Low arterial oxygen-content burden during the first 24 h and hospital mortality in mechanically ventilated adults: analyses of MIMIC-IV and eICU

BackgroundArterial oxygen content (CaO2) combines hemoglobin-bound and dissolved oxygen. A time-weighted CaO2 deficit could capture changing anemia and desaturation that a hemoglobin-only summary misses. We examined its association with hospital mortality and its added value beyond hemoglobin.MethodsIn this retrospective cohort study, we included mechanically ventilated adults from the Medical Information Mart for Intensive Care IV (MIMIC-IV) and eICU Collaborative Research Database (eICU). The MIMIC-IV 25th percentile of the first matched CaO2 value (12.48 mL O2/dL) was prespecified as an empirical cutoff and was applied unchanged to eICU. The first-24-h burden integrated the time-weighted deficit below this cutoff. Models were adjusted for demographics and diagnosis-derived case mix. We compared CaO2 burden with hemoglobin burden calculated at the same blood-gas time points and performed landmark, measurement-coverage, severity, and post hoc peripheral oxygen saturation (SpO2)-stratified analyses.ResultsBurden was available for 14,050 MIMIC-IV and 20,780 eICU stays. CaO2 was strongly correlated with hemoglobin (r = 0.965 and 0.949, respectively), which explained R2 = 0.932 and 0.901 of first-measurement CaO2 variation, respectively; median dissolved-oxygen contributions were 2.9% and 2.5%, respectively. Per 1-SD higher burden, adjusted odds ratios for hospital mortality were 1.33 (95% confidence interval 1.28–1.39) in MIMIC-IV and 1.29 (1.24–1.35) in eICU. Additional Acute Physiology and Chronic Health Evaluation adjustment attenuated the eICU estimate to 1.14 (1.10–1.19), while 24-h landmark estimates were 1.26 (1.21–1.32) and 1.19 (1.14–1.25), respectively. Cross-validated area under the receiver operating characteristic curve (AUROC) gains over matched hemoglobin burden were 0.006 and 0.010, respectively, and calibration was similar. In post hoc analyses, hemoglobin explained less CaO2 variation when first SpO2 was <90% (R2 = 0.787 and 0.682, respectively) than when it was ≥95% (R2 = 0.973 and 0.978, respectively); corresponding AUROC gains were 0.045 and 0.044 versus 0.002 in both databases.ConclusionIn cohorts dominated by preserved oxygen saturation, low CaO2 was predominantly an anemia-related signal, and its association with mortality was substantially attenuated by illness-severity adjustment. The burden is best viewed as a compact observational summary of the depth and duration of low arterial oxygen content, not as an independent measure of tissue oxygenation, a superior prediction model, or a treatment target.