Subphenotypes Of Infection In The Emergency Department: Clustering Analysis From Routine Parameters To Predict Early...

Anticipating sepsis and infection-related deterioration in the emergency department (ED) remains challenging due to the limited accuracy of available individual scores and biomarkers. We performed an unsupervised hierarchical cluster analysis using routine variables to identify subphenotypes associated with early deterioration and described their clinical and prognostic features. This retrospective study included patients presented in ED with suspected infection, excluding those without confirmed infection or septic shock. Prognostic phenotypes were identified based on clinical deterioration within 72 hours of admission, as adjudicated by an independent committee. Deterioration was defined by a composite outcome, including an increase in sepsis-related sequential organ failure assessment score ≥ 2 points, subsequent shock requiring full fluid resuscitation, acute respiratory failure needing invasive support, or death. Cluster frequency, characteristics, prognostic performance, and 28- and 90-day mortality were assessed. Among 965 consecutive patients [635 infections; 330 sepsis; mean age = 70 (52-82) years; Charlson score = 4 (1-7); sepsis-related sequential organ failure assessment score = 1 (0-3); lactates = 1.4 (1.1-2.1) mM; D-90 mortality = 11%], deterioration occurred in 155 patients (16%). The clustering analysis identified three phenotypes with distinct clinical and prognostic profiles. Early deterioration occurred in 32% of C#1, 15% of C#2, and 6% of C#3 ( P < 0.001). Cluster 1 (19%) comprised the most severe cases with pronounced organ dysfunction and tissue hypoperfusion. Cluster 2 (56%) showed moderate dysfunction and intermediate outcomes. In contrast, C#3 (25% of patients) represented younger patients with fewer comorbidities and concomitant failures. Recognition of these phenotypes based on available criteria highlights the heterogeneity of sepsis in the ED and may support phenotype-driven risk stratification and clinical management.