Background: Coagulase-negative staphylococci (CoNS) are one of the most common contaminant microorganisms isolated from blood cultures (BCs). Methods: We conducted a retrospective cohort study at St Luke`s International Hospital from 2004 to 2017. We collected a total of 1192 BCs due to CoNS. Of 1192 BCs, 143 pts with polymicrobial infection and 112 pts who were <18 years old were excluded. We defined the true infection (TI) on the following criteria; 1) patients with persistent bacteremia due to CoNS, 2) BCs that were positive more than 2 sets, 3) patients with foreign body implanted 28 days before taking BCs. Chi-square test, Fisherfs exact test and Student t test were used in univariate analysis and logistic regression analysis was used in multivariable analysis. Results: Of 927 patientsf BCs data, 201 pts were defined as TI, and 726 pts were defined as contamination. BCs collection location (general ward), inpatient, malignant tumor, steroids or chemotherapy use 90 days before collection, presence of central venous catheter (CVC), mortality in 90 days after collection, time to positivity (TTP : 24-48h, and <24h) differed significantly between these two groups. Based on multivariable analysis, TTP : 24-48h (adjusted odds ratio (OR) 1.93, p=0.019, 95% confidential interval (CI) 1.11-3.36), and < 24h : (OR 4.83, p<0.001, 95% CI 2.68-8.68), BCs collection location : general ward (OR 2.21, p<0.001, 95% CI 1.58-3.10), and presence of CVC (OR 1.91, p<0.001, 95% CI 1.37-2.66) were identified as independent risk factors of TI. Based on the estimates of OR, we assigned a score to each factors (Table 1). Area under the curve was 0.702 (95% CI 0.662 – 0.741) (Figure 1). ROC analysis showed that, if a patientfs score is 0 point, TI can be excluded with sensitivity: 98%. On the other hands, if a patientfs score is 7 points or more, TI can be diagnosed with specificity: 87%. Conclusion: Time to positivity, location of blood collection and the presence of CVC were associated with TI. The prediction rule developed in this study can be useful for clinicians for making decision whether to use antibiotics or not.
Table 1. Prediction rule of CNS true infection.
Figure 1. ROC analysis of prediction rule.
M. Rahman, None
N. Mori, None