Baseline, laboratory and imaging characteristics
In this retrospective study, a total of 432 patients with COVID-19 were enrolled, including 202(46.5%) women and 230(53%) men, the average age was 52.88 years. Fever (308, 71.3%), cough (270, 62.5%), expectoration (130, 30.1%) and fatigue (128, 29.6%) were the common symptoms. Hypertension (92, 21.3%) was the most common comorbidity.
The patients were divided into two groups: severe group (125/432, 28.94%) and non-severe (307/432, 71.06%) group based on the severity of the disease. Comparing with the non-severe group, in terms of the baseline characteristics, the average age of severe group was older (59.60±16.65 years vs 50.14±16.26 years, p<0.0001). Meanwhile, it is noticed that severe group has higher incidence of comorbidities, such as hypertension (p<0.0001), diabetes (p<0.0001), and COPD (p=0.009). As for the clinical laboratory findings, lower level of lymphocyte (p<0.0001) and higher level of white blood cell (p=0.023), neutrophil (p<0.0001), C-reaction protein (p<0.0001), LDH (p<0.0001), D-dimer (p<0.0001) and NLR (p<0.0001) were detected in severe group as compared with non-severe group. Regarding CT results, 96.0% (120/125) patients had bilateral lung involvement, 32% (40/125) consolidation, and 3.2% (4/125) pleural effusion among patients in severe group. There was significant difference in CT score (6.0 [4-9] for severe group vs 6[4-7] for non-severe group, p<0.0001) between two groups (Table 1).
Predictive value of NLR, LDH, D-dimer and CT score
As table 1 showed, NLR, LDH, D-dimer, and CT score were statistically significantly higher in the severe group. On the basis of receiver operating characteristic (ROC), the area under curve (AUC) was 0.716 for NLR, 0.740 for LDH, 0.650 for D-dimer, and 0.612 for CT score, indicating certain diagnostic value for the severity of disease (Figure 1 and Table 2). In addition, the optimum cutoff from ROC was 3.82, 246 U/L, 0.83μg/ml, and 7 for NLR, LDH, D-dimer, and CT score, respectively (Table 2).
We assumed that when the level of NLR, LDH, D-dimer, and CT score on admission exceeded the optimum cutoff, the patients were prone to develop severe or critical types. According to optimum cutoff, the patients were divided into different subgroups.
As table 3 showed, about 25.9% (112/432), 31.9% (138/432), 22.2% (96/432) and 25% (108/432) patient, respectively, had high level of NLR, LDH, D-dimer and CT score on admission. After grouping, the distribution of baseline NLR [63/125 (50.4%) vs 49/307(16%); p<0.0001], LDH [74/125(59.2%) vs. 64/307(20.8%); p<0.0001]; D-dimer[47/125 (37.6%) vs 49/307(16%); p<0.0001] and CT score [46/125 (36.8%) vs 62/307 (20.2%); p<0.0001] over optimum cutoff in two groups were significantly significant (Table 3).
Univariate analysis indicated that the high level of NLR, LDH, D-dimer and CT score positively correlated with the severity of disease (OR=5.350; 95%CI=3.361-8.518; p<0.0001 for NLR; OR=5.509;95%CI=3.511-8.646;p<0.0001 for LDH; OR=3.173; 95%CI=1.976-5.094; p<0.0001 for D-dimer; OR=2.301; 95%CI=1.455-3.638; p<0.0001 for CT score). However, after adjusting other statistically significant index, the predictive value of NLR>3.82, LDH>246U/L were persisted (OR=2.163; 95%CI=1.162-4.026; p=0.015 for NLR; OR=2.298;95%CI=1.327-3.979; p=0.003 for LDH). While the relationship between D-dimer> 0.83μg/ml, CT score>7 and the severity of disease was weakened (OR=1.209; 95%CI=0.626-2.334; p=0.571 for D-dimer; OR=1.519;95%CI=0.71-3.247; p=0.281 for CT score). In addition, fatigue (OR=1.978;95%CI=1.127-3.473; p=0.018), chest tightness (OR=2.265; 95%CI= 1.011-5.074; p=0.047), hypertension (OR=2.534, 95%CI=1.259-5.099; p=0.009), CRP (OR=1.013; 95%CI= 1.003-1.023; p=0.011), bilateral lung involved(OR=3.890; 95%CI=1.356-11.154;p=0.011) were still positively correlated with the severity of disease (Table 4).
Evaluation of multi-parameter model
According to the logistic regression, NLR>3.82 and LDH>246U/L were statistically significant risk factors (Table 4). And as table 2 showed, the sensitivity of NLR>3.82 and LDH>246U/L in predicting the severity of COVID-19 were 50.40% and 59.20%, respectively. Then, further evaluation was made to judge whether the combined diagnosis model of two indexes can improve the sensitivity of prediction.
The table 5 indicated that the combined diagnosis of NLR>3.82 and LDH>246U/L could increase the sensitivity in predicting the severity of disease [NLR>3.82(50.40%) vs combined diagnosis model (72.80%); p =0.0007; LDH>246(59.2%) vs combined diagnosis model (72.80%); p<0.0001].