Abstract Rationale Guidelines quantify acute respiratory distress syndrome (ARDS) severity by PaO2/FiO2 (PF). This index disregards differences in positive airway pressure applied by invasive mechanical ventilation (IMV), non-invasive ventilation (NIV) and high flow nasal cannula (HFNC), confounding assessment of lung injury severity. We hypothesized that modified oxygenation index (MOXI: mean airway pressure/PF), which incorporates positive airway pressure applied by each device, would better track clinical outcomes across respiratory support devices (RSD). Methods A 2-hospital cohort of adults with ARDS on HFNC, NIV or IMV was queried. Hourly MOXI and PF were obtained from medical record flowsheets and used to calculate 24-hour means for each index. A nomogram derived a priori from prior mechanistic trials was used to derive mean airway pressure from HFNC flow rates. Analyses assessed associations between respiratory indices (PF, MOXI) and clinical outcomes (30-day mortality, respiratory support-free days at 14 days RSFD), and evaluated for residual prognostic information from RSD. Results Of 414 included patients (63 years, 50% female), 37% required HFNC, 13% NIV, and 50% IMV at enrollment. There were 1.5 0,8.9 days of RSFD and 30-day mortality was 33%. Mean PF paradoxically was better (higher) with more invasive RSD (HFNC: 119, NIV: 137, IMV: 178, p 0.001). By contrast, mean MOXI was worse (higher) with more invasive RSD (HFNC: 3.8, NIV: 8.1, IMV: 9.6 p 0.001), demonstrating face validity with clinical severity. Lower PF (OR: 0.99, p = 0.005) and higher MOXI (OR: 1.12, p 0.001) predicted 30-day mortality. In a multivariable model with PF and RSD as covariates, for a given PF, odds of death were more than double with NIV or IMV compared to HFNC (NIV vs HFNC: OR 2.2, p 0.001; IMV vs HFNC: OR 2.74, p 0.001). In a similar model with MOXI and RSD, MOXI remained significantly predictive of mortality, while RSD afforded no additional prognostic information (Figure 1). Higher PF (β = 0.01, r2=0.05, p 0.01) and lower MOXI (β=-0.3, r2=0.15, p 0.01) were associated with more RSFD. When RSD was added to the model with PF, patients on IMV had 3.1 fewer RSFD (p 0.01) compared to HFNC, while there was no difference in RSFD between HFNC and NIV. When RSD was added to the model with MOXI, only MOXI and not RSD was significantly predictive of RSFD. Conclusions ARDS severity among patients on HFNC is likely overestimated by PF. Modified oxygenation index (MOXI) may more accurately represent ARDS severity across respiratory support devices. This abstract is funded by: NHLBI
Fox et al. (Fri,) studied this question.