Abstract
OBJECTIVE: To determine the optimal PaO2:FiO2 threshold in the first 24 hours of intensive care unit admission, and its associated discriminatory capacity, for prognostication of mortality among critically ill patients. METHODS: This bi-national registry included adult patients admitted to intensive care units in Australia and New Zealand from January-2018 to December-2022. The primary outcome was hospital mortality. Acute hypoxic respiratory failure was defined as PaO2:FiO2 of < 300 using the worst PaO2:FiO2 within the first 24 hours of intensive care unit admission. The unadjusted association between PaO2:FiO2 and hospital mortality was evaluated using restricted cubic splines with four knots to allow for continuous, non-linear associations. To determine the optimal threshold of the PaO2:FiO2 for predicting hospital mortality, Youden's method was used to identify the maximum sum of sensitivity and specificity. The area under the receiver operating characteristic curve and Youden's J-index were calculated to compare pre-specified subgroups. RESULTS: Among the 662,612 included patients, acute hypoxic respiratory failure was not present in 324,761 (49%) patients, mild in 181,499 (27%) patients, moderate in 128,277 (19%) patients, and severe in 28,125 (4%) patients. The hospital mortality rates, respectively, were 4.9% (15,797/324,761), 7.9% (14,291/181,499), 14% (18,247/128,277), and 31% (8,717/28,125). The association between PaO2:FiO2 and hospital mortality was non-linear with an inflection point at PaO2:FiO2 = 200. The area under the ROC curve was 0.677 (95%CI 0.675 - 0.679) with an optimum PaO2:FiO2 threshold of 230. (Youden's J-index of 0.267, sensitivity 56.1% and specificity 70.6%). The area under the ROC curve was 0.627 for patients who required invasive ventilation during their intensive care unit stay, compared with 0.698 for those who did not. CONCLUSION: The optimal PaO2:FiO2 threshold for predicting hospital mortality was 230. PaO2:FiO2 has low discriminatory capacity in predicting hospital mortality among intensive care unit patients.