Clinical course and predictors of 60-day mortality in 239 critically ill patients with COVID-19: a multicenter retrospective study from Wuhan, China

Background The global numbers of confirmed cases and deceased critically ill patients with COVID-19 are increasing. However, the clinical course, and the 60-day mortality and its predictors in critically ill patients have not been fully elucidated. The aim of this study is to identify the clinical course, and 60-day mortality and its predictors in critically ill patients with COVID-19. Methods Critically ill adult patients admitted to intensive care units (ICUs) from 3 hospitals in Wuhan, China, were included. Data on demographic information, preexisting comorbidities, laboratory findings at ICU admission, treatments, clinical outcomes, and results of SARS-CoV-2 RNA tests and of serum SARS-CoV-2 IgM were collected including the duration between symptom onset and negative conversion of SARS-CoV-2 RNA. Results Of 1748 patients with COVID-19, 239 (13.7%) critically ill patients were included. Complications included acute respiratory distress syndrome (ARDS) in 164 (68.6%) patients, coagulopathy in 150 (62.7%) patients, acute cardiac injury in 103 (43.1%) patients, and acute kidney injury (AKI) in 119 (49.8%) patients, which occurred 15.5 days, 17 days, 18.5 days, and 19 days after the symptom onset, respectively. The median duration of the negative conversion of SARS-CoV-2 RNA was 30 (range 6–81) days in 49 critically ill survivors that were identified. A total of 147 (61.5%) patients deceased by 60 days after ICU admission. The median duration between ICU admission and decease was 12 (range 3–36). Cox proportional-hazards regression analysis revealed that age older than 65 years, thrombocytopenia at ICU admission, ARDS, and AKI independently predicted the 60-day mortality. Conclusions Severe complications are common and the 60-day mortality of critically ill patients with COVID-19 is considerably high. The duration of the negative conversion of SARS-CoV-2 RNA and its association with the severity of critically ill patients with COVID-19 should be seriously considered and further studied.


Background
The epidemic of novel coronavirus disease 2019  struck Wuhan in late December 2019. As of June 8, 2020, the global number of confirmed cases has reached 6,931,000, with 400,857 deaths [1]. Although most patients recovered from COVID-19, critically ill patients need long-term hospitalizations and have a considerable risk of death. Our previous small sample-sized study showed that approximately 7% of inpatients with COVID-19 developed critical complications. Among the critically ill patients, by 28-day after admission to an intensive care unit (ICU), 61.5% deceased, and 60% of the survivors were still in hospital [2]. The long-term prognosis of all survivors is unknown.
Critically ill patients with COVID-19 are characterized by progressive respiratory failure due to lung infection of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) [3]. Recent evidence suggested that SARS-CoV-2 might directly dysregulate the normal functions of the kidney, liver, and peripheral blood components, which increases the risk of multiple organ failure [3][4][5]. To our knowledge, some potential high-risk factors of death were speculated in small sample-sized studies on COVID-19 [2,4,6]. In another study exploring the risk factors of death, only 11 (8%) were treated in ICUs and only 3 (2%) received mechanical ventilation during hospitalization in the survivor group, and severe complications during the progression of COVID-19, including acute respiratory syndrome (ARDS), acute cardiac injury, acute kidney injury (AKI), and liver dysfunction, were not considered at all [7].
Here, for this WHO-declared pandemic [8], we intended to report our findings on clinical course and the 60-day mortality in 239 critically ill patients with COVID-19 from Wuhan, China.

Study design and participants
We aimed to retrospectively study critically ill adult patients with COVID-19 admitted into ICUs from Wuhan Union Hospital, Jinyintan Hospital, and Wuhan Third Hospital, from January 12 to February 3, 2020. As previously described, SARS-CoV-2 infections were confirmed by a positive result on a reverse transcriptasepolymerase chain reaction (RT-PCR) assay of specimens from the respiratory tract according to guidelines released by National Health Commission of the People's Republic of China [9]. Critically ill patients were defined to be individuals admitted to ICU, who required mechanical ventilation or had a fraction of inspired oxygen (FiO 2 ) concentration greater than or equal to 60% as described in previous reports [2,8,10,11]. Critically ill patients who deceased within 48 h after ICU admission were excluded, because their durations in ICUs were too short to reveal the effectiveness of treatments received in ICUs and to eliminate the bias on data collection of organ function or complications. Research approval (KY-2020-23.01) was granted by the ethics board of Jin Yin-tan Hospital as the central coordinating center. The need for informed consent was waived.

Criteria for ICU admission and treatment
The ICU admission criteria and treatment decisions for all patients, including determination of the need for intubation and respiratory support, were made at the discretion of the treating physicians and were not standardized. In general, the goal is to ascertain adequate oxygenation to maintain SpO 2 ≥ 90% through high-flow nasal cannula (HFNC) and noninvasive ventilation (NIV) [12][13][14]. If the respiratory failure progressively deteriorated, the patients were considered to be eligible for noninvasive or invasive mechanical ventilation when PaO 2 /FiO 2 ≤ 200 mmHg. Where available, in patients with refractory hypoxemia (PaO 2 /FiO 2 < 80 or 60 mmHg) veno-venous extracorporeal membrane oxygenation (ECMO) might serve as a therapeutic option to stabilize gas exchange [13,14].

Data collection
Patient identification in the three hospitals was achieved by reviewing admission logs from available medical records. After several cycles of feedback and pilot testing, modified case report forms referencing the case record form shared by the International Severe Acute Respiratory and Emerging Infection Consortium for SARS-CoV-2 infection. Data were extracted from local servers by experienced research physicians at each center.
Demographic data, preexisting comorbidities, vital signs at ICU admission, laboratory values at ICU admission, complications, treatments, and test result of SARS-CoV-2 RNA on samples from the respiratory tract and of serum SARS-CoV-2 IgM were collected. For patients discharged, phone calls were made by April 5, 2020, to record their living status.

Outcomes and definitions
The primary outcome was 60-day mortality and its predictors. ARDS were defined according to the Berlin Definition [15]. AKI was diagnosed according to KDIGO clinical practice guidelines based on the serum creatinine levels [16]. Acute cardiac injury was diagnosed if the serum concentration of hsTNI was measured in the laboratory above the upper limit of the reference range (> 28 pg/mL) [2]. Liver dysfunction was diagnosed if serum ALT > 50 U/L or AST > 40 U/L during disease progression [5]. Coagulopathy was defined if PT > 13.5 s or APTT > 37 s. Negative conversion of SARS-CoV-2 RNA was defined as the last time when SARS-CoV-2 RNA was tested positive on samples from the respiratory tract. Hospital-acquired infection was diagnosed if the patients had a positive culture of a new pathogen obtained from lower respiratory tract specimens (bacterial pneumonia), blood samples (bacteremia), or urine (urinary tract infection) taken ≥ 48 h after ICU admission [17].

Statistical analysis
No hypothesis was made for the present study, so sample size estimation was unavailable. Data were expressed as mean ± standard deviation, median [interquartile range], or median (range) for continuous variables and number (%) for categorical variables. Differences between survivors and non-survivors were explored using two-sample t test for parametric variables, Wilcoxon rank-sum test for nonparametric variables, and Fisher's exact test for categorical variables. Kaplan-Meier plot was used for survival data. Age was dichotomized at 65 years. Lymphocyte counts at ICU admission were dichotomized at 1.1 × 10 9 /L, the lower limit of normal range, and at 0.55 × 10 9 /L and platelet counts at 125 × 10 9 /L. Dichotomized age, lymphocyte counts and platelet counts, and comorbidities and dichotomous complications showing a p value < 0.2 in univariate analysis were included for Cox proportional-hazards regression analysis.
All statistical tests were 2-tailed with significance set at p value less than 0.05. The Stata/IC 15.1 software (StataCorp, College Station, TX, USA) was applied for all analyses.

Demographic data and comorbidities of included patients
From January 12 to February 3, 2020, a total of 1748 patients with confirmed COVID-19 from the three study centers were screened, and 258 (14.8%) critically ill patients were identified. After excluding 19 patients who deceased within 48 h after ICU admission, 239 patients were included (Fig. 1)

Laboratory tests at ICU admission
The laboratory findings of all critically ill patients on ICU admission were summarized in Table 2. At ICU admission, 219 (91.6%) patients had lymphocyte counts less than 1.1 × 10 9 /L and 103 (43.1%) had lymphocyte counts less than 0.55 × 10 9 /L. A total of 59 (24.7%) patients had platelet count less than 125 × 10 9 /L. Among 165 patients with an analysis of arterial blood gas, their partial pressure of oxygen divided by fraction of inspired oxygen was 91.3 [IQR, 66.6-133.5] mmHg.
Of all critically ill patients, by April 5th, 2020, 85 (35.6%) patients had been discharged, including 5 patients on oxygen therapy using common nasal cannula at home. Seven patients were still being hospitalized, and their oxygen therapy comprised common nasal cannula in 5 patients, high-flow nasal cannula in 1 patient, and invasive ventilator in 1 patient.
The unviable analysis result (columns 2-4 in Table 4) shows the risk effect in survival time solely determined by a predictor, without considering its relation with    (Table 4).

Discussion
In this multicenter retrospective study on critically ill patients with COVID-19, the main findings include that the critically ill patients requiring ICU admission and mechanical ventilation or oxygen therapy with FiO 2 greater than or equal to 60% had a considerable 60-day mortality and that age older than 65 years, thrombocytopenia at ICU admission, ARDS, and AKI were independent predictors of 60-day mortality of these patients. After excluding all the patients included in the previous study and incorporating two more centers, the  The low limit of normal range of lymphocyte count was 1.1 × 10 9 /L. ‡ The low limit of normal range of platelet count was 125 × 10 9 /L mortality we found in the present study was similar to our previous study [2]. Compared to the present study, the previous study was a preliminary casualty report with a smaller sample size. On February 17th, the Chinese CDC stated that of 2087 critically ill patients with COVID-19, the case fatality rate was 49.0% [18]. The study was somewhat cross-sectional. It is impossible that 51.0% of patients in the denominator had been treated more than 60 days or discharged alive, so it is reasonable that the mortality rate of critically ill patients should be higher than 49.0%. According to the official guidelines, critically ill patients were those who need mechanical ventilator support, are in shock, or need supports for other dysfunctional organs [18]. The mortality of critically ill patients with COVID-19 reported in studies outside China was also high. In the first series of critically ill patients with COVID-19 in Washington, USA, the mortality was 67% [19]. In a study of 1591 critically ill patients from Lombardy Region, Italy, 26% deceased and 58% were still in ICU [20]. In another study, 282 (88.1%) of 320 patients who received mechanical ventilation deceased [21]. Advanced age, ARDS, and AKI influenced the mortality of critically ill patients with COVID-19 in an intertwined way. There were debates on whether COVID-19 caused typical ARDS, mainly because some patients were with preserved lung compliance [22]. But a recent postmortem study showed that the fundamental pathological characteristics of pulmonary infection caused by SARS-CoV-2 in critically ill patients were diffuse alveolar damage, which was also the typical pathological finding of severe acute respiratory syndrome and Middle East respiratory syndrome [23]. We believe alveolar damage occurs at early the stage of severe COVID-19, and some of these patients recovered gradually, some deceased in a short time, and the rest deteriorated with alveolar damage progressing, fulfilling the diagnostic criteria of ARDS at some point. Like in critically ill patients with SARS and MERS [24][25][26], older critically ill patients with COVID-19 had a higher mortality [2]. In older patients, both the incidence and mortality of ARDS were higher [27][28][29][30]. Both mild-moderate and severe ARDS are associated with a substantial increase in mortality in patients with AKI [31].
Besides the general association among advanced age, ARDS, and AKI, SARS-CoV-2 induces complications by binding to the angiotensin-converting enzyme 2 (ACE2) receptor [5,32]. Angiotensin-converting enzyme 2 (ACE2), the same cell entry receptor for both SARS-CoV and SARS-CoV-2, plays an important role in organ damages in patients infected by either virus [32][33][34]. The primary target organ of SARS-CoV is the lung [35]. In the respiratory tract, angiotensin-converting enzyme 2 (ACE2) is widely expressed on the epithelial cells and macrophages of alveoli, which facilitates the progression of ARDS, and thereafter causes death [3,34,36]. As a surface molecule, ACE2 also diffusely locates on epithelial cells of the renal tubules [37]. SARS-CoV particles have been detected in the cytoplasm of these cells in postmortem studies, which explains negative conversion of SARS-CoV in urine [38]. Recently, negative conversion of SARS-CoV-2 RNA in urine has also been confirmed [4]. The destruction of epithelial cells in the renal tubules leads to acute tubular necrosis, the most common form of AKI [37]. In a previous study, advanced age and ARDS are identified as independent risk factors for development of AKI, and AKI was known as an important indicator for death in patients with SARS [39].
The mechanisms of thrombocytopenia in patients with COVID-19 are not clear right now [40]. Mechanisms of SARS-CoV-2-induced thrombocytopenia have been suggested [41,42]. On one hand, SARS-CoV-2 infects bone marrow cells, thereby causing abnormal hematopoiesis. On the other hand, the lung tissue damages with SARS-CoV-2 infection also cause platelet aggregation and consumption of platelets. The latter has been demonstrated by the findings of hemorrhagic necrosis in pathological examination of lungs from critically ill patients with COVID-19 [43].
In the process of diagnosing and treating patients with COVID, qualitative evaluations (positive or negative) of samples from the respiratory tract are widely used [44,45]. In some studies, quantitative evaluations are mainly used in exploratory studies, showing that a high load of SARS-CoV-2 may be an indicator of the severity of infection [46,47]. A median (range) duration of 30  days between symptom onset and negative conversion of SARS-CoV-2 RNA in some critically ill survivors is surprising. The inability to clear off the virus in a short time may lead to a prolonged period of critical condition in these patients. But whether they can still spread the virus needs further study [48]. Studies on antibody responses against SARS-CoV-2 are also needed.
Our previous study showed that compared with survivors, non-survivors were more likely to have preexisting comorbidities, which were not identified in the present study, except malignancy in univariate analysis [2]. The major reason was the exclusion of patients who deceased within 48 h after ICU admission. This is one limitation of the present study. In a study of patients with COVID-19 including 137 survivors and 54 non-survivors, Zhou et al. conducted a multivariable logistic regression analysis to identify risk factors of death without including severe complications after hospital admission [7]. We conducted the present study to explore the treating effect of critical care deeply by excluding the dying patients who deceased in 48 h after ICU admission. Different predicting effects of preexisting comorbidities will be identified due to different study designs [49]. The second limitation was that some important information was not available, especially PO 2 /FiO 2, which caused a failure to include APACH II score into the Cox proportional-hazards regression model. Third, 80% of included patients were from Jin-yintan Hospital, and all included patients from Jin-yintan Hospital were transferred from other hospitals. The details of treatment before admission to Jin-yintan Hospital were unavailable. But treating patients in designated hospitals was one paramount measure in dealing with the COVID-19 epidemic [50]. And we conducted Cox proportional-hazards regression stratified by study centers. Fourth, the sample size is not large enough, and other variables remain to be explored. We are expecting further studies.

Conclusion
The mortality of critically ill patients with COVID-19 is considerably high and age older than 65 years, thrombocytopenia at ICU admission, ARDS, and AKI are independent predictors of 60-day mortality. The duration between symptom onset and negative conversion of SARS-CoV-2 RNA and its association with the severity of critically ill patients should be seriously considered in treating patients with COVID-19 and needs to be studied.