How to discriminate contamination from bloodstream infection due to coagulase-negative staphylococci: a prospective study with 654 patients

Authors


Corresponding author: A. F. Widmer, Division of Infectious Diseases and Hospital Epidemiology, University Hospital Basel, Petersgraben 4, 4031 Basel, Switzerland
E-mail:widmera@uhbs.ch

Abstract

Clin Microbiol Infect 2012; 18: E355–E361

Abstract

Coagulase-negative staphylococci (CoNS) are frequent contaminants of blood cultures. We aimed to evaluate the systemic inflammatory response syndrome (SIRS) criteria in patients with CoNS bacteraemia for discrimination between true bloodstream infection (BSI) and contamination. Prospective evaluation was carried out of clinical and laboratory parameters in adults with at least one positive blood culture with CoNS at the University Hospital of Basel between 2003 and 2007. Of 3060 positive blood cultures, 654 episodes of CoNS bacteraemia were identified. Of these, 232 (35%) were considered to be true BSI and 422 (65%) were considered to be contamination. Overall, 80% of study participants had at least one SIRS criterion, fever being the most common, and 49% had at least two SIRS criteria. In the multivariate analysis, independent predictors of BSI were fever or hypothermia (OR 2.93, 95% CI 1.91–4.5), tachycardia (OR 2.29, 95% CI 1.50–3.50), tachypnoea (OR 2.4, 95% CI 1.30–4.43), leucocytosis or leucopenia (OR 4.15, 95% CI 2.17–6.36) and the presence of a central venous line (OR 5.38, 95% CI 3.25–8.88). The probability of BSI increased with each additional SIRS criterion, ranging from 42.4% in patients with only one SIRS criterion to 56.7% for those with two criteria, and 72.3% for patients with three SIRS criteria. A positive blood culture with CoNS most likely represents true BSI if the patient has at least three SIRS criteria or two SIRS criteria and a central venous catheter. These simple bedside criteria may guide decision to treat, decreasing the use of glycopeptides.

Introduction

Coagulase-negative staphylococci (CoNS) are among the most frequently isolated pathogens in blood cultures and an important cause of nosocomial bloodstream infections (BSI) [1,2]. As ubiquitous skin commensals, CoNS are also the most common contaminants of blood cultures [3]. In clinical practice, it is important to distinguish between contamination and BSI [4,5] to prevent unnecessary prescription of antimicrobial agents leading to a selection of antimicrobial-resistant organisms such as vancomycin-resistant enterococci [6], longer hospitalization and increased costs [7,8]. The clinical relevance of a single blood culture positive for CoNS is difficult to assess, mainly because of the lack of a diagnostic reference standard for BSI [9]. Contamination is generally presumed if only one of at least two sets of blood cultures is positive for CoNS, whereas true BSI is assumed if at least two blood cultures yield CoNS [10–12]. However, several studies reported that about one-third of patients with true BSI had only one positive blood culture [13,14], and that contamination was possible even if two or more sets were positive [10,13]. In the literature, various clinical or laboratory definitions have been proposed to determine the clinical relevance of bacteraemia [15–20]. Of these, the CDC definition of a primary BSI [7] is the most commonly used, requiring clinical evidence of an infection plus an appropriate antibiotic therapy if an intravenous catheter is present; or at least two positive blood cultures. This definition was found to have a sensitivity of 67%, specificity of 56% and a positive predictive value of 31% [16]. Among laboratory-based methods, molecular genotyping such as pulsed-field gel electrophoresis and evaluation of biofilm-forming properties in CoNS were investigated as predictors of true BSI [10,21]. However, these laboratory methods are too time-consuming to guide a rapid start of an appropriate antimicrobial therapy in case of BSI. Moreover, molecular typing of CoNS from blood cultures did not seem to correlate with clinical criteria for BSI.[22] Clinical diagnosis of BSI relies on the presence of at least two of four systemic inflammatory response syndrome (SIRS) criteria. However, SIRS criteria have not been validated in the setting of bacteraemia with CoNS [23]. The aim of this study was to evaluate predictors of BSI and the usefulness of the SIRS criteria in patients with CoNS bacteraemia to discriminate between true BSI and contamination.

Methods

Study population and design

All adults with at least one positive blood culture with CoNS between 1 January 2003 and 31 December 2007 at the University Hospital Basel were eligible for this study. We excluded patients with neutropenia (leucocyte count <4 G/L) and those with growth of at least two different microorganisms in the same blood culture or in at least two separate blood cultures within 48 h [24,25]. The University Hospital Basel is a tertiary-care 800-bed teaching institution serving the northwestern part of Switzerland with a population of approximately half a million people, with about 31 000 admissions and >17 000 blood cultures taken annually. Clinical and laboratory data were prospectively collected by infection control practitioners using a standardized case report form. Data routinely collected included demographic characteristics, co-morbidities (diabetes mellitus, immunodeficiency, alcohol consumption, injecting drug use), hospital ward and clinical presentation including temperature, heart and respiratory rate. Immunodeficiency was defined as immunosuppressive treatment (immunosuppressant drugs, chemotherapy or prednisone with a dosage of at least 25 mg daily) or impaired immune system such as HIV infection. Further predisposing conditions recorded were surgical procedures, invasive ventilation and the presence of devices (i.e. central venous catheter, arterial catheters or peripheral lines) or implants. Blood tests included leucocyte count, C-reactive protein and creatinine. One episode of CoNS bacteraemia was defined as growth of CoNS in one or more blood cultures collected within 1 week. Nosocomial infection was defined as occurrence of BSI with CoNS after more than 48 h of hospitalization. Only the first episode was counted if there were multiple episodes for the same patient.

Microbiological methods

The identification of blood isolates was performed using the automated BacT/Alert 3D blood culture system (Organon Teknika Corp., Durham, NC, USA) [26], where samples were incubated for at least 7 days. Microbiology susceptibility tests were performed according to CLSI guidelines. Detection of CoNS relied on non-invasive colorimetric measurement of CO2 produced by growing microorganisms and was followed by Gram staining and subcultures of positive samples. The final identification of isolates was performed by standard procedures as previously described [27].

Blood stream infection

Diagnosis of BSI was assessed by a Fellow in infectious diseases and confirmed by a board-certified infectious diseases specialist in patients with at least one blood culture positive for CoNS based on the clinical presentation without an apparent infection at another site. SIRS was diagnosed if at least two of the following criteria were fulfilled: temperature >38°C or <36°C; heart rate >90/min; respiratory rate >20/min; leucocyte count >12 or <4 G/L or >10% immature neutrophil granulocytes [23].

Statistical analysis

Basic demographic characteristics, co-morbidities, clinical and laboratory parameters including SIRS criteria, and antibiotic therapy were compared according to the presence of BSI as evaluated by an infectious diseases specialist using the chi-square test or Fisher’s exact test for categorical variables and the Mann–Whitney U test for continuous variables. Logistic regression was used to estimate the predictors of true BSI in patients with CoNS bacteraemia.

All analyses were performed using STATA™ software version 11 for Windows (Stata Corp., College Station, TX, USA).

Ethical approval

The study was approved by the local ethical committee as part of the continuous quality assurance programme of the University Hospital Basel.

Results

Study population

A total of 2705 patients had positive blood cultures during the study period. Among them, we identified 676 patients with at least one positive blood culture with CoNS between 2003 and 2007 at the University Hospital Basel. Of these, 22 patients were excluded because of neutropenia. The final analysis was performed on 654 patients. Baseline characteristics are summarized in Table 1.

Table 1.   Baseline characteristics of the study population (n = 654) according to the clinical relevance of one single blood culture positive for coagulase-negative staphylococci
CharacteristicBloodstream infection (n = 232)Contamination (n = 422)p value
n (%)a n (%)a
  1. a Data given as n (%) except where indicated as median (IQR; interquartile range).

  2. b Bold values indicate p values that are statistically significant (i.e. below 0.05).

Median age, IQR 65 (49–74) 65 (49–75)0.427
Male gender140 (60)274 (59)0.652
Diabetes mellitus 45 (19) 77 (18)0.718
Alcohol use 32 (14) 40 (9)0.092
Intravenous drug use 17 (7) 30 (7)0.918
HIV infection 8 (3) 11 (3)0.540
Immunodeficiency 36 (16) 71 (17)0.665
Treatment with corticosteroids  6 (3) 17 (4)0.338
Ward
 Emergency room 43 (27)114 (73) 0.005
 Surgery 54 (45) 66 (55)
 Intensive-care unit 65 (41) 93 (59)
 Medicine 70 (32)149 (68)
Central venous catheter153 (67)121 (31) <0.001
Implant or device 54 (23) 65 (15)0.013
Artificial invasive ventilation107 (46)107 (25) <0.001
Fever (>38°C) or hypothermia (<36°C)154 (66)157 (37) <0.001
Heart rate >90/min145 (63)145 (34) <0.001
Respiratory rate >20/min 52 (22) 35 (8) <0.001
Leucocytes >12 or <4 G/L154 (66)142 (34) <0.001
Median C-reactive protein, IQR (mg/L)118 (58–196) 81 (31–147) <0.001
Median number of positive blood cultures within 1 week, IQR 2 (1–4) 1 (1–2) <0.001
Median number of blood cultures taken within 1 week, IQR  6 (4–8)  4 (4–8) <0.001

Predictors of bloodstream infection

Bacteraemia with CoNS was classified as BSI in 232 patients (35%) and contamination in 422 patients (65%). CoNS was isolated in more than one blood culture in 208 patients (32%), more frequently in case of BSI than contamination (68% vs. 12%, p <0.001).

Overall, 80% of study participants had at least one SIRS criterion, fever being the most common, and 49% had at least two SIRS criteria. Conversely, only 5% of patients with BSI compared with 29% of those with contamination (p <0.001) had no SIRS criteria when blood cultures were taken. BSI was more frequently diagnosed in patients with an implant or device (23% vs. 15%, p 0.013, Table 1). Patients with an implant or device were older (p <0.001) and less frequently injecting drug users (p <0.001), but no differences in the clinical presentation (SIRS criteria), C-reactive protein and central venous catheter were noted between both groups. According to susceptibility tests, 287 (45%) of CoNS isolates were resistant to oxacillin, more frequently in BSI (55% vs. 40%, p <0.001). Among 313 (48%) patients started on antibiotic treatment (β-lactam antibiotics or vancomycin), 161 had true BSI and 152 had contamination with CoNS, corresponding to a treatment rate of 69% for BSI and 36% for contamination (p <0.001). Compared with patients with contamination due to CoNS, those with BSI remained in hospital longer (median duration of 22 days, interquartile range (IQR) 11–39 days vs. 13 days, 6–24 days, p <0.001) and in the intensive-care unit (median stay of 9 days, IQR 4–21 days vs. 6 days, IQR 3–11, p <0.001). Death from all causes occurred more frequently in patients with true CoNS BSI compared with those with contamination (18% vs. 9%, p 0.001).

Predictors of BSI (Table 2) in univariate analysis were an increasing number of positive blood cultures, central venous catheter, stay in a surgery ward or intensive-care unit, invasive ventilation, carrying an implant or device, SIRS criteria and C-reactive protein. In multivariate analysis, a single blood culture positive for CoNS was likely to represent BSI if a central venous line (OR 5.38, 95% CI 3.25–8.88), fever or hypothermia (OR 2.93, 95% CI 1.91–4.5), tachycardia (OR 2.29, 95% CI 1.50–3.50), tachypnoea (OR 2.4, 95% CI 1.30–4.43) and leucocytosis or leucopenia (OR 4.15, 95% CI 2.17–6.36) were present.

Table 2.   Predictors of bloodstream infection, univariate and multivariate OR
CharacteristicUnivariate analysisMultivariate analysis
OR95% CIp valueORa95% CIp value
  1. aAdjusted for all variables listed in the Table.

  2. bReference.

Age, per 10 years increase0.970.88–1.060.4711.040.91–1.190.555
Female gender0.730.67–1.290.6521.080.70–1.660.735
Diabetes mellitus1.080.72–1.620.7180.880.52–1.500.640
Alcohol use1.530.93–2.510.0930.460.74–2.860.275
Intravenous drug use1.030.56–1.920.9180.890.38–2.090.784
Immunodeficiency0.910.56–1.410.6660.650.37–1.150.140
Ward
 Emergency room1b 1b
 Surgery2.171.31–3.58 0.003 1.140.56–2.320.720
 Intensive-care unit1.851.15–2.97 0.011 0.570.26–1.250.159
 Medicine1.250.79–1.960.3400.910.52–1.600.740
Central venous catheter4.523.19–6.39 <0.001 5.383.25–8.88 <0.001
Device or implant1.671.11–2.49 0.013 1.490.89–2.520.129
Artificial ventilation2.521.80–3.54 <0.001 1.140.62–2.110.674
Fever (>38°C) or hypothermia (<36°C)3.332.38–4.67 <0.001 2.931.91–4.50 <0.001
Heart rate >90/min3.182.28–4.44 <0.001 2.291.50–3.50 <0.001
Respiratory rate >20/min3.192.01–5.08 <0.001 2.401.30–4.43 0.005
Leucocytes >12 or <4 G/L3.892.77–5.46 <0.001 4.152.71–6.36 <0.001
C-reactive protein, per 50 mg/L increase1.191.08–1.30 <0.001 0.990.89–1.100.828

Algorithm to determine the clinical relevance of a single positive blood culture with CoNS

The probability of BSI increased with each additional SIRS criterion (test for trend p <0.001), ranging from 42.4% in patients with only one SIRS criterion to 82.6% for those with four SIRS criteria. Among patients with two SIRS criteria, the positive predictive value of BSI increased from 56.7% to 73.4% if there was a central venous catheter (Table 3).

Table 3.   Validity of a single positive blood culture for coagulase-negative staphylococcis as true bloodstream infection according to different diagnostic settings
SettingSensitivity (%)Specificity (%)PPV (%)NPV (%)Accuracy (%)
  1. CVC, central venous catheter; NPV, negative predictive value; PVP, positive predictive value; SIRS, systemic inflammatory response syndrome.

≥1 SIRS criterion95.3%28.9%42.4%91.7%52.4%
≥2 SIRS criteria78.0%67.3%56.7%84.8%71.1%
≥3 SIRS criteria36.2%91.2%69.4%72.2%71.7%
4 SIRS criteria8.2%99.1%82.6%66.2%66.8%
≥1 SIRS criterion and a CVC63.9%77.5%62.2%78.7%72.5%
≥2 SIRS criteria and a CVC51.7%89.1%73.4%76.1%75.4%
≥3 SIRS criteria and a CVC25.6%96.9%83.0%69.2%70.8%

The algorithm with the best combined sensitivity and specificity for CoNS BSI was defined as at least two positive blood cultures within 7 days, or one single positive blood culture plus at least two SIRS criteria and a central venous catheter, or one positive blood culture and three SIRS criteria (Fig. 1).

Figure 1.

 Algorithm to evaluate the clinical significance of a single blood culture positive for coagulase-negative staphylococci (CoNS) based on the presence of a central venous catheter (CVC) and systemic inflammatory response syndrome (SIRS) criteria.

Discussion

This study, involving 654 patients with CoNS bacteraemia, indicates that a single blood culture positive for CoNS is likely to represent BSI if at least three SIRS criteria or two SIRS criteria and a central venous catheter are present. These findings are consistent with previous studies, where a positive blood culture and a clinical picture compatible with infection had a similar positive predictive value as at least two positive blood cultures [4,16,28]. However, this is, to our knowledge, the first study assessing the clinical significance of CoNS bacteraemia using the SIRS criteria.

We have proposed an algorithm that can be easily used by clinicians to determine the clinical significance of CoNS isolated from blood cultures. Although only 36% of CoNS bacteraemias were defined as true BSI in our study, up to 50% of the study population fulfilled the definition of SIRS, which is defined as the presence of at least three SIRS criteria. The SIRS definition was therefore too sensitive and non-specific for bedside prediction of a BSI. This finding is in line with previous studies [29,30]. An antimicrobial therapy based only on the definition of SIRS would therefore lead to an overestimation of true BSI and to unnecessary prescription of antibiotics. This is particularly important in the setting of CoNS bacteraemia, which is considered to be a contamination on many occasions, and in regions where most CoNS isolates are resistant to methicillin, leading to an increased use of vancomycin and therefore to a selection of multiresistant bacteria. The positive predictive value of BSI in patients with only one SIRS criterion was too low (42.4%), but increased with each additional SIRS criterion to a maximum of 82.6% in patients who had four criteria. However, in our study, only a small proportion of patients with BSI presented with at least three SIRS criteria (28%) or four SIRS criteria (26%), indicating that an algorithm to identify true BSI due to CoNS should include also patients with only two SIRS criteria. When a central venous catheter was present, higher positive predictive value and accuracy were achieved. The algorithm with the best combined sensitivity and specificity for determining the clinical significance of CoNS was defined as at least two positive blood cultures, or one single positive blood culture and at least three SIRS criteria, or one single positive blood culture and at least two SIRS criteria and a central venous catheter.

In the literature, CoNS bacteraemias are reported to be highly associated with the use of intravascular devices [16,31–35], representing 30–60% of all catheter-related BSI [7,36]. These findings were supported by our results where the probability of BSI increased significantly among patients with two SIRS criteria in the presence of a central venous catheter and showed the highest positive predictive value for the definition of a BSI with CoNS. The type of positive SIRS criterion, however, was not useful in discriminating between BSI and contamination.

The overall prevalence of true BSI with CoNS was slightly higher than that reported in the literature, ranging from 10–12% to 20–30% [4,9,13,15,16,37]. Within our institution the occurrence of BSI with CoNS varied among different wards ranging from 27–32% in emergency and medical wards to 41–45% in the intensive-care unit and surgery wards. However, in multivariate analysis, after adjustment for the presence of a central venous line, these differences disappeared, indicating that central venous catheters were more frequently used in above-mentioned wards and were the main cause of CoNS BSI. Among other factors, demographic characteristics, co-morbidities, and the use of antibiotics did not vary between BSI and contamination.

We acknowledge some limitations. The lack of a reference standard to diagnose BSI with CoNS made it possible that misclassification of CoNS BSI had occurred during assessment of the clinical relevance of CoNS bacteraemia. However, this evaluation was consistently confirmed by an experienced infectious diseases specialist based on the clinical presentation and the lack of other infectious foci. Another limitation was the heterogeneity of patients regarding their underlying diseases and the cause of febrile episodes. Moreover, assessment of SIRS criteria might have presented some difficulties. For example, fever in patients already treated with antipyretic agents such as non-steroidal antirheumatics or steroids, and tachycardia in those treated with β-adrenergic blocking agents may be absent, leading to under-reporting of SIRS criteria. In contrast, tachypnoea in patients with heart failure or concomitant pneumonia might have overestimated the number of SIRS criteria. In addition, information on whether blood cultures were drawn from a central venous catheter was not available for this study. Moreover, we did not collect the time to positivity for CoNS bacteraemia. However, the scope of our study was the clinical assessment of true CoNS BSI and the evaluation of SIRS criteria. Finally, our algorithm was based on patients from a single hospital and results of this study may not be applicable to other populations, i.e. children and neutropenic patients.

Strengths of this study include the large sample size with 654 non-neutropenic patients presenting with at least one positive blood culture for CoNS. This is, to our knowledge, the first study focusing on SIRS criteria to discriminate between BSI and contamination. Our algorithm for the definition of a true BSI with CoNS represents an accurate and easy tool for clinicians to select those patients who require antibiotic treatment and catheter removal. In contrast to other studies, we included patients from different wards such as medical and surgical wards, emergency room and intensive-care units, allowing comparison within our hospital.

Conclusions

A positive blood culture with CoNS is likely to represent BSI if the patient has at least three SIRS criteria or two SIRS criteria and a central venous catheter. These findings may help clinicians to recognize true BSI in patients with a single positive blood culture with CoNS, providing an indication for starting adequate treatment.

Acknowledgements

This study was in part presented at the 21st ECCMID Meeting, Milan, Italy, 7–10 May 2011 (Poster # P1467).

Funding

This study has been supported by unrestricted grants of the Department of Internal Medicine, University Hospital Basel (L. Elzi).

Transparency Declaration

No potential conflicts of interest to declare.

Ancillary