[Skip to Navigation]
Access to paid content on this site is currently suspended due to excessive activity being detected from your IP address Please contact the publisher to request reinstatement.
Comment & Response
October 8, 2019

Identifying Sepsis Phenotypes—Reply

Author Affiliations
  • 1Department of Critical Care Medicine, University of Pittsburgh, Pittsburgh, Pennsylvania
  • 2Associate Editor, JAMA
JAMA. 2019;322(14):1417. doi:10.1001/jama.2019.12595

In Reply We agree with Dr Moser and colleagues that heterogeneity in sepsis is an important area of study. Many of the differences in patients with sepsis are clinically apparent, and yet unsupervised methods may identify emerging patterns or phenotypes not otherwise discernible at the bedside.

We also agree that “unsupervised” methods were not wholly unsupervised in our study because we relied on the inherent choices of clinicians to order examinations or laboratory tests or obtain measurements found in the electronic health record.1 Broadly speaking, these data do not capture the theoretical universe of randomly observable data in a patient with sepsis. It may even be that pertinent tests or data that offer a full permutation of this universe are not yet developed. For now, our models could run only on the observable data.