Journal Logo



A Dream of Nuanced Sepsis Care

Walker, Graham MD

doi: 10.1097/01.EEM.0000669352.71242.f6

    If you remember last month, the figurative bee in my figurative bonnet was sepsis, and I was excited to learn that there might be sepsis phenotypes, clusters of patients who all have “sepsis” but tend to have different body systems affected (neurologic v. cardiopulmonary v. renal) and also have widely different outcomes. (EMN. 2020;42[5]:15;

    These phenotypes were suggested by Christopher Seymour, MD, MSc, an associate professor of critical care and emergency medicine at the University of Pittsburgh School of Medicine, who published this article after struggling to figure out how to deliver better sepsis care. (JAMA. 2019;321[20]:2003;

    I emailed the very famous and very busy Dr. Seymour, and was lucky enough to have a fascinating discussion with him about his research, though my excitement was somewhat dampened that we may not see these phenotypes in practice any time soon.

    He is clearly an academic's academic, and is the first to point out that his phenotypes may not be exactly the right ones and that there are limitations to the machine learning analysis he did. Is it four phenotypes or five? Will we be referring to an alpha sepsis patient in the future? “Probably not. These phenotypes won't stand the test of time,” Dr. Seymour said, adding that hopefully we'll be talking about individual patient markers in the future.

    The biggest challenge he faced—the one probably all academic clinician-researchers like him face—is lack of data and lack of data standardization. He is actually working on a foundation grant to help address this. His data came from other studies' datasets, so data are missing and stored in different places in the EMR, among other issues.

    Epidemic of Classification

    I was surprised to hear Dr. Seymour's skepticism—or at least hesitation—about classification. He told me that he thinks we have an “epidemic of classification of patients into diseases” with subtypes and sub-subtypes, which can make it harder to, of course, group and study these patients. He mentioned that he is not even sure if his groups are right, but at least he is onto something.

    After developing these phenotypes, he ran simulations of several large, famous sepsis trials using different mixes of these phenotypes, and found that outcomes were different based on how many patients were alpha. That means, in effect, if you're attempting to use a novel drug or therapy to treat “sepsis,” your trial might show harm or benefit depending on how many patients in your group had a particular sepsis phenotype.

    If you're trying a new ventilator mode or technique, for example, it might show better outcomes in gamma and delta patients, who tend to have more pulmonary issues due to sepsis, but it might show no benefit at all if most of your patients are alpha or beta, who rarely have pulmonary issues and rarely require intubation.

    No Time Soon

    Dr. Seymour said he does not think we'll be seeing therapy changes based on his phenotypes any time soon. For one thing, he is not sure they are necessarily the correct groupings, and, as we all know, regulatory and metrics changes are incredibly slow to be updated. As an emergency physician, however, I can't wait for the day that I have a standardized way to provide customized care to these patients and can stop ordering blood cultures, lactate, vancomycin, Zosyn, and 30 mL/kg of IV fluids for every single patient every single time.

    As we get closer to personalized medicine, work like Dr. Seymour's may be the first stop. We have been promised the future of genome sequencing that may tell us specifics about whether drug X will work well based on your DNA, but hopefully we can at least start to go from one-size-fits-all sepsis care to general groupings of care based on bedside presentation.

    Wouldn't it be nice to have just a teeny, tiny, itsy-bitsy dose of nuance added to our sepsis care when sepsis care is defined so broadly that it is—at least for us in the ED—quite similar initially for the patient with a lactate of 2.0 and a mild UTI as for a patient with a lactate of 10 and necrotizing fasciitis? We do this with almost everything else in medicine, so why not sepsis?

    Share this article on Twitter and Facebook.

    Access the links in EMN by reading this on our website,

    Comments? Write to us at

    Dr. Walkeris an emergency physician at Kaiser San Francisco. He is the developer and co-creator of MDCalc (, a medical calculator for clinical scores, equations, and risk stratifications, which also has an app (, and The NNT (, a number-needed-to-treat tool to communicate benefit and harm. Follow him on Twitter@grahamwalker, and read his past columns at

    Read a special report, “Sepsis: What Works? What Doesn't?” on p. 24.

    Copyright © 2020 Wolters Kluwer Health, Inc. All rights reserved.