Reduced oxygen saturation entropy is associated with poor prognosis in critically ill patients with sepsis

Margaret Gheorghita, Matthew Wikner, Anika Cawthorn, Tope Oyelade, Kristóf Németh, Patrick Rockenschaub, Ferran Gonzalez Hernandez, Nel Swanepoel, Watjana Lilaonitkul, Ali R. ManiPhysiological 2022 · DOI

TL;DR

This study reveals that reduced complexity (entropy) in oxygen saturation (SpO2) signals is a significant predictor of poor prognosis in critically ill sepsis patients. By analyzing SpO2 variability, researchers developed a composite index with mean SpO2 that rivals the established SOFA score in predicting survival, offering a simple, non-invasive tool for ICU clinicians.

What Problem Does This Paper Address?

Critically ill patients with sepsis often suffer from impaired tissue oxygenation and disrupted cardiorespiratory system connectivity. While pulse oximetry (SpO2) is a standard monitoring tool, its potential for assessing the functional dynamics of these systems, particularly in predicting patient outcomes, has been underexplored. Existing methods for assessing sepsis severity, like the SOFA score, are valuable but may not fully capture the subtle physiological changes indicative of impending decline. This paper aims to bridge this gap by investigating whether the complexity, or entropy, of SpO2 signals can serve as a non-invasive, early indicator of survival in sepsis patients.

Methods

The researchers employed sample entropy and multiscale entropy to quantify the complexity of SpO2 time-series data. Sample entropy measures the regularity of a signal, with lower values indicating higher regularity. Multiscale entropy extends this by analyzing complexity across different time scales, providing a more nuanced view of the signal's dynamics. They analyzed 30-minute segments of continuous SpO2 and respiratory rate data. Standard statistical methods, including Cox regression analysis, were used to determine the independent predictive power of SpO2 entropy and other covariates (age, SOFA score, mean SpO2, ventilation status) on patient survival. They also developed a composite index by combining SpO2 mean and entropy, and evaluated its predictive performance using Receiver Operating Characteristic (ROC) curves and Kaplan-Meier survival analysis.

Datasets

The study utilized the MIMIC-III Clinical Database, a widely used, freely accessible critical care dataset. Specifically, they extracted data from patients meeting the Sepsis-3 criteria, requiring at least 30 minutes of continuous SpO2 and respiratory rate waveform data. The final cohort for analysis comprised 164 adult patients.

Key Insights and Ideas

  • Complexity as a Health Indicator: The core idea is that a healthy physiological system exhibits complex, dynamic fluctuations. When a system is stressed or failing, these fluctuations become more regular and less complex, a phenomenon known as "de-complexification." This study applies this concept to SpO2 signals in sepsis.
  • SpO2 Entropy as a Prognostic Marker: The research highlights that SpO2 entropy, not just its average value (mean SpO2), carries crucial prognostic information. Lower entropy suggests a less adaptable, more rigid physiological state, indicative of poorer outcomes.
  • Independence from Traditional Markers: A significant finding is that SpO2 entropy's predictive power for survival is independent of established clinical markers like age, SOFA score, and ventilation status, suggesting it offers a unique layer of information.
  • Synergy of Mean and Entropy: Combining mean SpO2 with SpO2 entropy creates a composite index that significantly enhances predictive accuracy, demonstrating that both the average oxygen level and its variability are important.

Key Results

  • Lower Entropy in Non-Survivors: Patients who did not survive their ICU stay had significantly lower SpO2 entropy compared to survivors (0.070 ± 0.083 vs. 0.107 ± 0.084, p < 0.05).
  • SpO2 Entropy as a Predictor: Cox regression analysis identified higher SpO2 entropy as a predictor of survival in sepsis patients.
  • Composite Index Performance: The composite index of SpO2 mean and entropy achieved an AUC of 0.705, comparable to the SOFA score's AUC of 0.707.
  • Enhanced SOFA Score: Combining SpO2 entropy with the SOFA score further improved predictive accuracy, yielding an AUC of 0.793.
  • Standard Deviation Not Predictive: Unlike entropy, the standard deviation of SpO2 signals did not differ significantly between survivors and non-survivors and showed no prognostic value.

Limitations and Future Work

The authors acknowledge several limitations:

  • Sample Size: While 164 patients is a reasonable cohort, a larger sample size could further strengthen the findings and allow for more granular subgroup analyses.
  • Data Source and Timeframe: The MIMIC-III data was collected between 2001 and 2012. Results may not be directly generalizable to current ICU practices or populations due to evolving treatment protocols.
  • Pulse Oximeter Accuracy: While the study focused on SpO2 values generally above 90% (where pulse oximeters are more accurate), potential inaccuracies at lower saturation levels could still influence results.
  • Limited Data Availability: Detailed hemodynamic data and ventilation parameters were not available for inclusion in the analysis, which could provide further insights.

Future work suggested includes:

  • Optimal Recording Length: Determining the precise optimal length of SpO2 recording for prognostic analysis.
  • Broader Application: Investigating the utility of SpO2 entropy in other critical care settings and for different diseases beyond sepsis.
  • Ventilated Patients: Further research into the relationship between SpO2 entropy and mean in mechanically ventilated patients, particularly concerning optimal oxygenation strategies.
  • Network Physiology: Expanding the analysis to include network physiology approaches, integrating multiple physiological variables for a more comprehensive understanding.

Summary updated

Abstract

Recent studies have found that oxygen saturation (SpO<sub>2</sub> ) variability analysis has potential for noninvasive assessment of the functional connectivity of cardiorespiratory control systems during hypoxia. Patients with sepsis have suboptimal tissue oxygenation and impaired organ system connectivity. Our objective with this report was to highlight the potential use for SpO<sub>2</sub> variability analysis in predicting intensive care survival in patients with sepsis. MIMIC-III clinical data of 164 adults meeting Sepsis-3 criteria and with 30 min of SpO<sub>2</sub> and respiratory rate data were analyzed. The complexity of SpO<sub>2</sub> signals was measured through various entropy calculations such as sample entropy and multiscale entropy analysis. The sequential organ failure assessment (SOFA) score was calculated to assess the severity of sepsis and multiorgan failure. While the standard deviation of SpO<sub>2</sub> signals was comparable in the non-survivor and survivor groups, non-survivors had significantly lower SpO<sub>2</sub> entropy than those who survived their ICU stay (0.107 ± 0.084 vs. 0.070 ± 0.083, p < 0.05). According to Cox regression analysis, higher SpO<sub>2</sub> entropy was a predictor of survival in patients with sepsis. Multivariate analysis also showed that the prognostic value of SpO<sub>2</sub> entropy was independent of other covariates such as age, SOFA score, mean SpO<sub>2</sub> , and ventilation status. When SpO<sub>2</sub> entropy was combined with mean SpO<sub>2</sub> , the composite had a significantly higher performance in prediction of survival. Analysis of SpO<sub>2</sub> entropy can predict patient outcome, and when combined with SpO<sub>2</sub> mean, can provide improved prognostic information. The prognostic power is on par with the SOFA score. This analysis can easily be incorporated into current ICU practice and has potential for noninvasive assessment of critically ill patients.
BibTeX
@inproceedings{gheorghita2022reduced,
  author = {Margaret Gheorghita and Matthew Wikner and Anika Cawthorn and Tope Oyelade and Kristóf Németh and Patrick Rockenschaub and Ferran Gonzalez Hernandez and Nel Swanepoel and Watjana Lilaonitkul and Ali R. Mani},
  title = {{Reduced oxygen saturation entropy is associated with poor prognosis in critically ill patients with sepsis}},
  booktitle = {Physiological},
  year = {2022},
  doi = {10.14814/phy2.15546},
}
Figure