A17-02 An EHR-embedded Pragmatic Stepped-wedge Randomized Controlled Trial: Optimizing Care in the Critically Ill by Liberalizing the Target O2 in Mechanically-ventilated ICU Patients

Rationale Occult hypoxemia (OH) is an oxygen saturation via arterial blood gas (SaO2) <88% despite an oxygen saturation via pulse oximetry (SpO2)≥90%. OH is common in patients with darker skin tones and has been associated with higher in-hospital mortality. We designed the Optimizing Care in the Critically Ill at UCHealth by Liberalizing the Target O22 in Mechanically-ventilated ICU Patients (OCCULT O2) trial to determine the effectiveness of limiting OH by standardizing an intermediate SpO2 range without increasing duration of mechanical ventilation. Methods This is an EHR-embedded, multisite, cluster-randomized, stepped-wedge trial implementing a multimodal educational and EHR order panel intervention aimed at achieving an intermediate oxygen saturation (SpO2 90-96%) compared to the previous standard of care (SpO2 88-94%) in mechanically ventilated adult patients admitted to the Intensive Care Unit (ICU) across a ten-hospital health system. The primary outcome is ventilator free days (VFD30), defined as days alive and not requiring invasive mechanical ventilation (IMV) in 30 days. Secondary outcomes include: ICU free days (IFD30), hospital free days (HFD30), and instances of OH. We analyzed primary and secondary outcomes using generalized linear mixed effects modeling frameworks. Results reported as median [IQR] unless noted otherwise. Results We enrolled 3,326 patients between September 15, 2024, and May 31, 2025. 1,351 (41%) were female, 2,317 (70%) self-identified as white, and 270 (8.1%) as black. Age was 63 years [48-73], with a baseline SOFA score of 8 [6-11], and SpO2/FiO2 of 167 [100-240]. Patients received 52 hours [24-121] of IMV, during which SpO2s were modestly increased (mean 96.31% vs 96.10%, Δ SpO2 0.21%, p = 0.006), and more patients experienced hyperoxemia (SpO2 >96% when FiO2 >0.21) (95% (1,424) vs. 93% (1,708), p = 0.025) in the intervention group. Overall, 327 (9.8%) patients had OH, but OH rates or the proportion with hypoxemia (SpO2 <88% when FiO2 <100%) (51% (938) vs. 52% (778), p = 0.7) were not different (primary and secondary outcomes in Table 1). Conclusion Embedding an EHR order panel intervention to standardize a higher, intermediate SpO2 target range across a 10-hospital health system facilitated successful execution of a cluster-randomized, stepped-wedge trial, but did not change clinically important outcomes, including VFD30. Both the mean SpO2 and hyperoxemia frequency indicate that higher than targeted SpO2s are quite common in clinical care among mechanically ventilated patients at UCHealth. Yet, because raising the SpO2 target did not reduce OH rates, targeting an intervention in real-time while patients are experiencing OH may be more effective. This abstract is funded by: National Center for Advancing Translational Sciences of the National Institutes of Health under Award Number UM1TR004399