Patients
From March 6, 2020, to April 30, 2021, a total of 7063 ICU patients with confirmed SARS-CoV-2 were reported to SIR. We excluded 574 patients with a primary diagnosis not associated with COVID-19 and 139 patients without data on follow-up (patients who are not Swedish residents and receive temporary patient numbers and patients emigrating during the study), yielding a total study cohort of 6350 ICU patients with COVID-19 (figure 1).
Approximately half of the patients (46.4%) were treated with prone position ventilation (table 1).
Patients in the total study cohort never receiving prone positions ventilation during ICU stay are hereafter referred to as “prone=no” and patients receiving prone position as “prone=yes”.
From the total study cohort, we extracted a sub-cohort of patients with PaO2/FiO2 ratio ≤20 kPa on ICU admission and receiving invasive mechanical ventilation within 24 hours from ICU admission. To generate this sub-cohort the following patients were excluded from the total study cohort: 1978 patients with no data on PaO2/FiO2 ratio on ICU admission, 568 patients with PaO2/FiO2 >20 kPa on ICU admission, 1877 patients who did not received invasive mechanical ventilation within 24 hours from ICU admission, 57 patients with no information on proning and 156 patients with no information on time of start of proning (Figure 2).
Thus, 1714 ICU patients were identified in this sub-cohort and included in the logistic regression models of 30- and 90-day mortality. Patients within this sub-cohort who were not treated with prone position initiated within 24 hours from start of invasive mechanical ventilation are hereafter referred to as “early prone = no” and patients treated with prone position within 24 hours from start of invasive mechanical ventilation as “early prone = yes”.
Total study cohort – baseline characteristics and process of care
Of the total cohort 6350 patients, 1843 (29.0%) were women. Median age was 64 (IQR 55–72) years. For 393 (6.2%) of the patients there were no data on proning. Most of the patients were admitted from hospital floor (77.6%). Median duration of symptoms before ICU admission was 10 (IQR 7–13) days. Approximately one quarter (27.8%) of the patients had no reported comorbidity on admission. Baseline characteristics are presented in detail in supplementary table 1.
Median total length of ICU stay was 10 (IQR 4–19) days; 5 (IQR 2-11) days for “prone=no” and 15 (IQR 9-26) days for “prone=yes”. Overall, 4174 patients (65.7%) patients received invasive mechanical ventilation. The median total duration of invasive mechanical ventilation was 11.6 (IQR 6.2-20.4) days. The median time from ICU-admission to start of invasive mechanical ventilation was 4.0 hours and the median time from ICU-admission to treatment in prone position was 20.4 hours. Renal replacement therapy was reported in 771 (13.5%) of the patients, 55 (1.6%) patients received extra corporeal membrane oxygenation (ECMO) and 1562 (24.6%) patients underwent tracheostomy. Care provided in the ICU for the total study cohort is presented in supplementary table 2.
Patients with PaO2/FiO2 ratio ≤20kPa on ICU admission and receiving invasive mechanical ventilation within 24 hours from ICU admission - baseline characteristics and process of care
Baseline characteristics are presented in table 1. Of the 1714 patients, 512 (29.9%) were women and the median age was 64 (IQR 55–71) years. Approximately one-third (28.8%) had no reported comorbidity on admission; 356 (31.1%) and 137 (24.0%) for “early prone=no” and “early prone=yes”, respectively. Median PaO2/FiO2 ratio on ICU admission was 11.6 (IQR 8.8-14.8) kPa for “early prone=no” and 9.9 (IQR 8.0-12.6) kPa for “early prone=yes”. For “early prone=no”, median SAPS 3 score was 58 (IQR 53–65) and “early prone=yes” was 58 (IQR 53-66).
The use of early prone increased from 8.5 % in March 2020 to 48.1% in April 2021. The median total duration of invasive mechanical ventilation was 11.0 (IQR 5.7-18.8) days for patients with “early prone = no” and 11.2 (6.9-18.5) days for patients with “early prone = yes”. Renal replacement therapy was reported in 20.2% and 16.7% for “early prone=no” and “early prone=yes”, respectively. Median total length of ICU stay was 14 (IQR 8-22) days for “early prone=no” and 14 (IQR 9-22) days for “early prone=yes”. Care provided in the ICU for the sub-cohort is presented in table 2.
Mortality
Overall, 30-day mortality was 24.3%, 22.3% and 26.4% for “prone=no” and “prone=yes”, respectively. For patients with no information on proning 30-day mortality was 23.2%. Corresponding figures for 90-day mortality were, 29.0%, 24.3, 33.6, and 28.8 (supplementary table 1).
For patients with PaO2/FiO2 ratio ≤20kPa on ICU admission and receiving invasive mechanical ventilation within 24 hours from ICU admission, 30-day mortality was 29.2%, 30.2% and 27.2% for “early prone=no and “early prone=yes” respectively. The corresponding figures for 90-day mortality were 33.5%, 34.8% and 31.1% (table 1.)
As shown in Figure 2, 1978 (31.1%) patients had missing data on PaO2/FiO2 ratio on ICU admission. A majority of these patients did not receive mechanical ventilation on admission, and under these circumstances FiO2 is not routinely reported to SIR. The median age among these patients was 64 (IQR 54-72) years, 542 (27.4%) were women and 393 (19.8%) patients had died within 30 days from ICU admission.
Logistic regression estimating the association between initiation of early prone position ventilation and mortality
The regression analyses included only patients with PaO2/FiO2 ratio ≤20kPa on ICU admission and receiving invasive mechanical ventilation within 24 hours from ICU admission. On univariate logistic regression analysis of 30-day mortality, the odds ratio for “early prone=yes” compared to “early prone=no” was 0.86 (95% CI 0.69-1.08), and after adjustment the corresponding odds ratio was 0.92 (0.71-1.19) (Table 3).
The odds ratios remained somewhat unchanged in the analyses restricted to patients with a) PaO2/FiO2 ratio ≤13.3 kPa on ICU admission and b) patients still in treated in ICU within 48 hours from ICU admission, Supplementary Table 3 and 4. We also performed a logistic regression model of 90-day mortality with almost identical results, Supplementary Table 5.