Overall
From 1 March 2020 until 30 September 2021, date of database closing for this analysis, 4041 critically ill COVID-19 patients were included in 90 centres over 22 countries into the RISC-19-ICU registry. The evolution of SARS-CoV-2 variants in the respective countries is depicted in Additional file 1: e-Figure 3. Similarly, the number of total hospitalized patients as well as patients requiring ICU care in selected countries participating in the registry is displayed in Additional file 1: Figure 4.
Overall, patients were mainly male (70%), aged 61 ± 14 years and presented with a body mass index (BMI) of 29 ± 6 kg/m2 (Table 1). They were admitted to the hospital a mean of 9 ± 12 days after symptom onset and had to be referred to the ICU 3 ± 13 days later. At ICU admission, 17% required vasopressor support and 42% were directly intubated and mechanically ventilated. Admission SOFA score amounted to 8 ± 5, and SAPS II and APACHE II scores were 36 ± 19 and 16 ± 8, respectively.
In total, 66% of all patients were invasively mechanically ventilated and 26% of the patients died during their ICU stay.
Demographics over the time course of the pandemic
One thousand seven hundred (42%) patients were admitted to the respective ICUs between March and September 2020, 1543 (38%) between October 2020 and January 2021, and 798 (20%) between February and September 2021 (Additional file 1: e-Figure 5).
Over the time course of the pandemic, severity scores at admission, namely SOFA (March 2020: 8.2 [7.6–9.0], September 2021: 5.8 [5.3–6.4]), age-corrected SAPS II (March 2020: 26 [24–29], September 2021: 19 [17–20]) and APACHE (March 2020: 12.5 [11.5–13.5], September 2021: 8 [7–9]), continuously decreased (p < 0.001) (Fig. 1, Additional file 1: e-Figure 6). The percentage of female patients admitted to the ICU started to increase after October 2020 and was highest in September 2021 (March 2020: 26 [23–29]%, October 2020: 26 [24–28]%, February 2021: 30 [28–33]%, September 2021: 41 [35–48]%, p < 0.001). On the other hand, while the mean age of patients increased during the first months of the pandemic, it steadily decreased after October 2020 (March 2020: 61 [60–63] years, October 2020: 64 [62–66] years, February 2021: 63 [61–65] years, September 2021: 55 [53–58] years, p < 0.001). Similarly, the average number of comorbidities remained constant until October 2020 and then proceeded to decrease until September 2021 (p < 0.001). Conversely, the latency between symptom onset and hospitalization (March 2020: 6.7 [6.2–7.2] days, October 2020: 7.7 [7.2–8.3] days, February 2021: 9.5 [8.8–10.3] days, September 2021: 9.7 [8.9–10.5] days, p < 0.001) as well as between hospital and ICU admission (March 2020: 1.7 [1.4–2.1] days, October 2020: 2.3 [1.9–2.8] days, February 2021: 3.6 [3.0–4.4] days, September 2021: 4.0 [3.3–5.0] days, p < 0.001) steadily increased until February 2021 and then remained constant for the remainder of the studied period.
Vitals and laboratory findings at ICU admission
Contrasting with the decreasing severity scores, patients admitted to the ICU presented continuously lower paO2/FiO2 ratios (March 2020: 132 [123–141] mmHg, October 2020: 131 [124–140] mmHg, February 2021: 120 [112–129] mmHg, September 2021: 101 [91–113] mmHg, p < 0.001) along with increasing ventilatory ratios (p = 0.01); these effects became especially pronounced after October 2020 (Fig. 2, Additional file 1: e-Figure 7). By contrast, D-dimer levels at ICU admission continuously decreased from March 2020 to September 2021 (March 2020: 1722 [1320–2241] μg/l, October 2020: 1581 [1276–1953] mmHg, February 2021: 988 [763–1272] mmHg, September 2021: 506 [338–759] mmHg, p < 0.001). Notably, whereas C-reactive protein and procalcitonin levels did remain constant between March 2020 and September 2021, ferritin levels decreased until August 2020 and then remained constant for the remainder of the pandemic. Conversely, leucocyte (March 2020: 8.1 [7.6–8.7] 109/l, October 2020: 8.9 [8.3–9.5] 109/l, February 2021: 9.8 [9.1–10.5] 109/l, September 2021: 9.1 [8.2–10.1] 109/l, p < 0.001) and neutrophil counts increased until August 2020 and then continuously decreased for the remainder of the pandemic (p < 0.001), whereas lymphocyte counts remained constant until August 2020 and then showed a similar decreasing dynamic as leucocytes and neutrophils (p = 0.08).
Disease progression over the ICU stay
Not only the admission characteristics, but also the dynamics of disease in response to ICU care can change over time. In order to capture these changes in disease progression over the first days of ICU stay, we computed the difference between day 5 and ICU admission and evaluated the change of this parameter (Δlate-early) over time.
At later stages of the pandemic, the PaO2/FiO2 ratio increased more pronouncedly during the first 5 days of ICU stay (March 2020: Δlate-early 34 [20–48] mmHg, October 2020: Δlate-early 29 [18–38] mmHg, February 2021: Δlate-early 37 [25–50] mmHg, September 2021: Δlate-early 70 [41–100] mmHg, p = 0.05), whereas the ventilatory ratio decreased more markedly (p < 0.001) (Fig. 3, Additional file 1: e-Figure 8–9, Additional file 1: e-Table 2). Similarly, C-reactive protein progressively experienced a stronger declining effect over the duration of the pandemic (March 2020: Δlate-early − 29 [− 102 to 44] mg/l, October 2020: Δlate-early − 47 [− 105 to 13] mg/l, February 2021: Δlate-early − 127 [− 189 to − 68] mg/l, September 2021: Δlate-early − 231 [− 352 to − 109] mg/l, p = 0.001). However, neither D-Dimer, ferritin and procalcitonin levels, nor leucocyte, neutrophil or lymphocyte count dynamics varied over the time course of the pandemic.
Medication management
The use of hydroxychloroquine and ritonavir/lopinavir while widely employed in the first months of the pandemic, dropped to 0% by June 2020 (Fig. 4). Similarly, therapeutic anticoagulation, whereas increasingly employed in the first year of the pandemic, experienced a decline during the second half of the pandemic (March 2020: 35 [24–48]%, October 2020: 77 [66–85]%, February 2021: 70 [57–80]%, September 2021: 45 [26–65]%, p < 0.001). Conversely, the use of corticosteroids increased steadily from 14 [9–22] in March 2020, reaching 86 [79–92] by October 2020 and 97 [94–99] by September 2021. Tocilizumab on the other hand was prescribed during the first months of the pandemic, but its prescription saw a halt between June 2020 and February 2021, after which its use steadily increased to 17 [5–47] in September 2021. Most prominently after February 2021, the proportion of vaccinated individuals admitted to the ICU steadily increased (p < 0.001) (Additional file 1: e-Figure 10).
Organ support management and outcomes
The proportion of patients receiving invasive mechanical ventilation (March 2020: 82 [76–86], October 2020: 76 [70–82]%, February 2021: 70 [61–77]%, September 2021: 74 [64–82]%, p < 0.001) and renal replacement therapy decreased (March 2020: 12 [9–16], October 2020: 9 [7–12]%, February 2021: 5 [3–7]%, September 2021: 3 [1–9]%, p < 0.001) throughout the pandemic (Fig. 5, Additional file 1: e-Table 3–5). Conversely, more invasively ventilated patients were treated with extracorporeal membrane oxygenation (ECMO) at later stages of the pandemic (March 2020: 0.4 [0.1–1.3] %, October 2020: 0.5 [0.2–2]%, February 2021: 1 [0.4–3]%, September 2021: 3 [1–9]%, p < 0.001) (Fig. 4). Overall, more patients were treated with non-invasive mechanical ventilation (March 2020: 14 [11–18], October 2020: 25 [20–31]%, February 2021: 39 [32–46]%, September 2021: 24 [17–33]%, p < 0.001) and high-flow oxygen therapy (March 2020: 5 [4–7], October 2020: 10 [8–14]%, February 2021: 24 [19–31]%, September 2021: 20 [14–29]%, v0.001) as the pandemic progressed. Additionally, awake prone position was increasingly employed from February 2021 onwards (March 2020: 50 [43–58]%, October 2020: 47 [40–54]%, February 2021: 45 [38–52]%, September 2021: 52 [42–62]%, p < 0.001). Finally, ICU mortality initially worsened until June of 2020 and then progressively improved until September 2021 (March 2020: 23 [19–26], October 2020: 23 [19–26]%, February 2021: 29 [24–33]%, September 2021: 18 [12–24]%, p < 0.001), whereas length of ICU stay continuously decreased over the time course of the pandemic (March 2020: 14 [13–16], October 2020: 13 [12–15] days, February 2021: 12 [11–13] days, September 2021: 11 [10–13] days, p < 0.001).
Development of differences between survivors and non-survivors over time
During the pandemic, patients surviving the ICU stay were characterized by a lower age than patients not surviving the ICU, albeit the difference in mean age decreased between March 2020 and October 2020 between both groups (p < 0.001) (Additional file 1: e-Figure 11). On the other hand, while non-survivors presented with higher SOFA scores at ICU admission during the whole pandemic, the difference in initial SOFA between non-survivors and survivors grew with the progress of the pandemic (p < 0.001).
The PaO2/FiO2 ratio at admission presented decreasing dynamics in both survivors and non-survivors during the pandemic (p < 0.001) (Additional file 1: e-Figure 12). Nevertheless, survivors were characterised by a more pronounced increase in PaO2/FiO2 ratio (p < 0.001) and had a stronger decrease in ventilatory ratio (p = 0.03) over the first 5 days of their ICU stay at later stages of the pandemic than non-survivors (Additional file 1: e-Figure 13). On the other hand, C-reactive protein (p < 0.001) as well as creatinine levels (p < 0.001) albeit lower at admission in survivors than non-survivors in March 2020 progressed to be similar in September 2021 (Additional file 1: e-Figure 12). Strikingly, C-reactive protein dynamics over the first days after ICU admission showed a more pronounced decline in non-survivors after June 2020, as compared to survivors (p < 0.001) (Additional file 1: e-Figure 13).
The proportion of survivors not being treated with mechanical ventilation decreased in the first year of the pandemic, to afterwards increase to its initial proportion by September 2021, while this effect was much less pronounced in non-survivors (p < 0.001) (Additional file 1: e-Figure 14). Conversely, more survivors than non-survivors were treated with high-flow oxygen over the course of the pandemic (v0.001), whereas a similar proportion of patients received non-invasive mechanical ventilation throughout the pandemic (p = 0.40). Interestingly, while survivors presented longer lengths of ICU stay than non-survivors in March 2020, this inverted with non-survivor requiring longer care in the ICU, especially during June 2020 and February 2021, as the pandemic progressed (p < 0.001).