4.1. Subject characteristics
Demographic characteristics of participants stratified by blood culture are summarized in Table 1. A total of 331 patients were eligible for this analysis. Depending on the results of blood culture, the patients were divided into the positive group and the negative group, 72 cases and 259 cases, respectively. Pulmonary infection in the positive group was less than that in the negative group, while hepatobiliary system infection and urinary tract infection were significantly more than those in the negative group. Patients with positive blood culture were the more likely to have an elevated PCT, lactate, CRP and a lower albumin. There was no difference in hospital stay and 28-day mortality between the two groups.
Table 1
Baseline characteristics of patients in this study according to blood culture results
Characteristic | Overall | Blood culture results | P Value |
negative | positive |
Patients | 331(100%) | 259(78.2%) | 72(21.8%) | |
Sex |
Female | 114(34.4%) | 86(26.0%) | 28(8.5%) | 0.18 |
Male | 217(65.6%) | 173(52.3%) | 44(13.3%) | 0.37 |
Age, median (IQR), y | 76(64–83) | 76(64–83) | 77(65–84) | 0.24 |
APACHE-II score median (IQR) | 20(15–24) | 19(14–24) | 20(15–25) | 0.37 |
Source of bacteremia |
Pulmonary | 114(34.4%) | 110(42.5%) | 4(5.6%) | < 0.01 |
Hepatobiliary system | 62(18.7%) | 38(14.7%) | 24(33.3%) | < 0.01 |
Abdominal | 24(7.3%) | 22(8.5%) | 2(2.8%) | 0.1 |
Urinary | 83(25.1%) | 46(17.8%) | 37(51.4%) | < 0.01 |
Skin and soft tissue | 22(6.6%) | 19(7.3%) | 3(4.2%) | 0.06 |
Unknown | 26(7.9%) | 24(9.3%) | 2(2.8%) | 0.07 |
Blood lactate ± SD (mg/dl) | 3.80 ± 3.05 | 3.68 ± 3.12 | 4.22 ± 2.77 | 0.18 |
WBC ± SD (10*9 /L) | 12.80 ± 6.91 | 12.53 ± 6.70 | 13.74 ± 7.60 | 0.19 |
C-reactive protein ± SD (mg/L) | 115.6 ± 96.78 | 108.42 ± 99.79 | 141.42 ± 80.53 | 0.01 |
Admission PCT ± SD (ng/ml) | 9.64 ± 17.57 | 7.82 ± 15.98 | 16.18 ± 21.22 | < 0.01 |
Reviewed PCT ± SD (ng/ml) | 18.08 ± 27.43 | 14.50 ± 24.86 | 30.96 ± 32.21 | < 0.01 |
IL-6 ± SD (Pg/ml) | 294.80 ± 870.56 | 265.16 ± 782.45 | 401.44 ± 1132.72 | 0.24 |
Pro-BNP ± SD (Pg/ml) | 1450.65 ± 2030.49 | 1514.07 ± 2163.35 | 1171.61 ± 1314.49 | 0.56 |
Albumin ± SD (g/L) | 31.23 ± 5.43 | 31.86 ± 5.36 | 29.00 ± 5.14 | < 0.01 |
CRP/ albumin ± SD | 4.0 ± 3.57 | 3.7 ± 3.64 | 5.06 ± 3.09 | 0.004 |
Glycosylated hemoglobin ± SD (%) | 6.67 ± 1.91 | 6.71 ± 1.94 | 6.54 ± 1.81 | 0.51 |
Mean days in ICU (IQR), d | 6(3–10) | 5(3–10) | 6(4–10) | 0.89 |
Mean days in hospital (IQR), d | 10(7–15) | 10(7–15) | 11(8–17) | 0.35 |
28-day mortality | 81(24.5%) | 66(25.5%) | 15(20.8%) | 0.42 |
4.2. Blood culture results.
The positive rate of blood culture was 21.8%. Since most of the pathogenic bacteria originate from the urinary tract(51.4%)and hepatobiliary system༈33.3%༉, the main pathogens cultivated were Escherichia coli (55.6%)and Klebsiella pneumoniae༈18.1%༉. The rests were Pseudomonas aeruginosa (5.6%), Staphylococcus aureus (4.2%), Enterobacter cloacae (4.2%) and some other pathogens (11.1%), as showed in Table 2. There were 9 cases of Staphylococcus epidermidis and Corynebacterium considered to be contaminants and excluded.
Table 2
Distribution of pathogens with positive blood culture
Organism | Number of isolates (%) |
Escherichia coli | 40(55.6) |
Klebsiella pneumoniae | 13(18.1) |
Pseudomonas aeruginosa | 4(5.6) |
Acinetobacter baumannii | 1(1.4) |
Staphylococcus aureus | 3(4.2) |
Enterobacter cloacae | 3(4.2) |
Others | 8(11.1) |
4.3. Univariate and multivariate analysis for discriminating BSI
Figure 1 showed the univariate logistic regression models between baseline variables and BSI. The univariate analysis indicated that admission PCT (OR = 1.023, 95% CI, 1.009–1.036, P = 0.001), review of PCT (OR = 1.019, 95% CI, 1.010–1.028, P < 0.0001), CAR (OR = 1.08, 95% CI, 1.01–1.16, P = 0.024), hepatobiliary system infection (OR = 17.37, 95% CI, 5.66–53.28, P < 0.001), urinary tract infection (OR = 22.12, 95% CI, 7.46–65.62, P < 0.0001) were positively correlated with the risk of BSI. Albumin (OR = 0.89,95% CI, 0.85–0.95, P < 0.0001) was negatively correlated with the risk of BSI.
Multivariate analysis results were presented in Fig. 2.The results also showed that reviewed PCT (OR = 1.01, 95% CI, 1.00-1.03, P = 0.007), hepatobiliary system infection (OR = 15.54, 95% CI, 4.87–49.55, P < 0.001) and urinary tract infection(OR = 19.34, 95% CI, 6.34–59.04, P < 0.001) (relative to lung infection) were independent risk factors for predicting BSI, but CAR and admission PCT were excluded(P > 0.05).
4.4 The ROC curve for BSI
We evaluated the predictive value of admission PCT, reviewed PCT, albumin, CRP and CAR alone using ROC analysis (Fig. 3). The area under the ROC curve for BSI was 0.714(0.662–0.762), 0.73(0.679–0.777), 0.669(0.615–0.719),0.6(0.545–0.653) and 0.62(0.565–0.672) for admission PCT, reviewed PCT, albumin, CRP and CAR, respectively. A prediction value of CAR > 5 was considered to a significant marker in predicting BSI (sensitivity:50%; specificity: 71.43%; PPV32.73%; NPV83.71%; P < 0.001).By comparing the AUC, admission PCT and reviewed PCT exhibited greater predictive significance compared with CRP or CAR. At a cut-off of 3.98 ng/ml, reviewed PCT offered the best accuracy in predicting BSI with the sensitivity, specificity, PPV, and NPV of 84.72%,54.83%, 34.08% and 92.76%; 3.97 ng/ml was the optimum cut-off of admission PCT, the sensitivity, specificity, PPV, and NPV were 66.7%, 69.1%,37.5% and 88.18%, respectively.
In order to explore the application value of multi-indicator joint prediction of BSI, We combined PCT with the site of infection and albumin through the Logistic analysis model to establish a regression equation to calculate the BSI probability comprehensively, and then used ROC analysis to calculate AUC. Through calculation, we found that reviewed PCT combined with the site of infection could more effectively predict the occurrence of BSI with AUC of 0.840(95%CI 0.796–0.878; sensitivity:84.72%; specificity:72.59%; PPV46.51%;NPV94.06%)(Fig. 4A). The combination of reviewed PCT, albumin and the site of infection maximally increased the predictive power with the highest AUC of 0.844(95%CI 0.801–0.882; sensitivity:87.50%; specificity:70.27%;PPV45%;NPV95.29%) (Fig. 4B).