Objectives: To assess the response to initial oxygenation strategy according to clinical variables available at admission.
Design: Multicenter cohort study.
Setting: Thirty French and Belgium medical ICU.
Subjects: Immunocompromised patients with hypoxemic acute respiratory failure.
Interventions: None.
Measurements and main results: Data were extracted from the Groupe de Recherche en Reanimation Respiratoire du patient d'Onco-Hématologie database. Need for invasive mechanical ventilation was the primary endpoint. Secondary endpoint was day-28 mortality. Six-hundred forty-nine patients were included. First oxygenation strategies included standard oxygen (n = 245, 38%), noninvasive ventilation (n = 285; 44%), high-flow nasal cannula oxygen (n = 55; 8%), and noninvasive ventilation + high-flow nasal cannula oxygen (n = 64; 10%). Bilateral alveolar pattern (odds ratio = 1.67 [1.03-2.69]; p = 0.04), bacterial (odds ratio = 1.98 [1.07-3.65]; p = 0.03) or opportunistic infection (odds ratio = 4.75 [2.23-10.1]; p < 0.001), noninvasive ventilation use (odds ratio = 2.85 [1.73-4.70]; p < 0.001), Sequential Organ Failure Assessment score (odds ratio = 1.19 [1.10-1.28]; p < 0.001), and ratio of PaO2 and FIO2 less than 100 at ICU admission (odds ratio = 1.96 [1.27-3.02]; p = 0.0002) were independently associated with intubation rate. Day-28 mortality was independently associated with bacterial (odds ratio = 2.34 [1.10-4.97]; p = 0.03) or opportunistic infection (odds ratio = 4.96 [2.11-11.6]; p < 0.001), noninvasive ventilation use (odds ratio = 2.35 [1.35-4.09]; p = 0.003), Sequential Organ Failure Assessment score (odds ratio = 1.19 [1.10-1.28]; p < 0.001), and ratio of PaO2 and FIO2 less than 100 at ICU admission (odds ratio = 1.97 [1.26-3.09]; p = 0.003). High-flow nasal cannula oxygen use was neither associated with intubation nor mortality rates.
Conclusions: Some clinical characteristics at ICU admission including etiology and severity of acute respiratory failure enable to identify patients at high risk for intubation.