Impact of an ROX index-guided intubation strategy on mortality in patients receiving high-flow nasal cannula: a target trial emulation

The ROX index (the ratio of peripheral oxygen saturation to the fraction of inspired oxygen, divided by respiratory rate) has been validated to predict high-flow nasal cannula (HFNC) failure, but whether acting on ROX to guide the timing of intubation improves patient outcomes is unknown. We estimated the per-protocol effect of ROX-guided strategies versus usual care on mortality after HFNC initiation. We emulated target trials using Medical Information Mart for Intensive Care IV (MIMIC-IV) electronic health record data. Adults initiated on HFNC within 7 days of intensive care unit (ICU) admission were eligible. Strategies were usual care or intubation within 2 h after ROX first fell below 3.85, 4.88, or time-varying thresholds (2.85 for hours 1–5, 3.47 for hours 6–11, and 3.85 from hour 12). We used clone-censor-weighting with pooled logistic regression, adjusted for baseline and time-varying covariates, to estimate 30-day mortality. We included 1,651 adults (median age, 66 years; women, 41%). Estimated 30-day mortality under usual care was 26.9% (95% CI 24.7–29.2). ROX < 3.85: 20.5% (95% CI 17.1–24.1; risk difference [RD], −6.5 percentage points [95% CI −9.1 to −3.8]; risk ratio [RR], 0.76 [95% CI 0.66–0.86]). ROX < 4.88: 19.8% (95% CI 16.3–24.0; RD, −7.1 percentage points [95% CI −10.3 to −3.5]; RR, 0.74 [95% CI 0.62–0.87]). Time-varying ROX: 21.3% (95% CI 18.5–24.3; RD, −5.6 percentage points [95% CI −7.9 to −3.4]; RR, 0.79 [95% CI 0.70–0.87]). ROX-guided intubation strategies were associated with lower 30-day mortality than usual care. These findings are hypothesis-generating and support prospective evaluation of the ROX index applied as a decision policy rather than as a prediction score alone; they should not be used to guide intubation decisions at present.