Predictors of mechanical ventilation and mortality in critically ill patients with COVID-19 pneumonia

Abstract

Objective: To determine if potential predictors for invasive mechanical ventilation (IMV) are also determinants for mortality in COVID-19-associated acute respiratory distress syndrome (C-ARDS).

Design: Single center highly detailed longitudinal observational study.

Setting: Tertiary hospital ICU: two first COVID-19 pandemic waves, Madrid, Spain.

Patients or participants: 280 patients with C-ARDS, not requiring IMV on admission.

Interventions: None.

Main variables of interest: Target: endotracheal intubation and IMV, mortality.

Predictors: demographics, hourly evolution of oxygenation, clinical data, and laboratory results.

Results: The time between symptom onset and ICU admission, the APACHE II score, the ROX index, and procalcitonin levels in blood were potential predictors related to both IMV and mortality. The ROX index was the most significant predictor associated with IMV, while APACHE II, LDH, and DaysSympICU were the most with mortality.

Conclusions: According to the results of the analysis, there are significant predictors linked with IMV and mortality in C-ARDS patients, including the time between symptom onset and ICU admission, the severity of the COVID-19 waves, and several clinical and laboratory measures. These findings may help clinicians to better identify patients at risk for IMV and mortality and improve their management.

Keywords: Acute respiratory distress syndrome; Aprendizaje automático; Artificial intelligence; COVID-19; Inteligencia artificilal; Invasive mechanical ventilation; Machine learning; Predictores; Predictors; Síndrome de distrés respiratorio agudo; Ventilación mecánica invasiva.