Why identical-looking ICU patients get different outcomes
Two patients roll into the emergency department with pneumonia.
Same fever. Same oxygen level. Same age. One recovers quickly. The other spirals into sepsis and needs the ICU.
Doctors have always known these patients aren't really the same on the inside. The question has been: how do you measure the difference?
Sepsis is a leading killer worldwide. It is the body's own immune response gone haywire after an infection.
For decades, drugs that calm that response — called immunomodulators, like hydrocortisone — have been tested in trials. The results keep coming out mixed.
The likely reason: trials enroll patients based on how sick they look, not on how dysregulated their immune system actually is. That mismatch may be why the same steroid helps some patients and harms others.
The old approach vs the new
Traditional sepsis trials use clinical severity to pick patients. Blood pressure, oxygen needs, and organ scores decide who enrolls.
But here's the twist. The authors argue severity doesn't equal dysregulation. Two equally sick-looking patients may have very different immune fingerprints.
So they built a tool that measures the fingerprint directly.
How it works, in plain terms
Think of the immune system like an orchestra.
In health, the sections play in balance. In sepsis, some sections play way too loud while others fall silent. "Severity" tells you the music is chaotic. It doesn't tell you which instruments are the problem.
The new tool listens to 35 specific blood messages — cytokines, receptors, and other signals — to pick out the true shape of the chaos.
Then, brilliantly, the team simplified that reading down to just three measurements: procalcitonin, soluble TREM-1, and IL-6. Three markers, one score (parsimonious = uses as few inputs as possible).
The team studied 398 adults with community-acquired pneumonia across emergency departments, hospital wards, and ICUs.
They measured 35 plasma biomarkers and used machine learning to find patterns without assuming what the answer should look like. That produced three "immune dysregulation profile" stages (DIP1, DIP2, DIP3) and a continuous 0-to-1 score called cDIP.
They then tested whether just three blood markers could predict the full 35-marker pattern. They validated the result in five independent groups totaling 1,191 more patients.
Finally, they reanalyzed the CAPE COD trial — a randomized study of hydrocortisone in severe community-acquired pneumonia — to see if their score predicted who actually benefited.
The three-biomarker model predicted the full 35-biomarker dysregulation with about 91% accuracy for the stage classification.
Higher dysregulation was linked to higher mortality and more secondary infections. This held up independent of how sick the patient looked clinically — meaning the tool captured something severity scores missed.
Now for the part that may change how sepsis trials are run.
In the CAPE COD reanalysis, hydrocortisone helped survival only in patients classified as severely dysregulated (DIP3, or cDIP scores at the high end). Those patients also recovered their immune balance faster when given the steroid.
In less dysregulated patients, no survival benefit appeared. When the same trial was re-sliced by clinical severity alone, no such pattern emerged.
Here's where it gets interesting
This is precision medicine for the messiest field in hospital care.
Instead of treating everyone with severe pneumonia the same way, doctors could someday run three blood tests and decide who truly needs hydrocortisone.
That could mean better outcomes for the right patients — and sparing others the known risks of steroids.
Sepsis experts have argued for years that the field needs "endotypes" — biological subgroups — to move beyond one-size-fits-all care.
This paper offers one of the most practical endotyping tools so far. The fact that it was validated in five external cohorts, across different infections and care settings, is a major strength. The publicly available framework also means other research teams can test it right away.
Right now, this is a research tool. You cannot walk into your local ER and request a "DIP score."
But if you or a loved one is in the hospital with pneumonia or sepsis, the story is still hopeful. Sepsis care is moving from gut calls toward biology-guided decisions. The days of handing the same drugs to every patient may be ending.
If you're in a clinical trial for a sepsis therapy, ask whether the trial uses an immune stratification tool. Increasingly, the answer may be yes — and that is good news.
The study is largely observational and retrospective for the key treatment findings.
The hydrocortisone analysis reused an existing randomized trial rather than testing the tool prospectively. That's a strong hint, not a final answer. Some cohorts were also small in certain subgroups.
Finally, real-world labs will need to validate that procalcitonin, soluble TREM-1, and IL-6 can be measured reliably and quickly enough for bedside decisions.
The next step is a prospective trial that assigns hydrocortisone based on the DIP or cDIP score and measures outcomes directly.
If that works, sepsis care could shift toward "right drug, right patient, right time" — a long-held goal of critical care. For now, the framework is publicly available for other teams to build on.