Clinical and inflammatory profiles of pulmonary versus non-pulmonary sepsis in ICU patients: a single-center retrospective study

In this single-center retrospective observational study, we compared the clinical characteristics, pathogen spectrum, inflammatory profiles, and outcomes between patients with pulmonary and non-pulmonary sepsis from January 2024 to October 2025. Based on the source of infection, a total of 110 patients with sepsis were included and categorized into groups of pulmonary (n = 53) and non-pulmonary (n = 57) infection. Clinical and laboratory data, treatments, and outcomes were compared between two groups. Patients with pulmonary sepsis were older and more likely to be female, with a higher prevalence of cerebrovascular disease, while non-pulmonary sepsis patients had more cases of cancer. Pathogen analysis revealed that Klebsiella pneumoniae (12/53) and COVID-19 (11/53) were prominent in pulmonary sepsis, whereas Escherichia coli (20/57) was the most common pathogen in non-pulmonary sepsis. Non-pulmonary sepsis patients had higher levels of PCT, IL-6, BNP and lower PaCO₂ (all p < 0.05). SOFA scores at 96 h were higher in the pulmonary group (7.0 vs. 4.0, p < 0.05). The in-hospital mortality rate was 60.38% in the pulmonary group and 42.11% in the non-pulmonary group ( p = 0.06). After adjusting for age, sex, cancer, cerebrovascular disease, baseline SOFA score, and vasopressor use, pulmonary sepsis remained significantly associated with higher in-hospital mortality (aOR: 3.70, 95% CI: 1.35–10.11, p = 0.011). Sensitivity analyses excluding extreme PCT and/or IL-6 values showed directionally consistent but statistically imprecise estimates, with only the IL-6 exclusion analysis remaining significant. The median time to in-hospital death was 18.0 days (95% CI: 13.8–22.2) for pulmonary sepsis and 24.0 days (95% CI: 11.5–36.5) for non-pulmonary sepsis (log-rank p = 0.402). The unadjusted mortality comparison was of borderline significance ( p = 0.06), but remained significant after adjustment for confounders. Our data showed that sepsis from pulmonary and non-pulmonary sources exhibits distinct pathogen distributions and inflammatory profiles. As infection source may contribute to clinical heterogeneity in sepsis, there is a need to explore source-specific management strategies.