Atropos Health raises $33M to improve value-based care through real world evidence

Steven Loeb · May 23, 2024 · Short URL:

The company launched its Generative AI solution in October, with direct chat-to-database capability

There's more healthcare data flowing through the system now than ever before, thanks to new sources like EMRs and wearables, that are being collected daily. Yet, despite that, there's still an evidence gap, one where 70% of the US population is systematically excluded from clinical trials, and only about 14% of daily medical decisions have any high quality evidence behind them.

"If you're a physician trying to treat a patient, you look for trials and literature and guidelines that give you evidence for what to do and, very often, you're left without high quality evidence to inform that decision. And the evidence they do have often would have excluded the very patient in front of them. What doctors have to do all the time is extrapolate from that imperfect evidence," Dr. Brigham Hyde, CEO and co-founder of Atropos Health, told VatorNews.

The company, which uses AI to translate real-world clinical data into personalized, real-world evidence for care, announced a $33 million Series B round on Thursday, bringing its total raised to $49 million.

One of the challenges being solved by Atropos is the difficulty of producing high quality evidence which has, until now, been a manual process, one with difficult to implement methodologies, and one that has required teams of people. Atropos focuses on automating that conversion.

"We're using its data to create high quality evidence that is built to be very transparent, using the best methodologies, highly accurate, and let's do that with automation. And that's what we do, we cut that timeline down to usually about a day right now and with our generative solution that drops to minutes," said Hyde.

Founded in 2019 as part of a research project at Stanford University, Atropos' platform aggregates data from millions of patient charts so it can answer physician questions, and then generates a report, which it calls a Prognostogram. For example, if a pediatric cardiologist is seeing a new patient with an incidental diagnosis of atrial septal aneurysm, they might want to find out about the need for anticoagulation. The existing literature, however, might not be helpful for a variety of reasons, perhaps because a study answering the question hasn't been done yet, the available evidence is unreliable, or the evidence doesn't apply to that patient.

When a physician poses the question, which can be done through their EHR or from Atropos' online portal, the company cleans up the data in the electronic health record and then uses a novel search engine to generate patient cohorts and perform statistical analyses to provide an answer. The Atropos team works with the physician to refine the question, perform the analysis, and to provide an interpretation of what the results mean for their patient. Atropos' clinicians will also discuss and clarify any follow-up questions. 

The Atropos Evidence Network includes de-identified medical records from over 200 million patients from over 10 years from around the globe. The data comes from a list of healthcare RWD sources, including from data platforms, aggregators, academic medical centers (AMCs), large health systems (IDNs), and community oncology practices, as well as claims data, genomics, and registry data, which comes from a partnership Atropos entered into with patient-centric healthcare data company SEQSTER last last year.  

For example, Atropos recently published a study using its data to see if concerns that patients who were undergoing surgery should be taken them of their GLP-1 medication were founded or not. In its results, Atropos found the evidence shwoing that it's fine for patients to take their medication; this study has since been used to change policy for patients.

"That type of story is very typical for us: often a user, a physician or researcher, will come to us with an idea or a hypothesis they have. By being able to present them with really clear, high quality evidence, run on often tens of thousands of patients that match specific criteria, they can not only make a change in daily patient care, but they can affect policy. By doing that, it can be used to advance evidence based care everywhere," said Hyde.

Generative AI solution

In October, Atropos launched its Geneva OS and ChatRWD user application in October, the first Generative AI application incorporating direct chat-to-database capability, to help healthcare and life science leaders advance and accelerate evidence generation.

When its technology originally spun out of Stanford, it didn't use any AI or large language models (LLMs); it still had the clinician in the loop, where they would take the question from the physician or researcher and Atropos would use its technology to program the study and run it, usually in about a day.

When ChatGPT came out, though, the company started to look at ways to automate its technology, though that was not without its own risks.  

"Our system, Geneva, is known for its accuracy, and transparency and one of the risks with LLM is that they can tend to hallucinate. There are different publications out there that show a hallucination rate of at least 10%, perhaps even higher than that. For the types of evidence we produce, that would be unacceptable," Hyde explained.

"We had to make sure this was still accurate. So, what we've replaced the human intake steps: now a user by themselves can just type a few sentences and they'll be guided by the co-pilot that’s LLM-backed to basically input a question into our system and it runs without that human needing to be in the loop." 

ChatRWD was built so that the company's core technology, Geneva, is the tech that does all the data analysis, meaning it's not letting an LLM touch a patient record or do that type of analysis. Geneva's still does the primary analysis, while the LLM is helping the user form and structure and define a good question, which then that goes into Geneva and runs automatically.

In its own study, Atropos found that, in comparison to other LLMs, ChatRWD performed much higher on physician trust.

"As generative AI has deployed so rapidly, we're quickly moving into a space, for certain types of applications, we really have to trust the accuracy and physicians are going to vote by knowing how much they trust the output For our setting, that's absolutely crucial."

Going deeper into value-based care

The company's newly raises Series B funding was led by Valtruis, with participation from new strategic investors including McKesson Ventures, Touchdown VC, Sanofi Ventures, and Merck GHI Fund, along with existing investors Breyer Capital, Emerson Collective, and Presidio Ventures. 

"We've included a number of strategic investors in this round and I would highlight both McKesson and Touchstone Ventures: they represent really significant players in vertical integration within healthcare. They touch many parts of the value chain and we think that evidence should spread throughout the healthcare value chain and when it does, we'll get better outcomes for patients and better financial outcomes for players involved," said Hyde.

With this funding, Atropos Health plans to further its stake in value-based care, helping healthcare organizations accelerate appropriate clinical decision-making for patients based on their unique physiology.

This has become especially important as what Hyde calls "the low hanging fruit" of value-based care such as deploying capabilities to improve chronic care management, have been solved. Now it's more about how to deal with complicated patients, especially now the way to get paid in those contracts has become harder; by helping physicians get the right evidence for their care, Atropos can help the physician by arming them with the evidence they need to treat a patient a certain way and another patient another way to achieve the best results. 


Along with the new funding, Atropos Health also announced that value-based care veteran Mike Spadafore has joined its board of directors. 

"Mike has a stellar reputation as an investor, and it's been active in value-based care for over a decade. One of the focuses to date of the vouchers has been in value-based care and he brings a great knowledge base there. He's also worked at the Blues Index Fund, and familiar with the payer landscape and those dynamics. So I'm looking forward to him contributing a lot to our strategy there, as well as thinking through where the puck is moving on that side of the wall. He will be a great partner to build this business with going forward," said Hyde.

The ultimate vision at Atropos is to enable personalized evidence for everyone, part of which is about building great user experience that can be used easily by clinicians, researchers, and policymakers.

"We bank that on standards of quality. It's not just good enough to do a little analysis on these patients, we're talking about clinical evidence for decision making, that's the standard we hold to, and we think Geneva should be the standard for converting data into high quality evidence," Hyde said.

Support VatorNews by Donating

Read more from our "Trends and news" series

More episodes