Pathogens analysis and modeling of mortality risk in sepsis patients with COVID-19 and without COVID-19

This study compared isolate-level pathogen profiles, antimicrobial susceptibility, clinical characteristics, and mortality predictors between sepsis patients with and without coronavirus disease 2019 (COVID-19), and developed cohort-specific nomograms for in-hospital mortality. This retrospective intensive care unit (ICU) cohort included 608 adults with sepsis: 158 in group COVID-19 and 450 in group non-COVID-19. All patients were assessed for severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) detection status. Patient-level comparisons and mortality modeling used the full cohort, whereas microbiological analyses were restricted to 235 eligible bacterial or fungal isolates from baseline cultures after exclusion of colonizers, contaminants, clinically non-causative organisms, and duplicates. Candidate predictors were selected by least absolute shrinkage and selection operator (LASSO) regression and entered into multivariable logistic regression; coefficients were pooled using Rubin's rules when multiple imputation was performed. External validation used MIMIC-IV with fixed internal coefficients. Gram-negative organisms predominated, mainly Acinetobacter baumannii and Klebsiella pneumoniae, with substantial antimicrobial resistance. The COVID-19 and non-COVID-19 models showed apparent internal area under the curve values of 0.938 and 0.871, conservative optimism-corrected values of 0.913 and 0.871, and external validation AUC values of 0.841 and 0.859, respectively. Cohort-specific nomograms may provide supplementary risk-stratification information in critically ill patients with sepsis.