Arterial blood gas as a predictor of mortality in covid pneumonia patients initiated on noninvasive mechanical ventilation: A retrospective analysis

Gupta, B.; Jain, G.; Chandrakar, S.; Gupta, N.; Agarwal, A.

Background: The alveolar–arterial oxygen (A–a) gradient measures the difference between the oxygen concentration in alveoli and the arterial system, which has considerable clinical utility. Methods: It was a retrospective, observational cohort study involving the analysis of patients diagnosed with acute COVID pneumonia and required noninvasive mechanical ventilation (NIV) over a period of 3 months. The primary objective was to investigate the utility of the A–a gradient (pre-NIV) as a predictor of 28-day mortality in COVID pneumonia. The secondary objective included the utility of other arterial blood gas (ABG) parameters (pre-NIV) as a predictor of 28-day mortality. The outcome was also compared between survivors and nonsurvivors. The outcome variables were analyzed by receiver-operating characteristic (ROC) curve, Youden index, and regression analysis. Results: The optimal criterion for A–a gradient to predict 28-day mortality was calculated as ≤430.43 at a Youden index of 0.5029, with the highest area under the curve (AUC) of 0.755 (p <0.0001). On regression analysis, the odds ratio for the A–a gradient was 0.99. A significant difference was observed in ABG predictors, including PaO2, PaCO2, A–a gradient, AO2, and arterial–alveolar (a–A) (%) among nonsurvivors vs survivors (p-value <0.001). The vasopressor requirement, need for renal replacement therapy, total parenteral requirement, and blood transfusion were higher among nonsurvivors;however, a significant difference was achieved with the vasopressor need (p <0.001). Conclusions: This study demonstrated that the A–a gradient is a significant predictor of mortality in patients initiated on NIV for worsening respiratory distress in COVID pneumonia. All other ABG parameters also showed a significant AUC for predicting 28-day mortality, although with variable sensitivity and specificity.