Temporal dimensions of oxygen saturation exposure and clinical outcomes in critically ill patients: a retrospective multi-database cohort study

Abstract

Most studies on SpO2 exposure patterns focus solely on SpO2 intensity and neglect exposure duration. We systematically investigated the associations between SpO2 exposure patterns and clinical outcomes from both intensity and temporal dimensions in critically ill patients. In this retrospective cohort study, 121,758 critically ill patients were analyzed. The patient data were obtained from three databases: SICdb (n = 11,324) for continuous exposure analysis using heatmaps, MIMIC-IV (n = 50,110) for cumulative exposure analysis, and eICU (n = 60,324) for validation. We examined the relationships between SpO2 exposure and 28-day mortality using Cox regression and logistic regression models. Restricted cubic spline analysis was performed using time-weighted SpO2 as a continuous variable to examine the continuous association between SpO2 values and mortality risk. Both analytical approaches revealed a U-shaped relationship between SpO2 and mortality. Heatmap analysis revealed a SpO2 of 93.5–97% as optimal, whereas Cox regression showed that a SpO2 of 96–97% conferred the strongest protection (HR 0.714, 95% CI 0.694–0.733; P < 0.001). Prolonged exposure to SpO2 > 98% had a diminishing effect, whereas SpO2 < 93.5% increased mortality risk (HR 1.236, P < 0.001). External validation confirmed these findings (AUC 0.813). Our study confirms a U-shaped relationship between SpO2 and outcomes, emphasizing that oxygenation assessments should consider both intensity and duration. A prolonged SpO2 above 98% or below 94% should be avoided in critically ill patients.