399

Abstract

Introduction: Measures of hypoxemia (Oxygenation Index (OI), Oxygen Saturation Index (OSI), PaO2/FiO2 (PF), and SpO2/FiO2 (SF)) have been used to stratify risk of mortality in children with ARDS, but there is limited data on their inclusivity and performance among all mechanically ventilated children without cyanotic heart disease. Methods: We reviewed health records from a tertiary care PICU to calculate OI, OSI, PF and SF for the first 3 days of mechanical ventilation (MV) for children admitted from 3/2009- 4/2013. We calculated SF and PF using the closest charted FiO2 ≤ 1 h prior to SpO2 (≤ 97%) or PaO2. We calculated OI and OSI using the closest mean airway pressure ≤ 6 h prior to PF or SF. We analyzed the first value for each metric after intubation, and average values for each of the first 3 days of MV. ICU mortality was the primary outcome, and we computed area under the curve of receiver operating characteristic plots. Results: 1833 children without cyanotic heart disease were mechanically ventilated (9.9% mortality). All 4 metrics had the highest discrimination for mortality using the average values on day 1 of MV. SF could be calculated on day 1 for 1174 children (median SF alive 238 IQR (190,286), dead 141 (102,208); 11.7% mortality; AUC 0.792), OSI for 990 (median OSI alive 4.3 IQR (3.1,6.8), dead 12.7 (5, 16.6); 12.8% mortality; AUC 0.762), PF for 718 (median PF alive 373 IQR (258,468), dead 221 (132,351); 15.6% mortality; AUC 0.718) and OI for 695 (median OI alive 2.7 IQR (1.7,5.1), dead 6.9 (3.2,17.9); 16.3% mortality; AUC 0.739). All 4 metrics were available in 424 patients, in whom average PF on day 1 of MV discriminated mortality worse than SF (p=0.003) and OSI (p=0.05), and was similar to OI (p=0.09). 556 patients had only SpO2 based metrics available, and mortality discrimination was similar (Day 1 SF AUC 0.735; Day 1 OSI AUC 0.713) to the group with PaO2 available. Conclusions: SF and OSI appear to discriminate mortality better than PF. SpO2 based markers of hypoxemia roughly doubled the number of patients available for risk stratification. Such data may facilitate revision of severity of illness scores and diagnostic criteria for ARDS.