Assessing mortality difference across COVID-19 intubation strategies

ABSTRACT

Rationale The optimal timing of invasive mechanical ventilation (IMV) among patients with COVID-19 related acute respiratory failure (ARF) is unknown. Use of high flow nasal cannula (HFNC) support could potentially avoid the need for IMV and related risks. However, patients failing HFNC may be at increased risk for peri-intubation complications such as cardiac arrest. At NewYork-Presbyterian Weill Cornell Medical Center (NYP-WCMC) and Lower Manhattan Hospital (LMH), an early IMV strategy prior to March 26th 2020. We subsequently switched to a prolonged observation strategy, supporting patients with non-invasive devices including HFNC. In this study, we compared in-hospital mortality in patients with ARF managed with early IMV strategy versus a prolonged observation strategy. Methods: This is a retrospective cohort study using the Weill Cornell COVID-19 Registry, which included 1869 patients admitted with a COVID-19 positive PCR test up until May 15, 2020. Patients at risk for intubation due to ARF, defined by requiring > 6 liters/min nasal cannula, were included. Patients who met ARF criteria prior to March 26, 2020 were in the early IMV strategy group, and those who met criteria on or after March 26, 2020 were in the prolonged observation strategy group. In-hospital mortality with intubation strategy as the main exposure was modelled with cox proportional hazards regression. Confounders included age, sex, BMI, comorbidities, severity of illness (SOFA) and hospital strain (difference between daily admissions and discharges). Both SOFA and hospital strain were calculated for each patient on the day that they developed ARF for modelling purposes. Results: We identified 774 patients at risk for intubation due to ARF (table), 141 were in the early IMV group and 633 were in the prolonged observation strategy group. Death occurred in 33.3% of patients in the early IMV group compared to 34.8% in the prolonged observation group. Patients in the early IMV group had a longer length of stay among survivors (27.2 ± 26.1 days vs 21.6 ± 22.8 days, p = .0213). Age-adjusted hazard ratio for death comparing early IMV versus prolonged observation was 1.35 (95% CI 0.86-2.12, which decreases to 0.87 (95% CI 0.52-1.45) after adjusting for confounders. Conclusion: In this retrospective observational study with a modest sized sample, early IMV strategy was not associated with excess mortality compared to prolonged observation. In resource constrained settings, prolonged observation with HFNC support is a reasonable hospital-level strategy in patients with ARF.