This study reporting the first pandemic wave results in Turkish intensive care units revealed 90-day mortality of 55.1%. Invasive mechanical ventilation, high lactate level, age more than 60 years, development of cardiac arrhythmia, need for vasopressor treatment, positive fluid balance, severe hypoxemia (PaO2/FiO2 ≤ 150) and not being fully active in performance status were determined to be the independent risk factors for 90-day mortality.
There is only one study published so far that demonstrated the 90-day mortality rate and the risk factors for mortality in critically-ill patients with COVID-19 . It was a multicenter, prospective, cohort study conducted in patients with laboratory-confirmed COVID-19 who were admitted to ICUs of 138 hospitals in France, Belgium, and Switzerland. In this study 90-day mortality was found to be 31% and older age, immunosuppression, severe obesity, diabetes mellitus, higher renal and cardiovascular SOFA score, lower PaO2/FiO2 ratio and shorter time between onset to first symptoms and ICU admission were found to be independent risk factors for 90-day mortality.
Mortality rates in critically-ill COVID-19 patients vary. In studies conducted in New York City, ICU and hospital mortality rates were found to be 39% and 60.4% respectively. [12, 24] In a multicenter, cohort study performed in 65 USA hospitals, the 28-day mortality was 35.4% . In another multicenter study conducted in Mexico, ICU mortality was determined as 51.8% . In the Italy-Lombardy region, ICU mortality was reported to be 48.7% and hospital mortality was 53.4% . In another study conducted in England, Wales and Northern Ireland region, hospital mortality was found to be 42% . ICU mortality was found to be 29%  in a study from Germany. In a retrospective study from China, 60-day mortality was reported as 61.5% . These differences might be due to differences in disease severity, in surge-capacity, in resources, etc. Turkey has sufficient number of intensive care beds, reported to be more than 30 adult ICU beds per 100,000 population, around 30% of them located in private hospitals which had limited roles in the pandemic. However, there is severe shortage of intensive care workers, especially nurses in Turkey. In the OECD countries, number of physicians per 100,000 population is 348 and number of nurses per 100,000 population is 938. In Turkey, these numbers are 348 and 301, respectively. During the COVID-19 pandemic, surge-capacity were enhanced in many countries [25–27]. In a study conducted in Australia, the number of intensive care beds in the first wave of pandemic was increased by 191%, the number of ventilators by 120%, the number of senior doctors by 240% and the number of intensive care nurses by 249% . However, in Turkey, even prior to COVID-19, number of patients per nurse could be as high as 4–5 in third level comprehensive units. Number of physiotherapists are even negligible.
As observed in our study, advanced age [13, 29] and accompanying comorbidities (hypertension, diabetes mellitus cardiovascular disease, etc.)  determine the short and long-term prognosis in patients with COVID-19.
Acute respiratory failure is the most common cause of intensive care admission in COVID-19 patients. [20, 30, 31]. Most of these patients require invasive mechanical ventilation. In studies, frequency of IMV varies between 12.2–88% [1, 16, 24, 32]. In our study, 52.4 % of the patients had PaO2/ FiO2 ratio below 150 and 57.6% of the patients needed IMV at admission or during the follow-up. In this study, both IMV and PaO2/FiO2 ≤ 150 were found to be risk factors determining 90-day mortality. In studies conducted in different parts of the world, mortality rates vary between 35.7–100% in invasively ventilated patients [14, 17, 19]. This variety among studies might be due to differences in standards of care, in surge-capacity, in resources and due to different treatment approaches in the first-wave.
In our study, the 90-day survival rate was found to be higher in patients who received HFNO, although patient numbers are few and HFNO or NIV failure rates have not been recorded in this study. With the COVID-19 pandemic, HFNO therapy has become widespread around the world . One of the advantages of this treatment is that it can be used in wards as well as in the ICU. It has been shown that HFNO therapy reduced the need for IMV in COVID-19 patients with respiratory failure .
Blood lactate levels are used mostly to follow tissue perfusion in critically-ill patients . Even minor increases in lactate levels are associated with higher mortality rates . In this study, non-survivors had higher lactate levels compared to survivors and lactate level > 2 mmol/L was an independent predictor of mortality. In COVID-19 patients, lactate levels were reported to be high in non-survivors [20, 24, 31], however, in none of them it was determined to be a risk factor in multivariate analysis.
In this study, vasopressor treatment, which was given to 41.1% of the patients, was also an independent predictor of mortality. Although we have not recorded the underlying reasons for vasopressor use, this is a predictor of disease severity and to our knowledge there is no other study reporting vasopressor treatment as an independent risk factor for mortality, except REVA network study revealing cardiovascular SOFA score being an independent risk factor for mortality.
In this study, positive fluid balance was found to be an independent risk factor for 90-day mortality in ICU patients. It was well known that positive fluid balance increases morbidity and mortality in critically-ill patients with sepsis, septic shock and ARDS [37, 38]. It had also been shown that negative fluid balance reduces the need for renal replacement therapy and respiratory failure in critically-ill patients [39, 40]. To our knowledge, this is the first study reporting positive fluid balance as an independent predictor of mortality. This finding necessitates meticulous control of fluid balance in critically-ill COVID-19 patients.
The ECOG scoring system is used to evaluate the physical performance of patients in chronic diseases such as cancer. In this study, we preferred to use ECOG score as it is easy to use and is widely known. Patients who were not fully active had poor prognosis. To our knowledge, this factor was not evaluated in other studies. However, in REVA network-COVID-ICU study, clinical frailty scale (CFS) was found to be higher in patients who died within 90 days , as was the VIP1 study where CFS was found to be related with 30-day mortality in critically-ill very elderly patients .
In studies, cardiac damage / arrhythmia was common in COVID-19 patients and it was related with increased mortality [29, 30, 43]. We observed cardiac arrhythmia in 33 (7.8%) patients, and this was a risk factor that increased 90-day mortality similar to other studies. This may be due to medications and/or due to the disease itself. In the early days of the COVID-19 pandemic, hydroxychloroquine alone or together with azithromycin was used frequently in some parts of the world [24, 29, 44] and in our country , as well. In this study, hydroxychloroquine had been given to 86% and azithromycin to 56.8% patients.
The major strength of this study is that it has been conducted as a multicenter study in Turkish ICUs. As the pandemic is a major threat globally, data from various countries and geographic regions with different income levels and resources are needed. In addition, there are few studies reporting long term mortality rates in COVID-19 patients, to our knowledge, this one being the second one reporting 90-day mortality. We even need longer term outcome. And, finally some new independent predictors of mortality such as positive fluid balance, baseline performance status, high admission lactate and vasopressor use have been determined which need to be validated in future studies.
However, this study has several limitations. Number of ICUs and patients are limited due to the short study duration, and although ICUs entering the study are from different regions of Turkey, the results might not reflect the whole country. The study was conducted during the first wave and might not reflect the current situation. In addition, there might be other factors influencing mortality, which could not have been included in the study due to its design.