Infection

Clinical and Epidemiological Study

Low Rate of Clinical Consequences Derived from Results of Blood Cultures Obtained in an Internal Medicine Emergency Department
B.P. Ehrenstein, T. Jarry, H. J. Linde, J. Schlmerich, T. Glck

Abstract
Background: Blood cultures detect bacteremia in individual patients and help define local pathogen and resistance spectra. At the same time, the benefits of blood culture results in the management of individual patients and therefore their cost-effectiveness are disputed. Patients and Methods: During 1 calendar year, we conducted a prospective study of emergency department admissions with blood culture draws and at least a 3-day hospitalization afterwards. We prospectively surveyed treating physicians on usefulness of blood culture results for patient management. Results: 428 diagnostic episodes (emergency visits) involving 390 patients occurred during the study period from 10/2002 to 10/2003. The analysis included 188/428 (44%) episodes with blood culture draws performed according to the predefined clinical standard where patients were hospitalized with sufficient duration. Absence of therapeutic consequences in response to blood culture results was reported for 138/142 (97%) of episodes with negative blood culture results, for 16/21 (76%) with blood culture results positive only for skin flora, and for 14/25 (56%) of episodes with blood cultures positive for obligate pathogens. Treating physicians regarded the blood culture results necessary for clarifying the etiology in 34/188 (18%) episodes, and rated blood culture results necessary for their therapeutic decisions in 29/188 (15%) episodes. Conclusion: Negative blood culture results rarely changed the management of medical inpatients. Our study suggests that in settings with broad-spectrum empirical antibiotic therapy positive blood culture results for obligate pathogens trigger adjustment of the antibiotic therapy in only about half of instances. Many blood cultures drawn in the emergency department where considered unnecessary by ward physicians. Guidelines for preventing unnecessary blood culture draws are warranted in order to increase the rate of their meaningful clinical consequences for medical inpatients initially treated with broad-spectrum empirical antibiotics.

Introduction
Blood cultures (BC) are important for defining local spectra of pathogens and resistance in severe infections. At the same time, the benefits of BC results for the management of individual patients and therefore their cost-effectiveness are disputed [16]. Two studies evaluating the clinical impact of BC obtained from adult and pediatric patients in emergency departments (ED) reported few relevant changes in the clinical management caused by BC results (1.6% and 0.4%, respectively) [1, 2]. Both studies were retrospective and were conducted by infectious disease specialists. Since medical charts do not document all factors that influence treatment decisions, results of retrospective studies could be biased. We conducted a prospective study aimed to observe specific consequences of BC results in routine clinical practice in a patient population that is commonly treated empirically with broad-spectrum antibiotics.

Patients and Methods

We conducted a prospective observational study of adult patients with BC diagnostic work-up, seen under routine conditions during 1 calendar year at the internal medicine ED of Regensburg University Medical Center. During the study period, ED resident or attending physicians ordered BC at their discretion. The implied rule although not an official guideline was to obtain BC whenever severe bacterial infections necessitating inpatient antibiotic therapy were suspected at the initial evaluation. A predefined clinical standard for BC diagnostics in the ED gave no rules for BC indication but provided specific guidelines on how to obtain BC, once deemed indicated

B.P. Ehrenstein (corresponding author), T. Jarry, J. Schlmerich, T. Glck Dept. of Internal Medicine I, University Medical Center Regensburg, 93042 Regensburg, Germany; Phone: (+49/941) 944-7003; Fax: -7049; e-mail: [email protected] H.J. Linde Institute of Medical Microbiology and Hygiene, University of Regensburg, Regensburg, Germany Received: April 29, 2005 Revision accepted: July 15, 2005 This paper is dedicated to the founders of the Walter Marget Foundation, D. Adam and F. Daschner, in gratitude for their support of the training in infectious diseases.

Infection 2005; 33: 314319 DOI 10.1007/s15010-005-5065-5

314

Infection 33 2005 No. 5/6 URBAN & VOGEL

B.P. Ehrenstein et al. Clinical Consequences of Blood Cultures

by the emergency physicians. 50 ml of blood were to be obtained from two separate blood draws and subsequently cultured in five bottles of a commercially available BC system (BacT/Alert, Organon Teknika, Durham, NC, USA). The first blood sample (20 ml, one aerobic and one anaerobic bottle) could be drawn through a freshly placed peripheral venous access catheter, while the second sample (30 ml, two aerobic and one anaerobic bottle) had to be obtained with a separate venipuncture. For both draws, skin was disinfected with an alcohol-based solution (SoftaseptTM N, B. Braun, Melsungen, Germany) for at least 1 min. BC bottles were transported, cultured and analyzed under routine conditions. For positive BC, a preliminary report was sent out immediately after microbiologic growth was observed, and for negative cultures, after 48 h of cultivation. At the time of BC draw, ED personnel was asked to complete a short written multiple-choice questionnaire documenting suspected type of infection, antibiotic pretreatment, timing of the BC draw, and compliance with the standard. Patients could enter the study several times if BC were obtained on separate visits to the ED. All BC of each separate visit were analyzed as a single observation and termed diagnostic episode (hereafter episode). The clinical relevance of bacteria commonly found in the skin flora isolated in BC is often difficult to determine, even prospectively. In our study, coagulase-negative staphylococci, Micrococcus spp., Peptostreptococcus spp., Propionibacterium spp., Acinetobacter lwoffii, and aerobic spore-forming gram-positive rods were termed skin flora (SF) and considered either contaminants or, if isolated at least out of two separate BC draws, a possible cause of infections. All other isolated pathogens were considered obligate pathogens (OP). For patients admitted and hospitalized for at least 3 days, with completed ED questionnaire and five BC bottles drawn, a second multiple-choice questionnaire was given to their treating (ward) physicians. The physicians were asked what changes in the antibiotic management, if any, resulted from BC drawn in the ED. In addition, the physicians were asked to rate (necessary, helpful, unnecessary) the importance of those BC results for determining infection etiology and for decisions regarding antibiotic management. The episodes from patients with completed ED questionnaire and five BC bottles drawn, and who were hospitalized for at least 3 days, were included in the study. For the included episodes, we retrospectively retrieved from medical charts information on empirical antibiotic therapy (1 day

after ED visit) and antibiotic therapy after BC results (3 and 5 days after admission). Whenever possible, we resolved discrepancies in questionnaire data and filled in missing information by personally recontacting physicians and/or by consulting medical charts. Data were entered into the clinical study database and then linked with a database of all BC results from the Department of Microbiology. Data were analyzed using SPSSTM software (version 12.0, SPSS Inc., Chicago, IL, USA). Crude associations were analyzed using 2 test or Fischers exact test as appropriate. We assumed that each episode was an independent observation. Results were considered statistically significant if p-values were less than or equal to 0.05. The study was performed in accordance with the rules set for observational studies by the local ethics committee. The patients informed consent was not required since the BC draws were done for clinical purposes and were to be done according to general clinical standards.

Results
During the observation period (October 26, 2002 to October 25, 2003; 5004 ED patient visits), 428 episodes originating from 390 patients occurred in the ED. The five-bottle BC standard was followed in 222/428 (52%) episodes. In 188/428 (44%) episodes, the standard was followed and the patient was hospitalized for 3 days or longer. These 188 episodes were included in the study analyses. Lack of questionnaire data and failure to follow the predefined five-bottle BC standard caused 206 episodes to be excluded. An additional 34 episodes were excluded because hospital stays were shorter than 3 days. Compared with the excluded episodes, the included episodes were more likely to yield SF (12% vs 7%) and had higher mean patient age (56 vs 49 years) (Table 1). Among the 188 included episodes, 25 (13%) yielded OP, and 23 (12%) yielded SF. Of the 25 cultured OP, six (24%) were detected only in the anaerobic BC bottles, seven (28%) only in the aerobic bottles, and 12 (48%) were detected in both aerobic and anaerobic media. 19 (76%) of the 25 OP detected were detected with the first pair of

Table 1 Comparison of included and excluded diagnostic episodes.

Episodes included analysis (n = 188) Patient age Mean ( SD) Range No. of BC bottles per episode Mean ( SD) Range No. (%) episodes with positive BC results No. (%) OP-positive episodes No. (%) SF-positive episodes 55.7 ( 17.7)a 17 89 5.0 ( 0.3)a 57 46 (24) 25 (13) 23 (12)b

Episodes excluded analysis (n = 240) 49.5 ( 19.7)a 16 93 2.6 ( 1.1)a 16 45 (19) 29 (12) 16 (7)b

All episodes (n = 428) 53.2 ( 19.1) 16 93 3.7 ( 1.5) 17 91 (21) 53 (12) 39 (9)

a Statistically significant difference (t-test, p < 0.01); b statistically significant difference (2 test, p = 0.05); BC: blood culture(s); OP: obligate pathogens; SF: skin flora

Infection 33 2005 No. 5/6 URBAN & VOGEL

315

B.P. Ehrenstein et al. Clinical Consequences of Blood Cultures

BC bottles. When two pairs (four BC bottles) were used, the detection rate increased to 24/25 (96%). Mean number of positive bottles was 3.2 for OP and 1.4 for SF. In the 121/188 (64%), episodes where the first pair of BC bottles was obtained through a newly placed peripheral venous catheter and the second, through a separate venipuncture, SF was more likely to be isolated from the venous catheter (8.3% vs 3.3%, p = 0.15 for trend by non-parametric McNemar test). Species isolated from the 188 episodes, stratified by pathogenicity, are described in table 2. The 188 episodes occurred among 177 patients (seven patients had two episodes and two patients had three episodes). 77 (41%) episodes originated from female patients. The mean age was 56 years (SD 17.7 years; range 1793 years). Active malignant disease as an underlying condition was known or strongly suspected in 67/188 (36%) episodes. The suspected types of infection before obtaining BC were respiratory (34%), abdominal (18%), urogenital (8%), (severe) sepsis (7%), musculoskeletal infection (4%), endocarditis (2%), and one patient was suspected to have meningitis. For the remaining 48 (25%) episodes, the ED physicians could not specify the type of infection before ordering BC. No empirical antibiotic therapy was started in only 10/188 (5%) episodes after BC drawn in the ED. The anti-

Table 2 Pathogens isolated from the 188 diagnostic episodes included in the study.

Isolated pathogens Obligate pathogens Escherichia coli Streptococcus spp. Enterococcus spp. Candida albicans Klebsiella pneumoniae Pseudomonas aeruginosa Staphylococcus aureus (methicillin sensitive) Streptococcus pneumoniae All obligate pathogens Skin flora Staphylococcus spp., coagulase-negative Aerobic spore-forming gram-positive rods Micrococcus spp. Peptostreptococcus spp. Acinetobacter lwoffii Propionibacterium acnes All pathogens representing skin flora All pathogens
a Two

No. of episodes 10 8 2 1 1 1 1 1 25 14 5 2 2 1 1 23a 46b

episodes yielded two separate isolates representing skin flora (1 x Acinetobacter lwoffii and coagulase-negative staphylococci, 1 x Micrococcus spp. and Propionibacterium acnes); b two epsisodes yielded one isolate of an obligate pathogen and one isolate representing skin flora

biotics used for empirical therapy on the day after admission (i.e. before results of the BC were available) and the susceptibilities of the subsequently isolated bacterial OP are given in table 3. The most commonly used antibiotic was piperacillin, mostly in combination with the betalactamase inhibitor sulbactam. More than 90% of the bacterial pathogens were susceptible to carbapenems and to the combination of piperacillin and sulbactam. The isolated organism was in vitro resistant to the empirically given antibiotic regimen in four (16%) of the 25 episodes yielding OP, resulting in the following therapy adjustments. One patient with endocarditis caused by Streptococcus mitis, empirically treated with ciprofloxacin was switched to penicillin G and gentamicin; one patient with neutropenic bacteremia caused by Klebsiella pneumoniae resistant to piperacillin/sulbactam empirically treated with these antibiotics and voriconazole was switched to meropenem and voriconazole (vancomycin and metronidazole were added simultaneously because of the patients rapid clinical deterioration); one patient with cholangitis and BC positive for ampicillin-resistant enterococci empirically treated with piperacillin/sulbactam was switched to piperacillin/sulbactam and vancomycin; and one patient with a presumed respiratory infection on the basis of a refractory grade IV non-Hodgkins lymphoma with candidemia caused by Candida albicans, empirically treated with piperacillin/sulbactam and erythromycin, was switched to piperacillin/sulbactam, erythromycin, and amphotericin B. The patient with neutropenic bacteremia and the patient with candidemia died, respectively, 5 and 4 days after admission. The treating physicians attributed the deaths to severe infections, although both patients had underlying malignant diseases with grave prognoses. The difference in hospital mortality between OP-positive and OP-negative episodes was not statistically significant (2/25 (8.0%) vs 9/163 (5.5%), p = 0.62). Table 4 summarizes the analysis of clinical consequences of BC obtained in the ED, stratified by isolated pathogens. Among the 142 episodes with negative BC results, antibiotic therapy was ended after obtaining the negative results in three episodes, and in one additional episode, the duration of the antibiotic therapy was shortened. No change of antibiotic therapy in response to BC occurred in 14/25 (56%) of OP-positive episodes. Among the 11 OP-positive episodes with any change in the antibiotic management in response to BC, in one episode, the empirically prescribed therapy was prolonged. Of the ten remaining episodes, antibiotic spectrum was narrowed in three and broadened in seven. Two of the ten episodes also received therapy prolongation. For 21 episodes with only SF isolated, no change in the antibiotic management occurred in 16 (76%) episodes, the spectrum of the antibiotic regimen was broadened in two (9%) episodes, and the duration of the antibiotic therapy was prolonged in three (14%) episodes.

316

Infection 33 2005 No. 5/6 URBAN & VOGEL

B.P. Ehrenstein et al. Clinical Consequences of Blood Cultures

The usefulness of the results of the BC at admission for elucidating infection etiology was rated by ward physicians as necessary for 34 (18%) episodes, as helpful for 53 (28%) episodes, and as unnecessary for 101 (54%) episodes. The corresponding physician ratings for the usefulness of BC for deciding on antibiotic therapy were 29 (15%), 56 (30%), and 103 (55%).

Table 3 Empirical antibiotic therapy for 188 diagnostic episodes and antibiotic susceptibilities of the 23 isolated bacterial obligate pathogens.

Antibiotic

Empiric therapy 1 day after ED admission for 188 episodes No. (%)

Antibiotic susceptibilities derived of the 24 bacterial OP from 188 episodes % 37a 67 71 71 92 96 79 79 83 87 50 50 42a 42a 50a

Discussion
We evaluated clinical consequences of BC obtained at ED admission for subsequently hospitalized patients. Almost invariably, treating physicians were reluctant to change the antibiotic therapy of patients in response to the negative BC results, indicating low relevance of negative results for decisions regarding de-escalation or ending empirical broad-spectrum antibiotic therapy. Furthermore, BC results positive for OP triggered any change in the antibiotic management of patients in less than half of the episodes. Contrary to our expectation that the isolation of a causative patho-

Penicillin G Ampicillin Piperacillin (ureidopenicillinsb) Amoxicillin + clavulanate Piperacillin + sulbactam Meropenem Cefazolin Cefuroxime Ceftriaxone Ceftazidime Gentamicin (aminoglycosidesb) Ciprofloxacin (fluoroquinolonesb) Erythromycin (macrolidesb) Clindamycin Metronidazole Vancomycin
a Gram-negative

3 (2) 3 (2) 3 (2) 35 (19) 80 (43) 7 (4) 0 (0) 4 (2) 23 (12) 1 (1) 6 (3) 41 (22) 35 (19) 4 (2) 22 (12) 7 (4)

rods are constitutively resistant to these antibiotic drugs. Therefore, these isolates were not specifically tested but rated as resistant. ED: emergency department; OP: obligate pathogens; b empiric antibiotics of the same class were reported together (the class is stated in brackets)

Table 4 Clinical consequences derived from blood cultures (BC) obtained in the emergency department, stratified by pathogen type.

Clinical consequences derived of the BC results and rating of their importance by treating physicians No change in management Antibiotic therapy discontinued Change to or addition of a specific antibiotic treatment for the detected pathogen Antibiotic spectrum broadened Antibiotic spectrum narrowed Duration of antibiotic therapy shortened Duration of antibiotic therapy prolonged Importance of BC results for elucidating etiology Necessary Helpful Unnecessary Importance of BC results for deciding on antibiotic therapy Necessary Helpful Unnecessary OP: obligate pathogens; SF: skin flora

25 episodes with BC positive for OP No. (%) 14 (56) 0 (0) 10 (40) 7 (28) 3 (12) 0 (0) 3 (12) 8 (32) 12 (48) 5 (29)

21 episodes with BC positive only for SF No. (%) 16 (76) 0 (0) 2 (9) 2 (9) 0 (0) 1 (5) 3 (14) 5 (24) 8 (38) 8 (38)

142 episodes with negative BC No. (%) 138 (97) 3 (2) 0 (0) 0 (0) 0 (0) 1 (1) 0 (0) 21 (15) 33 (23) 88 (62)

All 188 episodes No. (%) 168 (89) 3 (2) 12 (6) 9 (5) 3 (2) 2 (1) 6 (3) 34 (18) 53 (28) 101 (54)

9 (36) 10 (40) 6 (24)

5 (24) 8 (38) 8 (38)

15 (11) 38 (27) 89 (63)

29 (15) 56 (30) 103 (55)

Infection 33 2005 No. 5/6 URBAN & VOGEL

317

B.P. Ehrenstein et al. Clinical Consequences of Blood Cultures

gen would lead to narrowing of the antibiotic spectrum, in seven of ten patients with any change due to the OP-positive BC results, the regimen was broadened. In 20/188 (11%) episodes, BC results triggered any change of the antibiotic management. BC yielded isolates resistant to the initial empirical antibiotic regimen only in 5/188 (3%) episodes (four OP-positive and one SF-positive episodes), whereby two patients died despite the immediate switch to an adequate antibiotic regimen. Therefore, only in 3/188 (2%) episodes did patients clearly benefit from the results of the BC diagnostics, with an additional 15/188 (8%) episodes having a probable benefit. We did not observe any relevant reduction in antibiotic treatment for patients in response to BC results. These findings are corroborated by treating physicians low ratings of usefulness of blood culture results for elucidating the etiology and for deciding on the antibiotic therapy. In more than half of all evaluated episodes, BC results were rated as unnecessary for diagnostic and therapeutic management of patients, while only in 15% of episodes were the BC results rated as necessary for deciding on the antibiotic therapy. Interestingly, treating physicians, albeit not reporting any change in management for 97% of episodes with negative BC results, still rated these BC results necessary or helpful for deciding on the antibiotic therapy in 38% of the episodes. An explanation for this contradiction could be an inadequate interpretation of negative BC results by some of the treating physicians in our study. Despite the fact that negative BC results do not rule out a relevant transient bacteremia, physicians could judge infections with negative BC results as less severe and opt for shorter duration of antibiotic therapy, if the patient improved rapidly. We speculate that some physicians may have omitted the information of decreased therapy duration on the questionnaires. To our knowledge, no other prospective study evaluating the impact of BC results on the antibiotic management of inpatients with medical emergencies has been reported. A retrospective study, using chart review by an infectious disease specialist to evaluate the clinical impact of 1,062 BC obtained from adult patients in an ED, reported a similar, albeit lower, rate of 1.6% of BC results leading to any change in the management of patients [1]. Another retrospective study of pediatric patients with medical emergencies evaluating the clinical impact of results of BC found an even lower rate of 0.4% in response to BC results [2]. The retrospective design relying solely on medical charts, however, could have yielded a biased estimate of the magnitude of the true clinical impact of BC results on the antibiotic management of patients, because medical charts often do not directly list all the reasons for changes in the antibiotic therapy. We did not observe a high rate of unnecessary antibiotic therapy triggered by the isolation of pathogens considered SF. The treating physicians in our study changed the antibiotic management only in 5/21 (24%) of SF-posi-

tive episodes, rating most of these isolates as contaminants. Although we neither evaluated the change in the length of stay, nor the cost of treatment triggered by SF-positive BC, a comparison of our findings with a study from the early 1990s evaluating the consequences of false-positive BC results (e.g. contaminants) suggests a better recognition of possible contaminants nowadays [7]. The common use of broad-spectrum antibiotics for empirical therapy documented in our study (Table 3) reflects the clinical setting of a tertiary care hospital characterized by patients with complex medical histories, multiple comorbidities, common antibiotic pretreatment, and in the case of malignant diseases or therapeutic immunosuppression frequently fulminant courses of their infections. This use of broad-spectrum antibiotics probably contributed to the low rate of detected BC isolates resistant to the initial antibiotic regimen. Our study has several limitations. The decision to obtain blood cultures and interpretation of their results rested largely on the ED and ward physicians in our general internal medicine department, despite the availability of an infectious disease consult service. This study was designed neither to determine optimal utilization of BC, nor expert interpretation of their results, but rather to represent current clinical practice. By choosing the prospective design we avoided detection bias and thereby improved the generalizability of our results. Our distinction between OP and SF was based solely on the species of bacteria. Such classification avoids the dependence of pathogenicity assignment on retrospective classification, which was the case in earlier studies [8-10]; it has also some possible flaws because some of the streptococcal isolates grouped as OP could be contaminants and at least in one case in our study (a patient with Wilsons disease and acute spondylodiscitis), while the isolated coagulase-negative staphylococci, grouped as SF, may have represented the actual pathogen responsible for the infection (data not shown). During the study period, the predefined standard for performing BC in the ED was to obtain 50 ml of blood by two separate venipunctures. Unfortunately, the staff of the ED did not adhere to this standard and therefore did not document the BC diagnostics strictly in approximately half of all potentially eligible episodes. Since the distribution of the follow-up questionnaire to ward physicians depended on the initial study documentation of BC in the ED, only 188 episodes could be included in the analyses. The comparison of included and excluded episodes (Table 1) shows that the patients with included episodes tended to be older. The significantly lower number of BC bottles per episode for the excluded episodes could be explained by ED staff trying to circumvent the additional workload of a second veniuncture and study documentation. We acknowledge that using 44% of all potentially eligible episodes in our study limits its internal and external validity, but from the available data of the excluded episodes, no relevant sys-

318

Infection 33 2005 No. 5/6 URBAN & VOGEL

B.P. Ehrenstein et al. Clinical Consequences of Blood Cultures

tematic bias other than the lower number of BC bottles per episode could be detected. With six (24%) of the 25 OP detected only in anaerobic BC bottles, our results do not support the notion of reserving anaerobic BC media for clinical situations with a high likelihood for anaerobic bacteremia, as suggested by some other authors [1113]. Obtaining a second pair of BC bottles (e.g. culturing 40 ml instead of 20 ml of blood) increased the detection rate of OP by 26%, which was comparable to previous studies [1416]. Because previous studies failed to detect a significantly higher contamination rate with SF pathogens (with obtaining BC through a newly placed venous catheter [17, 18]), the predefined standard of the ED allowed the first pair of BC to be obtained by this method for convenience. A historical comparison of BC contamination rates with and without the use of venous catheters in pediatric patients (published after the initiation of our study) confirms our finding (although only a statistical trend) of higher contamination rates with BC obtained through venous catheters as compared with dedicated venipuncture [19]. In summary, our study showed that negative BC results rarely changed the management of medical inpatients of a tertiary care hospital. In the case of blood culture results positive for OP, less than half of the patients received any adjustment of their antibiotic management. Many blood cultures drawn in the ED were not regarded necessary by ward physicians. Our results support efforts to design and evaluate specific decision rules to focus BC on situations with reasonable chances for the isolation of OP, or in situations with infections necessitating long-term antibiotic therapies (e.g. endocarditis). Regardless of the limited relevance for adjusting the antibiotic therapy on an individual level, the importance of continued BC diagnostics for recognizing changes in pathogen spectra and resistance spectra locally remains untouched by our results.

4.

5.

6.

7.

8.

9.

10.

11.

12.

13. 14. 15.

16.

References
17. 1. 2. Kelly AM: Clinical impact of blood cultures taken in the emergency department. J Accid Emerg Med 1998; 15: 254256. Leonard P, Beattie TF: How do blood cultures sent from a paediatric accident and emergency department influence subsequent clinical management? Emerg Med J 2003; 20: 347348. Laupland KB, Davies HD, Church DL, Louie TJ, Dool JS, Zygun DA, Doig CJ: Bloodstream infection-associated sepsis and septic shock in critically ill adults: a population-based study. Infection 2004; 32: 5964.

18.

3.

19.

Ruggeberg JU, Ketteler K, MacKenzie CR, Von Kries R, Reinert RR, Schroten H: Blood culture sampling rates at a German pediatric university hospital and incidence of invasive pneumococcal disease. Infection 2004; 32: 7881. Anatoliotaki M, Valatas V, Mantadakis E, Apostolakou H, Mavroudis D, Georgoulias V, Rolston KV, Kontoyiannis DP, Galanakis E, Samonis G: Bloodstream infections in patients with solid tumors: associated factors, microbial spectrum and outcome. Infection 2004; 32: 6571. Anderson DJ, Murdoch DR, Sexton DJ, Reller LB, Stout JE, Cabell CH, Corey GR: Risk factors for infective endocarditis in patients with enterococcal bacteremia: a case-control study. Infection 2004; 32: 7277. Bates DW, Goldman L, Lee TH: Contaminant blood cultures and resource utilization. The true consequences of false-positive results. JAMA 1991; 265: 365369. Leibovici L, Greenshtain S, Cohen O, Mor F, Wysenbeek AJ: Bacteremia in febrile patients. A clinical model for diagnosis. Arch Intern Med 1991; 151: 18011806. Bates DW, Cook EF, Goldman L, Lee TH: Predicting bacteremia in hospitalized patients. A prospectively validated model. Ann Intern Med 1990; 113: 495500. Bates DW, Sands K, Miller E, Lanken PN, Hibberd PL, Graman PS, Schwartz JS, Kahn K, Snydman DR, Parsonnet J, Moore R, Black E, Johnson BL, Jha A, Platt R: Predicting bacteremia in patients with sepsis syndrome. Academic Medical Center Consortium Sepsis Project Working Group. J Infect Dis 1997; 176: 15381551. Hellinger WC, Cawley JJ, Alvarez S, Hogan SF, Harmesen WS, Ilstrup DM, Cockerill FR, 3rd: Assessment of routine use of an anaerobic bottle in a three-component, high-volume blood culture system. J Clin Microbiol 1996; 34: 25442547. Morris AJ, Wilson ML, Mirrett S, Reller LB: Rationale for selective use of anaerobic blood cultures. J Clin Microbiol 1993; 31: 21102113. Mylotte JM, Tayara A: Blood cultures: clinical aspects and controversies. Eur J Clin Microbiol Infect Dis 2000; 19: 157163. Li J, Plorde JJ, Carlson LG: Effects of volume and periodicity on blood cultures. J Clin Microbiol 1994; 32: 28292831. Ilstrup DM, Washington JA, 2nd: The importance of volume of blood cultured in the detection of bacteremia and fungemia. Diagn Microbiol Infect Dis 1983; 1: 107110. Mermel LA, Maki DG: Detection of bacteremia in adults: consequences of culturing an inadequate volume of blood. Ann Intern Med 1993; 119: 270272. Isaacman DJ, Karasic RB: Utility of collecting blood cultures through newly inserted intravenous catheters. Pediatr Infect Dis J 1990; 9: 815818. Smart D, Baggoley C, Head J, Noble D, Wetherall B, Gordon DL: Effect of needle changing and intravenous cannula collection on blood culture contamination rates. Ann Emerg Med 1993; 22: 11641168. Norberg A, Christopher NC, Ramundo ML, Bower JR, Berman SA: Contamination rates of blood cultures obtained by dedicated phlebotomy vs intravenous catheter. JAMA 2003; 289: 726729.

Infection 33 2005 No. 5/6 URBAN & VOGEL

319