Identification of Distinct Clinical Phenotypes of Heterogeneous Mechanically Ventilated ICU Patients Using Cluster...

Area of Science:

• Critical Care Medicine
• Data Science in Healthcare
• Patient Stratification

Background:

• Ventilated intensive care unit (ICU) patients exhibit significant heterogeneity.
• Predicting outcomes for these patients remains challenging.
• Understanding patient subgroups is crucial for optimizing treatment strategies.

Purpose of the Study:

• To derive and validate distinct clinical phenotypes of mechanically ventilated ICU patients.
• To assess the predictive value of these phenotypes for 28-day mortality and extubation success.
• To explore differential responses to ventilation strategies among phenotypes.

Main Methods:

• Retrospective cohort study using the eICU Collaborative Research Database (eICU) for phenotype derivation.
• Cluster analysis applied to identify clinical phenotypes.
• Validation of identified phenotypes using the Medical Information Mart for Intensive Care (MIMIC-IV) cohort.
• Comparison of 28-day mortality and extubation success rates across phenotypes.

Main Results:

• Four distinct clinical phenotypes (A, B, C, D) were identified and validated.
• Phenotype A (respiratory) had lowest mortality (16%) and high extubation success (~80%).
• Phenotype B (cardiovascular) and C (renal) showed highest mortality (28%), with lower extubation success.
• Phenotype D (neurological/traumatic) had intermediate mortality (22%) and high extubation success (>80%).
• Phenotypes demonstrated varied responses to ventilation strategies regarding treatment duration but not mortality.

Conclusions:

• The identified clinical phenotypes effectively capture the heterogeneity of ventilated ICU patients.
• These phenotypes serve as valuable tools for predicting 28-day mortality and extubation success.
• Clinical phenotyping can potentially guide personalized ventilation strategies in the ICU.