A tool to diagnose sepsis

In the United States, at least 1.7 million adults develop sepsis each year and at least 350,000 die as a result, according to the Centers for Disease Control and Prevention. The condition is a life-threatening emergency in which the body’s extreme response to an infection damages its own tissues and organs.

Purvesh Khatri, PhD, a professor of computational medicine, set his sights on developing the first rapid molecular sepsis test and in 2016 co-founded Inflammatix Inc., which licensed his Stanford Medicine research to commercialize it. TriVerity was approved by the U.S. Food and Drug Administration in 2025.

The assay helps doctors make better decisions in emergency situations, when it’s not clear whether a patient has sepsis. Other tests that are in use take hours to days to produce definitive answers, so clinicians must make educated guesses based largely on symptoms. The new test can use one blood sample to quickly tell whether the patient has an infection, what kind it is and how serious it’s likely to be — all within 30 minutes.

In this Q&A, we asked Khatri, a professor of computational medicine, to fill us in on the test and its trajectory of success.

How does the test work?

It assesses how the patient’s immune system is responding — measuring a panel of 29 mRNA molecules from a small blood sample — and applies machine learning to interpret those signals. Instead of hunting for a single pathogen, it reads the body’s response and translates it into actionable insight.

What sparked the idea?

Tim Sweeney, MD, PhD, a postdoc in my lab and a co-founder of Inflammatix, was a resident in general surgery at Stanford Medicine from 2011 to 2015. The need became apparent during his time treating patients in the emergency department, surgery ward and intensive care unit.

The computational framework for how to get it done — combining multiple cohorts to find repeatable signals in the immune system — came from the method I had developed for an analysis of heterogeneous data.

The idea that immune response can diagnose infection without looking for a pathogen came from an analysis carried out by Charles Liu, who was a high school summer intern in my lab in 2013. His analysis of more than 8,500 blood samples from more than 40 diseases showed that a patient’s immune response differed by the type of immune insult, such as infection, organ transplant, autoimmune disease or cancer.

What key scientific or engineering advance made this possible?

Two things came together to turn noisy biology into clear guidance: the ability to measure many immune signals at once from a tiny blood sample and a computational framework that can spot patterns humans can’t in highly heterogeneous datasets.

The test platform on which TriVerity runs is called Myrna (think: My RNA) and it took about nine years of development. The test can measure up to 64 genes simultaneously from a single blood draw, making it the highest multiplexed platform to date with FDA clearance.

The computational framework leverages machine learning and AI to flip the traditional biomedical research paradigm. Traditionally, those of us looking to create diagnostic tests believed a test would be a failure if it was built on data from a variety of sources — with biological differences between people, clinical differences due to their medical conditions and technical differences depending on how the data is collected, for example. Our framework turns this curse into a blessing in disguise to identify a signal that persists despite heterogeneity.

What was the toughest challenge and how did you overcome it?

Disbelief. Clinicians have always dreamed of being able to read out immune signals, but solving the technical challenges and aligning commercial incentives were thought to be impossible. I can’t tell you the number of times we were told we would not be able to find a stable signal in the highly variable immune response and would not be able to build a sample-to-answer system with the 30-minute turnaround time. In addition, hundreds of millions of dollars have been spent on sepsis tests (usually tests looking for bacteria in the blood), and they have all been commercial failures. In fact, someone once told me, “Sepsis is the graveyard of startups.”

How did you show it works and is safe?

We had a successful trial designed to provide the evidence needed for FDA approval of the test’s accuracy, sensitivity and specificity. The trial [published in September 2025 in Nature Medicine] enrolled 1,222 patients from 22 emergency departments in the U.S. and Europe. Most importantly, we were able to show that the test we developed could have substantially reduced the number of missed infections and patients who were judged to be clinically stable but were, in fact, critically ill.

What early outcomes or adoption are you seeing?

Several health care systems have implemented the test, and the results are extremely encouraging. In several implementations, we are seeing faster decisions, more confidence in care plans and substantially fewer unnecessary admissions.

Can you share a patient story that captures its real-world impact?

Here is one of many: A man in his 60s arrived at the ED with a fever and shortness of breath. The team was working him up for pneumonia when the test score came back as low bacterial, low viral and high illness severity. Surprised that this was noninfectious, they ordered a heart attack workup — and it turned out he was having a heart attack. The patient had a good outcome. Thank goodness they didn’t treat a heart attack with antibiotics!

What’s next?

First, we are improving the accuracy of the test and demonstrating its utility in improving patient care. Second, we are working to demonstrate that the test can also help identify which intensive care unit patients with sepsis would respond well to immune modulatory therapy — and which would have negative consequences.

How does it feel to see this test FDA-approved and available to patients?

I have several overwhelming emotions. As someone who grew up in India and trained as an electronics engineer, I never imagined having opportunities to work on a disease that, according to the World Health Organization, accounts for 20% mortality globally each year. So, this success has given me a sense of fulfillment and accomplishment — to be able to go from an idea, which many people told us could not be done, to an FDA-cleared test with a potential for global impact. More importantly, it is humbling to realize how lucky I have been. I am eternally grateful for the singular privilege of mentoring and working with so many talented individuals both in academia and industry who committed to making that idea a reality.