Clinical models for predicting 30-day mortality in ARDS: A focus on ventilatory ratio-defined subgroups

Area of Science:

• Critical Care Medicine
• Pulmonary Medicine
• Data Science in Healthcare

Background:

• Elevated ventilatory ratio (VR) in acute respiratory distress syndrome (ARDS) is linked to increased mortality.
• Predictors of mortality may differ between ARDS patients with high versus low VR.
• Understanding these differences is crucial for targeted interventions and improved patient outcomes.

Purpose of the Study:

• To investigate distinct mortality risk factors in ARDS subgroups based on baseline VR.
• To develop and validate subgroup-specific risk prediction models for 30-day mortality in ARDS patients.
• To assess the performance of these tailored models compared to general models.

Main Methods:

• Retrospective analysis of ARDS patients from MIMIC-IV, eICU-CRD, and ARDSnet databases.
• Patients stratified into high VR (≥2) and low VR (<2) subgroups.
• Development and validation of logistic regression models for 30-day mortality prediction in each subgroup, using internal and external validation cohorts.

Main Results:

• Two distinct models were developed: one for high VR and one for low VR ARDS patients.
• Both models showed good predictive performance (AUC ~0.76 in training cohorts).
• Subgroup-specific models significantly outperformed models developed for the general population when applied to the alternative subgroup.

Conclusions:

• Prognostic risk factors for mortality in ARDS differ significantly between high and low VR subgroups.
• Developing and utilizing VR subgroup-specific prediction models enhances accuracy in predicting 30-day mortality.
• These findings support personalized risk stratification for ARDS management.