Interpreting peripheral oxygen saturation variability in critical illness: A directional framework adjusted for...

Area of Science:

• Physiology
• Biomedical Engineering
• Critical Care Medicine

Background:

• Peripheral oxygen saturation (SpO2) fluctuations during hypoxia can be quantified using entropy measures.
• SpO2 entropy analysis offers insights into physiological regulation and adaptation to challenges.
• Interpreting SpO2 entropy alone is limited without considering the severity of physiological challenge, such as hypoxia.

Purpose of the Study:

• To develop and validate a method to contextualize SpO2 entropy analysis.
• To assess the utility of directional parenclitic deviation (δ) as a biomarker for predicting 30-day ICU mortality in patients with sepsis, COPD, ALF, and cirrhosis.

Main Methods:

• Retrospective analysis of continuous 1 Hz SpO2 recordings from the MIMIC-III dataset.
• Calculation of sample entropy from raw 20-minute SpO2 signals.
• Normalization of SpO2 entropy using directional parenclitic deviation (δ) derived from a healthy hypoxia-exposure reference dataset.
• Application of Cox-regression models to assess 30-day ICU mortality.

Main Results:

• In sepsis patients, δ was significantly higher in non-survivors and independently predicted 30-day mortality (HR=1.79, P<0.0001).
• δ was not predictive of mortality in the COPD, ALF, or cirrhosis cohorts.
• The cirrhosis cohort exhibited unexpected negative mean δ values, suggesting potential aberrant regulatory engagement related to hepatopulmonary syndrome.

Conclusions:

• Directional parenclitic deviation (δ) provides crucial physiological context to entropy-based SpO2 analysis.
• This framework links SpO2 variability to hypoxia severity, enabling a more interpretable and potentially clinically applicable biomarker for systemic regulation in critical illnesses.
• Further validation in diverse cohorts could support the use of δ for personalized critical care.