Clinical characteristics and laboratory findings
796 (51.99%) COVID-19 patients with ARDS and 735 (48.01%) COVID-19 patients without ARDS were statistically matched by age, sex, and comorbidities in this study. The COVID-19 patients with ARDS were further classified into survivors (65.83%) and non-survivors (34.17%). Expectedly, COVID-19 patients with ARDS were more prevalent to be admitted to the intensive care unit (ICU) (19.24% vs 0.42%; P<0.001) and had a higher proportion of death (34.17% vs 1.22%; P<0.001) than those without ARDS (Table 1).
As shown in Table 1, symptoms such as fever (84.73% vs 61.23%; P<0.001), dyspnea (43.51% vs 35.63%; P=0.003) , fatigue(18.78% vs 14.22%; P=0.022), and vertigo (5.00% vs 2.25%; P=0.006) were more prevalent in COVID-19 patients with ARDS. Fever (89.52% vs 82.32%; P=0.010), dyspnea (50.00% vs 40.20%; P=0.012) and vertigo (7.30% vs 3.90%; P=0.045) were more pronounced in non-survivors (Table2). Moreover, patchy shadows and bilateral pulmonary involvement were found in majority of COVID-19 patients, while the difference of patchy shadows was not statistically significant between patients with and without ARDS (Table 1).
We observed that the levels of inflammatory cytokines were substantially elevated in patients with ARDS, including IL-6 (13.24 vs 3.11 pg/mL; P<0.001),hs-CRP (47.80 vs 5.20 pg/mL; P<0.001), TNF-α (8.30 vs 7.30 pg/mL; P<0.001). Conversely, the counts of immune cell subsets, including CD3+CD19- T cells (845.00 vs 1073.00 /μL; P<0.001), CD8+ T cells (279.00 vs 344.00 /μL; P <0.001), CD3-CD16+CD56+ NK cells (138.00 vs 192.00/μL; P<0.001) as well as the total number of T cells, B cells and NK cells (1194.00 vs 1545.00 /μL; P<0.001), were significantly decreased in the patients with ARDS (Table 1).
We also observed dysregulated cardiac injury biomarkers, liver damage indexes and coagulation biomarkers were more pronounced in patients with ARDS. Our study showed that significantly increased CK-MB (0.80 vs 0.70 U/L; P<0.001), creatinine (72.50 vs 70.00 μmol/L; P=0.009), D-Dimer (1.12 vs 0.43 ug/mL; P<0.001) and FDP (5.20 vs 4.00 g/L; P<0.001), while decreased ALB (34.30 vs 38.60 g/L; P<0.001) presented in ARDS group (Table 1). Compared with survivors, the same results can be found in non-survivors (Table 2). Other laboratory indexes in patients with and without ARDS, as well as non-survivors and survivors, were presented in Supplementary Table 1 [see Additional file 2] and Supplementary Table 2 [see Additional file 3]. These findings indicated that aggravated inflammatory responses, lymphopenia, and multiple-organ damage might be correlated with the development of ARDS and subsequently worse clinical outcomes.
Comorbidities, complications, and clinical treatments
Consistent with previous studies, COVID-19 patients with ARDS had an increased risk of developing other serious complications, more patients with ARDS presented with acute liver injury (9.92% vs 1.36%; P<0.001), acute kidney injury (18.22% vs 2.31%; P<0.001), heart failure (27.89% vs 2.31%; P<0.001), cardiac injury (20.40% vs 5.71%; P<0.001), and disseminated intravascular coagulation (DIC), 6.91% vs 0.27%; P<0.001) (see Additional file 2). Similar results were found in the comparison of non-survivors and survivors in COVID-19 patients with ARDS (see Additional file 3).
In terms of clinical treatments, more patients with ARDS frequently received antibiotic treatment (85.55% vs 65.31%), antiviral therapy (54.52% vs 42.59%), glucocorticoid therapy (63.44% vs 23.95%), and intravenous immunoglobulin therapy (41.33% vs 18.78%; P<0.001). Remarkably, patients with ARDS required more clinical oxygen support, including high-flow oxygen ventilation (55.53% vs 41.50%; P<0.001) and mechanical ventilation (23.87% vs 2.04%; P<0.001) (see Additional file 2). Besides, except for antiviral therapy and high-flow oxygen therapy, all forms of clinical treatments were given more in non-survivors (see Additional file 3).
Risk factors
Multivariate Logistic models were calculated to explore risk factors for development of ARDS and death among COVID-19 patients with adjustment of age, sex, and comorbidities. (Table 3). We observed that elevated inflammatory cytokines were substantially associated with the higher risk for disease progression of ARDS and death in COVID-19 patients, such as IL-6 (OR=1.021, 95%CI=1.016-1.026; P<0.001), TNF-α (OR=1.146, 95%CI=1.100-1.194; P<0.001), IL-10 (OR=1.126, 95%CI=1.087-1.166; P<0.001), and hs-CRP (OR=1.016, 95%CI=1.013-1.019; P<0.001). Conversely, the decreased level of counts of immune cell subsets, such as lymphocytes (OR=0.365, 95%CI=0.244-0.545; P<0.001), CD8+ T cells (OR=0.983, 95%CI=0.976-0.990; P<0.001), CD3-CD19+ B cells (OR=0.992, 95%CI=0.988-0.997; P=0.003) were significantly related to the lower risk of the progression from ARDS to death. Additionally, we found that multiple-organ damage biomarkers, including decreased ALB (OR=0.834, 95%CI=0.799-0.872; P<0.001), elevated AST (OR=1.004, 95%CI=1.001-1.007; P=0.017), CK-MB (OR=1.350, 95%CI=1.180-1.545; P<0.001), creatinine (OR=1.007, 95%CI=1.003-1.012; P<0.002), D-Dimer (OR=1.272, 95%CI=1.207-1.340; P<0.001) and FDP (OR=1.052, 95%CI=1.038-1.066; P<0.001), might associate with high risk of progression of ARDS and outcome of COVID-19 patients.
Comprehensive prediction models
All four predictive score systems, including SOFA, qSOFA, APACHE II and SIRS, performed good prediction capacities for assessing the development risk of ARDS among COVID-19 patients and death risk among patients who developed ARDS. AUCs(95%CI) of SOFA, qSOFA, APACHE II and SIRS scores in assessing the progression of ARDS in COVID-19 patients were 0.857(0.820-0.894), 0.701(0.659-0.744), 0.729(0.679-0.780) and 0.672(0.618-0.725) (Figure 1a). Furthermore, all of them had prominent prediction capacities evaluating death risk. AUCs (95%CI) of SOFA, qSOFA, APACHE II and SIRS scores were 0.853(0.807-0.900), 0.743(0.686-0.799), 0.907(0.870-0.944) and 0.693(0.627-0.758) (Figure 1b).
Furthermore, based on inflammatory-related indexes, multiple-organ damage biomarkers and immune cell subsets, we constructed another three prediction models, all of which were significantly associated with the ARDS development and death. Discrimination of ARDS development risk models was better using multiple-organ damage biomarkers (AUC=0.779, 95%CI=0.746-0.812). The AUCs of inflammatory-related indexes model and immune cell subsets model were 0.729(0.701-0.757) and 0.687(0.629-0.746) (Figure 1a). Conversely, the highest predictive capacities of three prediction models of death was the immune cell subsets group model (AUC=0.954, 95%CI=0.924-0.985) (Figure 1b). AUCs of multiple-organ damage biomarkers model and inflammatory-related indexes model were 0.927(0.901-0.952), and 0.899(0.874-0.924).
Overall, we established a combined group by integrating four predictive scores, inflammatory-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 is potential of great value in the evaluation of the development and death risk of ARDS in patients with COVID-19 (Figure 1).