Association between the PaO2/FiO2-to-PaCO2 ratio and mortality in patients with sepsis: a retrospective study with dual-cohort validation

BACKGROUND: ratio (PFP) in sepsis. METHODS: This study analyzed MIMIC-IV data on 16,546 patients with sepsis admitted to the ICU and categorized into three groups by PFP values calculated from the first arterial blood gas within 24 hours of admission. Associations between PFP and mortality were assessed using Kaplan-Meier, Cox regression, and restricted cubic splines (RCS) analyses, and validated using clinical data from a regional hospital. Machine learning was used to develop predictive models for 28- and 90-day mortality, with and without PFP, to evaluate its predictive strength against existing models. RESULTS: Mortality rates at 28 and 90 days varied significantly across PFP groups (P<0.001). Cox regression and Kaplan-Meier analyses showed that lower PFP was associated with higher mortality risk (log-rank P<0.001), Especially among patients receiving mechanical ventilation,which was confirmed by subgroup and sensitivity analyses.RCS analyses demonstrated a nonlinear association between PFP and mortality (cutoff: 4.85). In external validation, lower PFP was linked to higher mortality risk (cutoff: 4.85; log-rank P<0.001). Machine learning models for 28- and 90-day mortality identified PFP as a significant predictor. Models including PFP improved integrated discrimination improvement (IDI) compared with baseline models (IDI=0.6%, P=0.001; IDI=0.3%, P=0.002), indicating improved discrimination. CONCLUSION: Lower PFP values correlate with higher 28-day and 90-day mortality in septic ICU patients. PFP has modest supplementary prognostic value, mainly in mechanically ventilated patients, not a universal biomarker for all sepsis patients.