Discovery of eight stable subgroups of sepsis in the emergency department with consensus clustering

Abstract Purpose Sepsis heterogeneity is an obstacle to therapeutic advances, which may be addressed by identifying more homogenous patient subgroups. We aimed to validate the four subgroups previously identified in the SENECA study in our cohort of patients with sepsis in the emergency department (ED) and potentially propose new subphenotypes. Materials and methods This was a retrospective cohort study with laboratory and clinical data from patients admitted to the ED at Oslo University hospital with sepsis according to the Sepsis-3 definition. 953 adult patients admitted between January 4, 2019 to October 09, 2023 were included. As in SENECA, Consensus clustering was used to subgroup patients, with minor methodological adjustments. Results Four subgroups could be identified in our dataset, but they differed from the SENECA phenotypes. Eight subgroups were stable with more granular and distinct characteristics, described as: (1) Mild young, (2) Mild older, (3) Mild respiratory failure, (4) Severe respiratory failure, (5) Inflammatory, (6) Renal and multiorgan failure, (7) Hepatic and (8) Cytopenic. When using the inclusion criteria from SENECA in the Oslo cohort, only 49% were “true” sepsis patients according to Sepsis-3 criteria. Conclusions We were unable to replicate the four phenotypes identified in the SENECA study. In our data, eight subgroups were stable and demonstrated more granular characteristics. When defining sepsis cohorts, ensuring included patients fulfil current sepsis criteria will facilitate cross-cohort validation, which is essential for clinical applications.