Fig 1.
Study cohort and emulated trial flow.
Abbreviations: ECMO, extracorporeal membrane oxygenation; IMV, invasive mechanical ventilation.
Table 1.
Baseline characteristics.
Fig 2.
Mortality in early versus late initiation of IMV groups.
Abbreviations: IMV, invasive mechanical ventilation.
Fig 3.
Primary and subgroup analyses examining mortality in early versus late initiation of IMV groups.
The hazard ratios are adjusted for the following covariates: age; sex; race; body mass index; hypertension; diabetes mellitus; coronary artery disease; congestive heart failure; chronic lung disease; active malignancy; days from symptom onset to ICU admission; hospital size (number of pre-COVID-19 ICU beds); and severity-of-illness covariates assessed on the day of IMV initiation (the renal, liver, and coagulation components of the Sequential Organ Failure Assessment score [13], the PaO2:FiO2 ratio, shock, concurrent therapies received [corticosteroids; tocilizumab; prone positioning; neuromuscular blockade], and inflammation. Sensitivity analysis #1 kept discharged patients in the risk set until day 90. Sensitivity analysis #2 excluded patients who died in the first 7 days of ICU admission, with follow-up for each patient starting on day 8 after ICU admission. Sensitivity analysis #3 was further adjusted for date of ICU admission, hospital geographic region, and hospital type. The exploratory analysis included patients not intubated during ICU days 1–7 in the “late IMV” group. Abbreviations: BMI, body mass index; HR, hazard ratio; IMV, invasive mechanical ventilation. aInflamation was defined as at least one of the following on the day of IMV initiation or prior: C-reactive protein >100 mg/L, interleukin-6 >80 pg/ml, or ferritin >1,000 ng/mL. Non-inflamed was defined as at least one value below the thresholds above, and no values that were above the thresholds. These thresholds above were selected based on prior studies [20–22].