Mechanical power normalisation methods to predict ICU mortality: a retrospective cohort study

Area of Science:

• Critical Care Medicine
• Respiratory Physiology
• Mechanical Ventilation

Background:

• Ventilator-induced lung injury (VILI) is a significant concern in mechanical ventilation.
• Optimal strategies to minimize VILI are still under investigation.
• Mechanical power (MP) is a recognized determinant of VILI, but its normalization requires further study.

Purpose of the Study:

• To assess if normalizing MP to patient-specific factors improves its accuracy in predicting ICU mortality.
• To explore the relationship between MP, MP ratio, and mortality, determining if it's linear or has a threshold effect.

Main Methods:

• Retrospective analysis of electronic health records from 3,578 mechanically ventilated adults.
• Evaluated MP normalized to dead space (ventilatory ratio, etc.), lung compliance, and a normal ideal MP (MP ratio).
• Used logistic regression and Area Under the Receiver Operating Characteristic Curve (AUROC) to assess mortality prediction and linearity.

Main Results:

• MP normalized to compliance (AUROC 0.71) and MP ratio (AUROC 0.71) showed improved discrimination for ICU mortality compared to unnormalized MP (AUROC 0.69).
• Other normalization methods did not significantly enhance predictive accuracy.
• A statistically significant, though small, non-linear relationship was observed between MP/MP ratio and ICU mortality.

Conclusions:

• Mechanical power normalized to compliance and MP ratio offer superior discrimination for ICU mortality.
• The improvement in predictive accuracy is modest, and the absolute predictive power remains limited.
• Further research into VILI determinants and optimal ventilation strategies is warranted.