From October 2020 to May 2021, 583 patients with SARS-CoV-2-related respiratory failure were admitted at study centers (Figure 1). After exclusion of 47 patients because of intubation at hospital admission, 536 patients were finally included in the study. Main patients’ features are reported in Table 1.
The mean age of study patients was 69±15 years, ranging from 17 to 98 years old. One-hundred-fifty-three patients (28.5%) had no known comorbidities at hospital admission.
Blood gas parameters before the starting of oxygen therapy were available in 284 patients as the remaining were tested on supplemental oxygen.
Overall, 373 patients received steroids treatment during the hospital stay (69.6%). The use of antiviral and immunomodulant therapies was at discretion of the attending physician.
NIV was used for the management of respiratory failure in 187 patients (34.9%) (Table 1).
Use of PP
PP was used in 114 patients (21.3%), 86 males (75.4%) (Table 1). Patients managed by PP more commonly had severe respiratory failure (PaO2/FiO2 < 300 mmHg in 83.3% vs 61.3% with vs. without PP, respectively) and less commonly known comorbidities (no known comorbidities in 37.7% vs 22.5% with vs. without PP, respectively) with respect to patients managed without PP. NIV was used in 81 and in 106 patients managed or not managed by PP (43.3% vs 25.1%, respectively).
Death or intubation during the hospital stay and within 30 days from admission
In-hospital death or ICU admission at 30 days from hospitalization occurred in 163 patients (30.4%). Death occurred in 129 patients (24.1%).
In-hospital death or ICU admission at 30 days occurred in 39 patients who underwent PP (34.2%) and in 124 of those who did not underwent PP (29.4%) (Rate Ratio 1.16, 95% CI 0.87-1.56). Death occurred in 23 patients who underwent PP (20.2%) and in 102 of those who did not underwent PP (24.2%) (Rate Ratio 0.83, 95% CI 0.56-1.25).
Increasing age, severity of respiratory failure and the use of NIV were independent predictors of death or ICU admission at 30 days (Table 2). PP was not associated with the risk of death or ICU admission at 30 days (HR 1.17, CI 95% 0.78-1.74). An increased risk for death or ICU admission was observed for patients with PaO2/FiO2 between 200 and 100 mmHg and for PaO2/FiO2 < 100 mmHg at admission (Table 2 and Figure 2).
Increasing age, severity of respiratory failure, the use of NIV and qSOFA ≥2 were independent predictors of death at 30 days. PP was not associated with the risk of death at 30 days (HR 1.01, CI 95% 0.61-1.67). An increased risk for death was observed for patients with PaO2/FiO2 between 200 and 100 mmHg and for PaO2/FiO2 < 100 mmHg (Table 2).
The use of NIV (HR 3.16, 95% CI 0.75 6.19 , p 0.001) and PP (HR 2.31, 95%CI 1.35-3.95, p 0.001)(Table2) were independent predictors of ICU admission at 30 days. No association was found between age, severity of respiratory failure and ICU admission.
Sensitivity analyses
In 353 patients entering the study with ARDS, no association was observed between PP and the risk for death or ICU admission (HR 1.07, 95% CI 0.70-1.65). Mild ARDS (PaO2/FiO2 between 200 and 300 mmHg) was significantly associated with reduced risk for death or ICU admission (HR 0.33, 95% CI 0.20-0.56).
In 138 patients with moderate to severe ARDS (PaO2/FiO2 ≤200 mmHg), the only independent risk factor for death or ICU admission was age (HR 1.03 per year, 95% CI 1.01-1.04); no association was found with PP or NIV and the risk for death or ICU admission in these patients.
In 187 patients receiving treatment with NIV, no association was found between PP (HR 1.43, 95% CI 0.92-2.22) and the risk for death or ICU admission.
When 418 patients with qSOFA available were considered, increasing age and severity of respiratory failure were independent predictors of death or ICU admission while PP was not (HR 1.21, CI 95% 0.77-1.90) (Table 3 and Figure 2).
When the 114 patients who had PP were considered, the only independent predictor of death or ICU admission at 30 days was NIV (HR 3.20, 95% CI 1.13-9.06) (Table 4).
Propensity score matching analyses
Among patients’ caracteristics reported in Table 1, age, sex, PaO2/FiO2 levels, number of comorbitidies and treatment with NIV differed between patients managed with or without PP. Propensity score matching was used to balance patients’ characteristics between the cohort including the 114 patients managed by PP and a control cohort including 114 matching patients managed without PP. The final propensity score matching sample comprised 228 subjects.
The balance of measured covariates showed negligible difference between matched cohorts (standardised mean differences were all ≤0.1).
In the propensity score matching population, PP was not associated with death or ICU admission at 30 days (HR 1.33, 95% CI 0.83-2.14; p=0.193) and not with death at 30 days (HR 1.26, 95% CI 0.70-2.20; p=0.403). An association was confirmed between PP and ICU admission
(HR 2.29, 95% CI 1.29-3.86; p=0.004) (Table 5)