Risk factors of mortality for intensive care COVID-19 patients: A retrospective cohort study

Yanli Gu Jinling Hospital, Nanjing Medical University Donghui Wang Jinling Hospital, Nanjing Medical University Cen Chen Jinling Hospital, The First School of Clinical Medicine, Southern Medical University Wanjun Lu Jinling Hospital, Medical School of Nanjing University Hongbing Liu Jinling Hospital, Medical School of Nanjing University Tangfeng Lv Jinling Hospital, Medical School of Nanjing University Yong Song Jinling Hospital, Medical School of Nanjing University Fang Zhang (  zhangfanglab@163.com ) Jinling Hospital, Medical School of Nanjing University


Introduction
Since the rst case was diagnosed in Wuhan, China, COVID-19 which caused by severe acute respiratory syndrome coronavirus 2(SARS-Cov-2) has spread worldwide. Up to September 14, 2020 The clinical manifestations of COVID-19 include asymptomatic infection, mild upper respiratory symptoms, and respiratory failure requiring advanced life support 3,4 . The severity of COVID-19 was classi ed into mild, common, severe and critically ill according to the guidelines on the diagnosis and treatment of new coronavirus pneumonia published by the National Health Commission of China. A signi cant proportion of severe and critically ill patients required intensive care, and had high mortality rates. If patients with high risk of mortality can be identi ed early on ICU admission, it will help to focus treatment efforts on these patients and might reduce the mortality rate of COVID-19.
To identify the risk factors of mortality for COVID-19 patients in ICU, we retrospectively analyzed clinical data of 57 patients admitted to Huoshenshan Hospital in Wuhan, China. After analyzed by logistic regression model, we identi ed IL-6 and PaO2/FiO2 as independent risk factors.

Results
Demographic and Clinical data From February 10 to April 10, 2020, 71 con rmed COVID-19 patients were admitted to the ICU, including 28 deaths and 43 survivors. As shown in table 1, after excluding patients with incomplete information, 20 cases (14 males, 6 females) in deceased group and 37 cases (24 males, 13 females) in surviving group were included. The mean time from illness onset to ICU admission was 29.07 days. No statistical difference was observed in age and sex between deceased group and surviving group. There was no statistically signi cant difference in underlying diseases, including hypertension, diabetes mellitus, coronary heart disease, cerebrovascular disease, COPD, and hepatic/renal insu ciency between the two groups (all p>0.05). The most common symptom of COVID-19 patients was cough, followed by fever, dyspnea, chest tightness, fatigue, poor appetite, muscle soreness, which were all similar in the two groups (all p >0.05). No statistically difference was observed in vital signs including body temperature, respiratory rate, heart rate, systolic blood pressure (SBP) and diastolic blood pressure (DBP) (all p >0.05) at ICU admission.
As shown in gure 1, The AUC of IL-6 and PaO2/FiO2 were 0.9 (95%CI:0.823-0.977, p<0.0001) and 0.865 (95%CI:0.774-0.956, p<0.0001) respectively. The cut-off value of IL-6 was 25.69 pg/mL, the sensitivity was 95% and the speci city was 75.7%, while the cut-off value of PaO2/FiO2 was 167.79 mmHg, the sensitivity was 75.7% and the speci city was 85%. Furthermore, we observed the variation trend of IL-6 in six patients (3 in deceased group and 3 in surviving group) who had tested plasma IL-6 for more than 5 times during hospitalization. The levels of IL-6 decreased gradually as the condition of patients improves in surviving group, while increasing when patients' condition deteriorated in deceased group (Fig 2).

Discussion
The SARS-CoV-2, known as the seventh human coronavirus, belonging to β coronavirus, is the pathogen of COVID-19. There are six coronaviruses infecting humans previously, including SARS-CoV-1 and middle east respiratory syndrome (MERS)-CoV in 2003 and 2012 respectively. But the global pandemic caused by SARS-CoV-2 is unprecedented. The mortality rates of SARS and MERS were more than 10% and 35% respectively 5,6 , while it is approximately 3.12%-5.43% in hospitalized COVID-19 patients according to data from National Health Commission of the People's Republic of China and World Health Organization.
Although the mortality rate of COVID-19 is lower than that of SARS and MERS, the overall number of deaths is higher due to the larger number of infections. Clinical data from Wuhan, China, indicate that approximately 17.7% to 32% of patients require ICU admission, and the mortality rate of critically ill patients is as high as 49% to 61.5% 7,8 . Early identi cation of individuals at high risk of mortality might help reduce mortality of COVID-19.
Angiotensin-converting enzyme 2 (ACE2), expressed on pulmonary epithelial cells, vascular endothelial cells and macrophages, is the target receptor of SARS-COV-2 [9][10][11] . Decreased expression of ACE2 in the lungs may lead to acute lung injury 12,13 . Cell apoptosis could induce local in ammatory response, which leads to the release of proin ammatory cytokines and chemokines into blood, including IL-1, IL-6, interferon Gamma (IFN-γ), Monocyte chemoattractant protein 1 (MCP1), IFN-γ -induced protein 10 (IP-10) and so on 3 . Under normal circumstances, this in ammatory response helps eliminating microorganism and facilitating patients recover. But excessive release of in ammatory cytokines, called cytokine release syndrome (CRS), can exacerbate in ammation response and damage lung tissues. Evidence shows that CRS plays an important role in the pathogenesis of COVID-19. Given the precise role of CRS in severe COVID-19, early recognition of this excessive in ammatory response and early intervention, such as glucocorticoids, immunoglobulin, and selective cytokine inhibitors, might help reducing severe COVID-19 mortality 14 . IL-6, one of the in ammatory cytokines signi cantly elevating in COVID-19 patients, is a key driver of the in ammatory process 3,15 . Excessive IL-6 can lead to organ damage, such as increasing vascular permeability 16 and decreasing myocardial contractility 17 , meanwhile serving as a biomarker for predicting disease severity 18 . A previous large retrospective cohort study showed that IL-6 was associated with death in COVID-19 patients 19 , which was consistent with our results.
Studies had shown the e cacy of tocilizumab against COVID-19 22,23 . However, all of them are retrospective studies and the number of reported cases is small. Larger random control trials are needed in the future to con rm the therapeutic effect of tocilizumab on COVID-19.
As the most commonly used oxygenation index, PaO2/FiO2 is included in sepsis management guideline 24 and acute respiratory distress syndrome(ARDS) 25 although it may overestimate the incidence and underestimate the mortality of ARDS 26 . In this study, oxygenation index PaO2/FiO2 was another independent predictor of COVID-19 death.
This study has several limitations. First, as a single-center retrospective study with relatively small sample size, the results needs to be validated by more studies. Second, because the imaging severity was not evaluated by computed tomography, the chest X-ray results might be inconsistent with real lung lesion. Third, not all laboratory tests were done in all patients, such as serum ferritin, T lymphocyte subpopulation.
In conclusion, IL-6 and PaO2/FiO2 are independent risk factors for predicting death in COVID-19 patients requiring intensive care, especially when IL-6 >25.69 pg/ml and PaO2/FiO2<167.79 mmHg. Clinicians should pay enough attention to these two indicators and take active intervention measures as early as possible in order to reduce mortality.

Study design and participants
After excluding patients with incomplete information, a total of 57 patients admitted to ICU in Huoshenshan Hospital from February 10 to April 10, 2020 were included in this single-center, retrospective study. All patients were con rmed SARS-CoV-2 infection by reverse transcription-polymerase chain reaction. Patients who admitted into ICU should satis ed any of the following criteria: 1. respiratory failure requiring mechanical ventilation; 2. unstable vital signs requiring electrocardiogram monitoring; 3.
combining with other organ failure such as gastrointestinal bleeding, heart failure, renal failure, etc. This study was approved by the Ethics Committee of the Wuhan Huoshenshan Hospital and informed consents was obtained from all individual participants or their families included. All methods were performed in accordance with the relevant guidelines and regulations.

Data collection
All data were obtained from the electronic medical system and was independently collected by two researchers to check the accuracy of data. Detailed demographic information, underlying diseases, clinical symptoms, vital signs, laboratory ndings, imaging severity and treatment strategies of all patients were recorded when they entered the ICU. Demographic information included age and sex. Underlying diseases included hypertension, diabetes mellitus, coronary heart disease, cerebrovascular disease (cerebral infarction / hemorrhage), chronic obstructive pulmonary disease (COPD) and hepatic/renal insu ciency, etc. Clinical symptoms included fever, cough, dyspnea, chest tightness, fatigue, poor appetite, and muscle soreness. Vital signs included body temperature, heart rate, respiratory rate, blood pressure, etc. Laboratory tests included blood routine, liver, kidney, heart, coagulation indexes, biological indicators related to in ammation or infection, oxygenation index (PaO2/FiO2), PaCO2, etc. Some indicators had been tested several times during hospitalization but the most recent examination results were analyzed in this study. According to the lung lesion range, the chest X-ray nding were divided into mild, moderate, severe. Mild was de ned as the lesion area involving 1-2 lung elds, moderate involving 3-4 lung elds and severe involving 5-6 lung elds. Therapeutic agents included antivirus drugs, antibiotics, corticosteroid, tocilizumab, convalescent plasma, immunoglobin, albumin, thymalfasin. Life support treatments included high-ow nasal cannula oxygen therapy (HFNC), non-invasive mechanical ventilation (NIV), invasive mechanical ventilation (IMV), extracorporeal membrane oxygenation (ECMO), continuous renal replacement therapy (CRRT).

Data analysis
Student's t test and chi-square test were used to compare continuous variables and categorical variables respectively. Continuous variables and categorical variables were expressed as mean ± standard deviation (SD) and frequency respectively. The multivariate logistic regression model was used to identify independent risk factors of mortality. P<0.05 two-tailed was considered statistically signi cant. All statistical analyses were performed using SPSS (version 25.0).