Patient characteristics
The study included 214 adult COVID-19 patients, including 129 men and 85 women. The median age was 77 years old, with ages ranging from 65 to 87 years old. In contrast to the non-severe group, the severe group involved more males than females, older patients, longer hospital stays, and higher inflammatory markers. Additionally, co-morbidity was connected to COVID-19 severity. Initial characteristics of the two groups were presented in Supplementary Table 1.
Correlation between HBP, IL6, and CRP and organ function
Spearman rank correlation analysis suggested that HBP levels were significantly positively linked with total bilirubin (TBIL) (r = 0.153, P = 0.025), lactate dehydrogenase (LDH) (r = 0.337, P < 0.001), serum creatinine (Scr) (r = 0.240, P < 0.001), international normalized ratio (INR) (r = 0.346, P < 0.001), and D-dimer (r = 0.398, P < 0.001), and significantly negatively linked with albumin (ALB) (r = -0.430, P < 0.001). IL6 levels were significantly positively linked with TBIL (r = 0.195, P = 0.004), aspartate aminotransferase (AST) (r = 0.247, P < 0.001), LDH (r = 0.239, P < 0.001), Scr (r = 0.202, P = 0.003), INR (r = 0.284, P < 0.001), and D-dimer (r = 0.326, P < 0.001), and significantly negatively linked with ALB (r = -0.299, P < 0.001). CRP levels were significantly positively correlated with TBIL (r = 0.234, P = 0.001), AST (r = 0.246, P < 0.001), LDH (r = 0.530, P < 0.001), Scr (r = 0.398, P < 0.001), INR (r = 0.495, P < 0.001), and D-dimer (r = 0.439, P < 0.001), and significantly negatively linked with ALB (r = -0.476, P < 0.001) (Fig. 1A).
Comparison of HBP, IL6, and CRP between non-severe and severe groups
The severity group showed higher admission levels of HBP, IL6, and CRP than the non-severe group, and the difference was statistically significant (all P < 0.001, Fig. 1B-D).
Risk factors associated with severe COVID-19
Univariate analysis indicated that HBP, IL6, CRP, gender (male), age, hypertension, diabetes, chronic kidney disease, cardiovascular disease (CVD), and nervous system disease were significantly positively correlated with the occurrence of the severe group (Table 1). By using step-wise forward logistic regression analysis, the independent risk factors for the severe group were shown to be age (OR = 1.09, 95% CI: 1.05–1.13, P < 0.001), HBP (OR = 1.07, 95% CI: 1.04–1.11, P < 0.001), IL6 (OR = 1.01, 95% CI: 1.00-1.03, P = 0.029), and CRP (OR = 1.03, 95% CI: 1.01–1.04, P = 0.001).
Table 1
Univariate and multivariate logistic regression analysis of parameters related to the incidence of severe COVID-19
Variable | Non-severe | Severe | Univariate analysis | Multivariate analysis Model |
n = 93 | n = 121 | OR (95% CIs) | p value | OR (95% CIs) | p value |
Gender (male), n (%) | 45(48.39) | 84(69.42) | 2.42(1.38–4.25) | 0.002 | | |
Age (years), median (IQR) | 67(59–78) | 83(75–88) | 1.08(1.06–1.11) | < 0.001 | 1.08(1.03–1.13) | 0.002 |
Hypertension, n (%) | 39(41.94) | 76(62.80) | 2.34(1.35–4.07) | 0.003 | | |
Diabetes, n (%) | 16(17.20) | 38(31.40) | 2.20(1.13–4.27) | 0.019 | | |
Chronic kidney disease, n (%) | 15(16.12) | 47(38.84) | 3.30(1.70–6.41) | < 0.001 | | |
Cardiovascular diseases, n (%) | 26(27.96) | 62(51.24) | 2.71(1.52–4.82) | 0.001 | | |
Nervous system disease, n (%) | 8(8.60) | 37(30.58) | 4.68(2.06–10.64) | < 0.001 | | |
HBP (ng/mL), median (IQR) | 16.35(7.44–25.32) | 51.43(45.98-103.91) | 1.09(1.06–1.10) | < 0.001 | 1.08(1.04–1.11) | < 0.001 |
IL6 (pg/mL), median (IQR) | 24.09(4.26–81.23) | 70.10(32.97-112.47) | 1.02(1.01–1.02) | < 0.001 | 1.01(1.00-1.03) | 0.030 |
CRP (mg/L), median (IQR) | 16.63(6.03–44.07) | 87.64(52.03-141.01) | 1.03(1.02–1.04) | < 0.001 | 1.03(1.01–1.04) | 0.001 |
CIs: confidence intervals, OR: odds ratio, IQR: interquartile range. |
Combined detection of HBP, IL6, and CRP predicted the severity of COVID-19
The ROC curve was drawn using HBP, IL6, CRP, and the combined prediction probability value as test variables and the occurrence of the severe group as the state variable (Fig. 2). The results suggested that the area under the curve (AUC) predicted by HBP, IL6, CRP, and combined prediction probability values for the severe group were 0.895 (95% CI: 0.853–0.938), 0.710 (95% CI: 0.641–0.780), 0.869 (95% CI: 0.821–0.917), and 0.947 (95% CI: 0.919–0.976), respectively (Table 2). Accordingly, the combination of HBP, IL6, and CRP had a higher AUC value than HBP, IL6, and CRP alone, with a sensitivity of 85.10%, specificity of 95.70%, positive predictive value of 96.30%, and negative predictive value of 83.20%. When the Youden index was at its maximum, the cutoff values of HBP, IL6, and CRP were 49.71 ng/mL, 11.24 pg/mL, and 39.67 mg/L, respectively. Combined grouping of HBP, IL6, and CRP was conducted according to the optimal cutoff values: Group 1, HBP ≤ 49.71 ng/mL, IL6 ≤ 11.24 pg/mL, and CRP ≤ 39.67 mg/L; Group 2, only one of the three indicators exceeded the optimal cutoff value; Group 3, two or more indicators exceeded the optimal cutoff value.
Table 2
Analysis of receiver operating characteristic parameters to predict the severity of COVID-19
Parameters | AUC | 95% CIs | Cutoff | Sensitivity (%) | Specificity (%) | PPV (%) | NPV (%) | p value |
HBP | 0.895 | 0.853–0.938 | 49.71 | 75.2 | 97.0 | 96.8 | 75.0 | < 0.001 |
CRP | 0.869 | 0.821–0.917 | 39.67 | 85.1 | 74.2 | 81.1 | 79.3 | < 0.001 |
IL6 | 0.710 | 0.641–0.78 | 11.24 | 91.7 | 43.0 | 67.7 | 80.0 | < 0.001 |
HBP + IL6 + CRP | 0.947 | 0.919–0.976 | 0.62 | 85.1 | 95.7 | 96.3 | 83.2 | < 0.001 |
CIs: confidence intervals, PPV: positive predictive value, NPV: negative predictive value. |
Comparison of HBP, IL6, and CRP between non-organ failure and organ failure groups
The organ failure group showed higher admission levels of HBP, IL6, and CRP than the non-organ failure group, and the difference was statistically significant (all P < 0.001, Fig. 3).
Risk factors associated with organ failure
Univariate analysis revealed that HBP, IL6, CRP, HBP + IL6 + CRP combination, age, diabetes, chronic kidney disease, CVD, and nervous system disease are significantly positively correlated with the occurrence of organ failure (Table 3). After adjustment for age, diabetes, chronic kidney disease, CVD, and nervous system disease at baseline, with HBP, IL6, and CRP as independent variables (model 1). By using step-wise forward logistic regression analysis, the independent risk factors for organ failure were shown to be age (OR = 1.07, 95% CI: 1.03–1.11, P = 0.001), CVD (OR = 2.74, 95% CI: 1.13–6.14, P = 0.140, HBP (OR = 1.01, 95% CI: 1.00-1.02, P = 0.018), IL6 (OR = 1.01, 95% CI: 1.00-1.01, P = 0.005), and CRP (OR = 1.02, 95% CI: 1.01–1.02, P < 0.001). When the baseline remained unchanged, with HBP + IL6 + CRP combination as the independent variable (model 2), the independent risk factors for organ failure were shown to be age (OR = 1.05, 95% CI: 1.02–1.08, P = 0.003), CVD (OR = 2.34, 95% CI: 1.14–4.80, P = 0.021), and HBP + IL6 + CRP combination (P < 0.001). Group 3 had the highest risk of organ failure (3 vs 1, OR = 9.47, 95% CI: 1.99–45.13, P = 0.005; 3 vs 2, OR = 10.00, 95% CI: 3.50-28.54, P < 0.001).
Table 3
Univariate and multivariate logistic regression analysis of parameters related to the incidence of organ failure
Variable | Non-organ failure | Organ failure | Univariate analysis | Multivariate analysis Model 1 | Multivariate analysis Model 2 |
n = 137 | n = 77 | OR (95% CIs) | p value | OR (95% CIs) | p value | OR (95% CIs) | p value |
Age (years), median (IQR) | 71(61–84) | 83(77–89) | 1.08(1.04–1.10) | < 0.001 | 1.07(1.03–1.11) | 0.001 | 1.05(1.02–1.08) | 0.003 |
Diabetes, n (%) | 26(17.20) | 28(31.40) | 2.44(1.30–4.58) | 0.006 | | | | |
Chronic kidney disease, n (%) | 27(18.98) | 35(45.45) | 3.40(1.84–6.28) | < 0.001 | | | | |
Cardiovascular disease, n (%) | 41(29.93) | 47(61.04) | 2.67(2.04–6.59) | 0.001 | 2.74(1.13–6.14) | 0.140 | 2.34(1.14–4.80) | 0.021 |
Nervous system disease, n (%) | 21(15.33) | 24(31.17) | 2.50(1.28–4.89) | 0.007 | | | | |
HBP (ng/mL), median (IQR) | 21.00(11.97–51.43) | 51.43(46.81-101.04) | 1.02(1.01–1.02) | < 0.001 | 1.01(1.00-1.02) | 0.018 | | |
IL6 (pg/mL), median (IQR) | 40.88(8.13–81.23) | 71.57(35.68-181.72) | 1.01(1.01–1.02) | < 0.001 | 1.01(1.00-1.01) | 0.005 | | |
CRP (mg/L), median (IQR) | 30.09(10.15-71.00) | 101.63(55.79-147.93) | 1.02(1.01–1.02) | < 0.001 | 1.02(1.01–1.02) | < 0.001 | | |
HBP + IL6 + CRP | | | | < 0.001 | | | | < 0.001 |
2vs1 | | | | | | | | |
3vs1 | | | 16.29(3.72–71.42) | < 0.001 | | | 9.47(1.99–45.13) | 0.005 |
3vs2 | | | 12.55(4.70-33.49) | < 0.001 | | | 10.00(3.50-28.54) | < 0.001 |
Model 1 was adjusted for age, diabetes, chronic kidney disease, cardiovascular disease, and nervous system disease at baseline, with HBP, IL6, and CRP as independent variables. |
Model 2: the baseline remained unchanged, with the HBP + IL6 + CRP combination as the independent variable. |
CIs: confidence intervals, OR: odds ratio, IQR: interquartile range. |
Comparison of HBP, IL6, and CRP between survivor and non-survivor groups
The non-survivor group had considerably higher admission levels of HBP, IL6, and CRP than the survivor group, with statistically significant differences. (all P < 0.001, Fig. 4).
Kaplan-Meier survival curves for HBP, IL6, CRP, and HBP + IL6 + CRP combination
KM survival curves revealed that HBP (hazard ratio (HR) = 2.497, log-rank P < 0.001), IL6 (HR = 3.437, log-rank P = 0.018), and CRP (HR = 3.631, log-rank P = 0.003) greater than the optimal threshold were correlated with a substantially increased risk of mortality (Fig. 5). Group 3 had a greater mortality risk than Group 1 did (3 vs 2, HR = 3.631, log-rank P = 0.003).