Predicting the early incidence and stratification of acute respiratory distress syndrome (ARDS) using alveolar-arterial oxygen partial pressure difference (AaDO2): A retrospective study from the MIMIC-IV database

Abstract Purpose: To investigate the feasibility of alveolar-arterial oxygen partial pressure difference (AaDO2) for early warning and prediction of severity in patients with acute respiratory distress syndrome (ARDS). Methods: This retrospective cohort study is based on the MIMIC-IV database. In this study, both patients without and with ARDS were included. The AaDO2 value was calculated by blood gas analysis and ventilation parameters from the first day to the seventh day of the patient's admission to the ICU, and the most sensitive AaDO2 value was found by comparing the different days with the AUC curve. The Youden index was used to calculate the best threshold for predicting the incidence of ARDS. AUC curve was drawn to analyze the cutoff value of AaDO2 value for predicting mild, moderate, and severe ARDS, as well as the difference. Logistic regression analysis, lowess smoothing, and restricted cubic spline model were used to predict the association between AaDO2 and the incidence of ARDS. Results: The study included 2640 patients, of whom 1817 had no ARDS and 823 had ARDS (401 mild, 377 moderate, and 55 severe). Our study found that the earliest time point at which AaDO2 predicted the incidence of ARDS was on day 2 of ICU admission AUC was 0.815, 95% CI (0.80-0.83) with a cut-off value of 170.477 mmHg. AaDO2 on day 2 predicted mild ARDS: AUC was 0.748, 95% CI (0.727-0.769) with a cut-off value of 136.169 mmHg; AaDO2 on day 2 predicted moderate ARDS: AUC of 0.872, 95% CI (0.856-0.888), cut-off value 196.037 mmHg; AaDO2 on day 2 predicted severe ARDS: AUC of 0.932, 95% CI (0.907-0.956), cut-off value 233.896 mmHg, linear Correlation analysis found a negative correlation between AaDO2 and PaO2/FiO2 (R=0.55, R2=0.252, P<001). In addition, logistic regression analysis, Lowess smoothing, and Restricted cubic spline model analysis showed a correlation between elevated AaDO2 and poor outcome. Conclusion: AaDO2 might be utilized to predict the incidence of ARDS as well as to stratify the severity of ARDS.