidence introduced also generates an AI Peer Review report with findings from the LLMs analysis of the research methods and design and a comparison of the findings against existing literature, according to Atropos. Further, Atropos AI, called Geneva OS, ranks all content within the Alexandria library as a "pass," "pass with notes" or "pass with limitations."
Users can request a manual clinical expert to evaluate the evidence presented in an answer.
"The rigorous vetting process we have introduced will verify the accuracy of novel evidence at scale and allow clinicians to use it to make informed medical decisions and produce better outcomes," said Neil Sanghavi, Atropos' president and head of product, in a statement on April 14.
THE LARGER TREND
While Atropos is currently integrating medical evidence that hasn't been published yet into giant healthcare tech developers such as Meta, Microsoft and others, the company has offered these insights before.
Previously, care teams using the HIPAA-secure Microsoft Teams platform had access to Atropos' AI Evidence Agent to improve access to real-time, patient-specific insights during meetings.
All insights generated by the company's agent offer citations, Hyde told Healthcare IT News last year.
Clinicians can read gold-standard, peer-reviewed studies right in the application. But when this kind of medical evidence is unavailable to answer their question, insights based on other medical evidence are marked with a badge for transparency, along with the sources.
"We view that eval step as a really critical part of it," he said.
Of note, the clinical AI company Abridge, another Atropos partner that previously integrated peer-reviewed medical evidence through Wolters Kluwer, entered into new content partnerships with the New England Journal of Medicine and the JAMA Network to add more, the company said last week.
ON THE RECORD
"The promise of precision medicine and the dream of the learning health system is available to physicians and patients, in existing workflows," said Hyde in a statement.
Andrea Fox is senior editor of Healthcare IT News.
Email: afox@himss.org
Healthcare IT News is a HIMSS Media publication.
