Risk Factors Related to Acute Respiratory Distress Syndrome and Death in Patients with COVID-19: A Retrospective Cohort Study in Wuhan, China

Background: The COVID-19 pandemic has been considered as the great threat to global public health. We aimed to clarify the risk factors associated with the development of acute respiratory distress syndrome (ARDS) and progression from ARDS to death and construct a risk prediction model. Methods: In this single-centered, retrospective, and observational study, 796 COVID-19 patients developed ARDS and 735 COVID-19 patients without ARDS were matched by propensity score at an approximate ratio of 1:1 based on age, sex and comorbidities. Demographic data, symptoms, radiological ndings, laboratory examinations, and clinical outcomes were compared between with or without ARDS. Univariable and multivariable logistic regression models were applied to explore the risk factors for development of ARDS and progression from ARDS to death and establish a comprehensive risk model. Results: Higher SOFA, qSOFA, APACHE II and SIRS scores, elevated inammatory cytokines, dysregulated multi-organ damage biomarkers, decreased immune cell subsets were associated with higher proportion of death (34.17% vs 1.22%; P<0.001) and increased risk odds of death (OR=57.216, 95%CI=28.373-115.378; P<0.001) in COVID-19 patients with ARDS. In addition to previous reported risk factors related to ARDS development and death, such as neutrophils, IL-6, D-Dimer, leukocytes and platelet, we identied elevated TNF-α (OR=1.146, 95%CI=1.100-1.194; P<0.001), CK-MB (OR=1.350, 95%CI=1.180-1.545; P<0.001), declined ALB (OR=0.834, 95%CI=0.799-0.872; P<0.001), CD8 + T cells (OR=0.983, 95%CI=0.976-0.990; P<0.001) and CD3 - CD19 + B cells (OR=0.992, 95%CI=0.988-0.997; COVID-19, Corona virus disease 2019; CT, Computerized tomography; CD, Cluster of differentiation; NK cells, Natural killer cells; IL-6, Interleukin 6; IL-10, Interleukin 10; IL-8, Interleukin TNF-α, Tumor necrosis factor Interleukin transaminase; transpeptidase; dehydrogenase; brain natriuretic SIRS, The systemic inammatory response syndrome; ECMO, extracorporeal membrane oxygenation; Interquartile range; SD, standard deviation. Continuous variables were described as median (IQR) or mean and standard deviation (SD). P values were calculated by Mann-Whitney U non-parameter test for skewed distributed data or Student’s t-test for normal distributed data (b). Categorical variables were expressed as

We also observed dysregulated cardiac injury biomarkers, liver damage indexes and coagulation biomarkers were more pronounced in patients with ARDS. Our study showed that signi cantly  (Table 1). Compared with survivors, the same results can be found in non-survivors (    (Fig. 1).
Overall, we established a combined group by integrating four predictive scores, in ammatory-related indexes, immune cell subsets and multiple-organ damage biomarkers. The accuracy of combined score for predicting ARDS development was 0.904(0.866-0.942), and that for the death was 0.959(0.931-0.986). The predictive accuracy of the combined model demonstrated highest among all models in development of ARDS and progression from ARDS to death, indicating that the combined score system of great value in the evaluation of the development and death risk of ARDS in patients with COVID-19 (Fig. 1).

Discussion
In this retrospective cohort study, we found that COVID-19 patients with ARDS were more likely to develop into death. Moreover, patients with elevated in ammatory-related indexes, decreased immune cell subsets, abnormal multiple-organ damage biomarkers and higher scores were more likely to develop ARDS and progress from ARDS to death. Speci cally, in addition to previously reported risk factors such as neutrophils, IL-6, D-Dimer, leukocytes and platelet, we identi ed novel risk factors including elevated TNF-α, ALB and CK-MB as well as decreased CD8 + T cell and CD3 − CD19 + B cells. Furthermore, we established a comprehensive risk prediction model by combining risk factors and scores which showed good prediction accuracy for ARDS development (AUC = 0.904) and ARDS death risk (AUC = 0.959). This model will be bene cial for early identi cation of development of ARDS and determination of effective treatments in COVID-19 patients.
Enrolled 796 COVID-19 patients with ARDS were statistically matched by age, sex, and comorbidities to those without ARDS. We noted that patients with ARDS presented excessive in ammation, dysregulated immune response and critical multiple organ damage. Moreover, SOFA, qSOFA, APACHE II and SIRS scores appeared to be higher in COVID-19 patients with ARDS as well as nonsurvivors. These ndings provided supporting evidence that COVID-19 patients with ARDS were associated with higher risk of worse clinical outcomes.
Multivariable logistic regression models were performed to explore risk factors related to ARDS development and death in COVID-19 patients. In terms of in ammatory-related indexes, in addition to previous reported risk factors such as neutrophils, IL-6 and CRP, we found TNF-α was a novel risk factor of ARDS development and death in COVID-19 patients. Cytokine storm has been reported to mediate extensive lung in ammation and ultimately induce ARDS in COVID-19 patients through severe systemic secretion of proin ammatory cytokines. 17,18 IL-6 induces various pro-in ammatory cytokines and chemokines through the regulation of STAT3 and activation of the NF-κB pathway, which ultimately causes pneumocyte and endothelial injury, vascular leakage and alveolar edema in cytokine storm. [19][20][21] During pulmonary infections in COVID-19 patients, intracellular IL-6 acts in concert with CXCL1 and CXCL2 to recruit polymorphonuclear leukocytes to the lung, resulting in killing pathogens but generating brosis. 22 TNF-α, a potential novel risk factor for the development of ARDS identi ed here, has been reported to mediate airway hyper-responsiveness through the induction of M2 muscarinic receptor dysfunction. 23  are associated with the poor clinical outcomes in patients with autoimmune diseases and viral infectious disease. [31][32][33] In addition to elevated in ammatory-related indexes and decreased immune cell subsets, we found abnormal multiple-organ damage biomarkers were another risk factors for development of ARDS and progression from ARDS to death in COVID-19 patients.
Previously reported risk factors of COVID-19, such as ALT, AST, hs-cTnI, NT-proBNP, eGFR, D-Dimer and FDP were also con rmed in our study. Furthermore, we identi ed novel risk factors including CK-MB and ALB. The systemic in ammatory response and immune system disorders during disease progression in patients with COVID-19 result in the high incidence of organ failure symptoms and dysregulated tissue damage biomarkers. 34 Additionally, respiratory dysfunction and hypoxemia in patients with COVID-19 and ARDS caused damage to myocardial cells through the regulation of type 1 and type 2 T helper cells. 35 Moreover, The high serum levels of CK-MB suggested myocardial damage in early-stage which were correlated with the severity and case fatality rate of COVID-19. 36,37 ALB was reported to selectively suppress the expression of pro-in ammatory TNF-α by inhibiting the activation of NF-kB pathway. 38 Additionally, the decreased levels of ALB contribute to in ammatory response and oxidative stress injury through the upregulating the levels of glutathione, which partly accounted for disease severity in MERS patients. 38,39 However, it remains insu cient to explore the association between the multiple-organ damage and the pathogenesis of development of ARDS.
Previous researches have reported that SOFA, qSOFA, APACHE II, and SIRS scoring systems shows good prediction accuracy for evaluating septic shock, multi-organ failure and ICU mortality. 7,8,[10][11][12]40,41 Here, we con rmed that those with higher SOFA, qSOFA, APACHE II or SIRS scores were more likely to develop ARDS and progress to death in COVID-19 patients. Moreover, we found that elevated in ammatory-related indexes, decreased immune cell subsets and abnormal multiple-organ damage biomarkers were associated with ARDS development and progression from ARDS to death in COVID-19 patients. In addition, the patients with increased levels of the biomarkers were more likely to develop complications such as heart failure, acute cardiac, kidney, liver injury and DIC. Therefore, the combined model integrating risk factors and scores showed the best predictive capacities of both development of ARDS and progression from ARDS to death in COVID-19 patients.
It is critical to recommend standardized and effective treatment protocols for SARS-COV-2 infection worldwide to improve poor clinical outcomes. Consistent with other studies, antibiotics and antivirals were widely used in COVID-19 treatment. 13,42 Notably, COVID-19 patients with ARDS received more supportive therapy, such as ventilation treatments and glucocorticoid therapy.
Preliminary evidence suggested the standard supportive care for ARDS patients was a protective ventilatory strategy, which might improve prognoses of patients with ARDS. 43,44 Additionally, recent studies showed that corticosteroids could be an effective treatment. The administration of dexamethasone and methylprednisolone could decrease duration of mechanical ventilation and reduce the death risk of patients with COVID-19 who develop ARDS. 45,46 In view of the excessive in ammation, dysregulated immune response and multiple organ damage in COVID-19 patients with ARDS, the current management of COVID-19 should be focused on in ammatory immune response, treatment of complications and supportive care, especially oxygen support.
Based on our study, several management strategies are warranted for COVID-19 patients with ARDS. First, in ammatory-related indexes and organ damage indexes should be monitored at different stages to prevent the development of ARDS in COVID-19 patients. Second, early application of systemic immune modulators such as intravenous immunoglobulin should be considered to reduce aberrant immune responses at the early stage of ARDS, which is helpful to prevent the progression of ARDS. 47 Third, Supportive therapy such as ventilation treatment is also critical for COVID-19 patients with ARDS.
Our study has several limitations. First, this was a retrospective cohort study and not all laboratory tests were done in all patients.
The missing data might affect the interpretation of the results. Second, the potential mechanisms of multiple organ damage and ARDS development and progression to death need to be further investigated and con rmed. Third, our study was performed in single-center hospital, which need be further con rmed in multi-center studies in the future. Finally, given that age and comorbidities, such as diabetes and hypertension, have been reported as the common risk factors of COVID-19 with ARDS, these factors were statistically matched between patients with ARDS and those without ARDS and were not included in this study.

Conclusion
In summary, we found higher SOFA, qSOFA, APACHE II and SIRS scores, elevated in ammatory-indexes, decreased immune cell subsets and abnormal multiple-organ damage biomarkers related to higher death risk of COVID-19 with ARDS. Furthermore, we  Comprehensive prediction models for ARDS development and progression from ARDS to death in patients with COVID-19 a. Receiver operating characteristic (ROC) curves and area under the curve (AUC) were employed to assess the predictive accuracy of models evaluating the risk of ARDS development for COVID-19 patients with SOFA, qSOFA, APACHE II and SIRS scores, in ammatory-related indexes, immune cell subsets, organ damage indexes, and combined group integrating abovementioned these factors. The multivariate logistic regression model analysis was used to establish a risk model. The stepwise regression was used for the prediction selection for the model. ROC curves and AUCs (95%CIs) values were generated to assess prognostic accuracy for each model. A two-sided P value < 0.05 was considered statistically signi cant. b. Receiver operating characteristic (ROC) curves and area under the curve (AUC) were employed to assess the predictive accuracy of models evaluating the death risk of ARDS patients with SOFA, qSOFA, APACHE II and SIRS scores, in ammatory-related indexes, immune cell subsets, organ damage indexes, and combined group integrating abovementioned these factors. The multivariate logistic regression model analysis was used to establish a risk model. The stepwise regression was used for the prediction selection for the model. ROC curves and AUCs (95%CIs) values were generated to assess prognostic accuracy for each model. A two-sided P value < 0.05 was considered statistically signi cant.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.