Introduction: ARDS (Acute Respiratory Distress Syndrome) is a respiratory condition with mortality rates up to 40%. Several studies have focused on understanding ARDS, however most studies use the 2012 ARDS Berlin definition (PaO2:FiO2) without covering severity evolution. This study focus on ARDS severity evolution for ICU patients with the 2022 Berlin criteria (PaO2:FiO2 and SpO2:FiO2) and its impact on hospital and ICU mortality and LOS.
Methods: The Philips eICU dataset was used in this study. 220,259 patients were selected from these criteria: 2016-2022 ICU adults patients (18 year old or more), intubated and on mechanical ventilation, and with at least one ARDS event for the ICU stay. The severity levels of ARDS are mild, moderate and severe. Level evolution was divided into two groups: BW (Better-to-Worse) with the progressions mild to severe, mild to moderate, moderate to severe; and WB (Worse-to-Better) with the regressions severe to mild, severe to moderate, moderate to mild.
Results: Subjects’ mean age was 58±15 years old and 60% were male. Majority of patients were Caucasian (78%), African American (8%) and Other/Unknown (14%). ARDS prevalence is: 33% mild, 46% moderate and 21% severe. Hospital mortality and ICU mortality have worse outcomes when severity increases (hospital vs ICU mortality): mild (12% vs 7%), moderate (18% vs. 13%), severe (44% vs. 37%). Hospital LOS for all levels is 23 days, while ICU LOS are: 14 (mild), 17 (moderate) and 16 days (severe). Prevalence for BW is 44% and for WB is 56%; average duration for BW and WB is 33 vs. 25 hours. Hospital and ICU mortality values for the severe to moderate are 24% and 19%, respectively, and 33% and 26% for moderate to severe. The most discriminative values from labs and vitals for BW and WB are base excess (presence of strong acid or base): 1.66 meq/L (BW) vs. 1.23 meq/L (WB) (p <.001), FiO2: 45% (BW) vs. 57% (WB) (p <.001) and SBP: 122 mmHg (BW) vs. 120 mmHg (WB) (p <.001).
Conclusions: ARDS is a severe condition that has a high mortality rate in critical care with worst outcomes when severity increases. The results of this study provide statistics that expand the understanding of ARDS and open opportunities for new prediction models and data-oriented approaches to support critical care.
