European Journal of Internal Medicine 24 (2013) 541–545

Contents lists available at ScienceDirect

European Journal of Internal Medicine

j ourna l homepage: www.e lsev ie r .com/ locate /e j im

Original article

Clinical diagnostic accuracy of suspected sources of bacteremia and its effecton mortality☆

Jose M. Ruiz-Giardin a,⁎, Beatriz C. Jimenez a, Rosa M. Martin a, Judith Ortiz a, Myrna Hilda Condori Arenas b,Juan V. SanMartin a, Noemi Cabello a, Ana M. Barrios a, Diego Rejas a,Marialving Fuentes Salazar c, Jeronimo Jaquetti b

a Internal Medicine Department, Fuenlabrada University Hospital, Madrid, Spainb Clinical Analysis Department, Microbiology Unit, Fuenlabrada University Hospital, Madrid, Spainc Intensive Care Unit, Fuenlabrada University Hospital, Madrid, Spain

☆ Part of this workwas presented in the 21th European Cand Infectious Diseases, Milan, May 7–10, 2011: Congruencin the origin of the bacteremia and the final diagnosis: Isbacteremia?⁎ Corresponding author at: Internal Medicine Dep

Fuenlabrada Camino del Molino, nº 2 28942 Fuenlab91 6006179.

E-mail address: jruiz.hflr@salud.madrid.org (J.M. Ru

0953-6205/$ – see front matter © 2013 European Federhttp://dx.doi.org/10.1016/j.ejim.2013.05.010

a b s t r a c t

a r t i c l e i n f o

Article history:

Received 19 December 2012Received in revised form 7 May 2013Accepted 10 May 2013Available online 12 June 2013

Keywords:BacteremiaSourceMortalityAntibiotic treatment

Background: Although much has been written about bacteremia, evidence of the clinical diagnostic accuracy ofbacteremia sources in the absence of microbiological results and the impact of diagnostic accuracy on mortalityis scarce.Methods: This is a retrospective study of bacteremia episodes over a 2-year period at a general hospital inMadrid. Congruence analyses between clinically presumed and definite sources, acquisition, causative organism,empirical treatment and progression to death were performed.Results: The study included 323 bacteremia episodes. Clinicians' diagnostic accuracy was higher for gastrointes-tinal (88.8%; 95% CI: 79%–84%), respiratory (93.9%; 95% CI: 79%–99%) and urinary tract sources (83.6%; 95% CI:75%–89%) and lower for skin and soft tissues (77.2%; 95% CI: 54%–92%) and, notably, intravascular sources(56%; 95% CI: 39%–71%).

Overall, a non-significant (3.45%; 95% CI: −0.6%–13.5%, p = 0.47) increase in mortality was observed in theincorrectly suspected bacteremia source group.Mortality related to a definitive source was significantly higher when an intravascular origin was not suspected,resulting in a 26% increase in mortality (95% CI: 1%–52%, p = 0.03). Differences in mortality related to inaccu-rate source assumptions were non-significant when the definitive bacteremia sources were gastrointestinal,urinary, respiratory, skin and soft tissues or unknown.Mortality in the group receiving appropriate empirical antimicrobial treatment was 10.6% compared with 19.7%mortality in the group receiving inappropriate empirical antimicrobial treatment (OR 2; 95% CI: 1.01–4.25).Conclusions: The diagnostic accuracy of bacteremia sources is high in all but intravascular sources. Anon-suspected intravascular source and inappropriate empirical treatment are related to a higher mortality.

© 2013 European Federation of Internal Medicine. Published by Elsevier B.V. All rights reserved.

1. Introduction

Blood stream infections and infections of obligate anaerobes resultin high mortality, especially when antibiotic and surgical treatmentsare inappropriate [1–4]. For example, the attributed mortality ratesfor intravenous line-associated bacteremia range from 32% to 82%[5,6].

ongress of Clinical Microbiologye of suspected clinical diagnosesit important in the prognosis of

artment, University Hospitalrada, Madrid, Spain. Tel.: +34

iz-Giardin).

ation of Internal Medicine. Publishe

The source of bacteremia has prognostic and therapeutic impor-tance, as it is suggestive to clinicians of the most probable causativemicroorganisms. Therefore, the present study aimed to accomplishthe following: first, to analyse clinicians' diagnostic accuracy indetermining the source of bacteremia at a general hospital in Madrid,considering the influence of microbial aetiology, the definitive sourceof bacteremia and the location of bacteremia acquisition; and second,to assess the mortality associated with diagnostic accuracy and withappropriate empirical antimicrobial therapy.

2. Materials and methods

A retrospective analysis of all sources of adult bacteremia at theUniversity Hospital Fuenlabrada (UHF) was conducted. The UHF is a350-bed institution with surgical and medical services, with theexception of cardiovascular, neurology and thoracic surgery. Adult

d by Elsevier B.V. All rights reserved.

542 J.M. Ruiz-Giardin et al. / European Journal of Internal Medicine 24 (2013) 541–545

bacteremia episodes, including those due to strict anaerobes (highrisk bacteremias) [7], that occurred over a 2-year period (January2005–January 2007) were included in the study procedures. Thesuspected source of bacteremia before blood culture (BC) results be-came available was noted by the treating clinicians on the patient'selectronic records and was subsequently included in a global data-base. Patients with positive, non-contaminant BC were prospectivelyfollowed. The microbiology department alerted the clinicians on adaily basis to positivity of a BC or gram stain. This information waspassed on to one of the six staff physicians from the Infectious DiseaseUnit who were in charge of data collection. A case evaluation wasthen conducted, and blood cultures were classified as true positivesor contaminants. The definitive source of bacteremia was identifiedby three physicians after a thorough review of medical records oncethe causal microorganisms were identified.

2.1. Statistics

Comparisons of categorical variables were performed using theChi-square test (or Fisher test, as appropriate). Confidence intervalsfor proportions were calculated using the following SPSS macro to gen-erate 95% confidence intervals: Macro !CIP V2003.07.15, (c) A.Bonillo,JM.Domenech & R.Granero. To compare proportions, we calculatedthe confidence intervals of the differences between 2 independentproportions using the following SPSS macro: Macro !IC2PA V20.05.99(c) J.M.Domènech-Massons y R.Granero-Pérez. The level of statisticalsignificance level was set at p b 0.05. The analyses of sensitivity (S),specificity (E), and positive (PPV) and negative (NPV) predictive valueswere conducted with EPIDATA 3.1.

2.2. Definitions

a) True bacteremia/fungemia and contamination: True bacteremia/fungemia was defined by the presence of at least 1 of the following3 criteria:

1 One of the following microorganisms was isolated from at least oneBC bottle: gram-positive cocci (other than coagulase-negativestaphylococci (CNS) and Micrococcus sp.), gram-negative cocci(mainly Neisseria meningitidis), gram-negative bacilli or fungi.

2 Isolation of the same CNS species in at least 1 bottle from each pairof 2 BCs drawn on separate occasions from a patient exhibitingrecognised clinical signs of bacteremia.

3 Isolation of a CNS in a single BC bottle from a patient with aCNS-colonised intravascular line (>15 colony-forming units) forwhom the physician considered the institution of antibacterialtherapy.A diagnosis of contamination was made when CNS, Bacillus sp.,Propionibacterium sp., Micrococcus sp. or Corynebacterium sp. wereisolated in a single bottle from a patient with no clinical signs.A continuous episode bacteremia was considered when the samemicroorganism was isolated in a second BC drawn within 7 daysof the initial BC.

b) Source: The definitive source of bacteremia was defined based onclinical and/or microbiological findings. The source of bacteremiawas considered clinically documented if there were focal signsand symptoms and microbiologically documented when the samemicroorganisms were isolated from the blood and the infectedsite. In the absence of any identified source, the source of bacter-emia was classified as unknown.The bacteremia source was classified as intravascular (intravascularlines and endocarditis), gastrointestinal, respiratory, urinary, centralnervous system, osteoarticular, skin and soft tissues or unknown.

c) Acquisition: Acquisitionwas classified as nosocomial (N) when bac-teremia occurred at least 48 h after admission or when patientscame from nursing homes or long-term care facilities [8]. All otherbacteremias were classified as community-acquired (CA).

d) Empiric antimicrobial therapy: Empiric antimicrobial therapy wasdefined as treatment initiated after BC extraction and before theavailability of the antibiogram. Therapy was classified as correct ifit exhibited “in vitro” sensitivity and if the dosage and method ofadministration were appropriate.

e) Progression: The progression of bacteremia was classified as recov-ery according to clinical and/or microbiological outcomes of death ifmortality was directly or indirectly related to bacteremia (within30 days of bacteremia diagnosis).

f) Blood cultures: BC extraction depended on the treating physician'scriteria. The BACT-Alert detection system was used. Two pairs ofbottles were drawn per patient, each with a 20–40 mL recom-mended blood volume.

3. Results

There were a total of 323 bacteremia episodes.

3.1. Type of bacteremia according to aetiology

3.1.1. Bacteremia due to aerobic or facultative aerobic–anaerobicmicroorganisms

This group included 267 cases of bacteremia. Escherichia coliwas themost frequently isolatedmicroorganism (119 cases. 44.6%), followed bycoagulase-negative staphylococci (23 cases, 8.6%) and Staphylococcusaureus (20 cases, 7.5%). There were 7 (2.6%) polymicrobial bacteremias.The overall mortality was 10.9% (29 patients).

The most common sources of bacteremia were urinary (114 cases,42.2%), gastrointestinal (44, 16.4%) and intravascular (41, 15.3%); in26 (8%) cases, the source was unknown.

The diagnostic accuracy of the bacteremia source was as follows:gastrointestinal, 41/44 (93.1%); urinary, 96/114 (84%); skin and softtissue, 11/14 (78%); and intravascular, 23/41 (56%). The diagnosticaccuracy reached 92.8% for respiratory sources (26/28 bacteremias)(Table 1).

Misdiagnosis of the sources of aerobic/facultative aerobic–anaerobicbacteremia occurred in 19.2% of cases (95% CI: 1.54%–50%; p b 0.001).

3.1.2. Bacteremia due to strict anaerobesWe identified 56 episodes of bacteremia caused by strict anaerobes.

Bacteroides spp. and Clostridium spp. were responsible for 38 (67.8%)and 12 episodes (21.4%), respectively. Themortality due to strict anaer-obe bacteremia was 17.8% (10 patients). The most common source ofthe bacteria was gastrointestinal, with 37 cases (66.1%).

Diagnostic accuracy was as follows: gastrointestinal, 31/37 (83.8%);skin and soft tissue, 6/8 (75%); and unknown, 1/4 (25%). The accuracyfor respiratory sources of bacteremia approached 100% (Table 1).

Misdiagnosis of the anaerobic bacteremia source occurred in21.42% of cases (95% CI: 10.6%–32.1%; p = 0.0002).

There were no significant differences between the diagnostic ac-curacy of the bacteremia source for strict anaerobes compared withthe accuracy for other microorganisms (2.3%; 95% CI: −9%–4%; =0.68).

3.1.2.1. Microorganisms. Themost commonmicroorganisms causing in-travascular bacteremias included coagulase-negative staphylococci,Candida spp., and S. aureus. Gastrointestinal bacteremias were typicallycaused by gram-negative bacilli (E. coli and Klebsiella spp.).

The diagnostic accuracies of the bacteremia sources of themost com-monly isolated microorganisms were as follows (Table 1): Streptococcuspneumoniae, 91.6%; Enterococci, 90.9%; S. aureus, 90%; E. coli, 89.9%;Klebsiella sp., 88.8%; Bacteroides sp., 84.2%; and Candida sp., 75%. Thediagnostic accuracy for bacteremia sources was low in cases caused byEnterobacter sp., 66.6%; Clostridium sp., 58.3%; non-fermenting gram-

Definitive source

0

20

40

60

80

100

120

GI U R S Un iv

Nu

mb

er o

f p

atie

nts

Accurate diagnosis

All other diagnosis

GI: gastrointestinal; U: urinary; R: respiratory; S: skin and soft tissues; Un: unknown;

IV: intravascular

Fig. 1. Diagnostic accuracy of bacteremia sources in 236 community-acquiredbacteremias.

Table 1Microorganisms: Initial bacteremia source suspicions and definitive sources.a

Definite bacteremia source

Microorganisms;number of cases

Suspectedbacteremiasource

GI U R S Un IV O CNS Total

Polimicrobian: 9 GI 1 1 2Un 1 1 2S 2 2

E. coli: 119 U 77 77GI 19 4 23R 3 8 11Un 3 1 3 7

Bacteroides sp.: 38 GI 26 3 29Un 1 1 1 3R 1 2 3S 2 2

Coagulase negativeStaphylococci: 23

IV 9 9Un 1 2 5 8GI 1 3 4

S. aureus: 20 IV 7 7U 3 1 4Un 3 1 4S 2 2R 2 2

Klebsiella spp.: 18 GI 6 1 7U 6 6Un 1 1 2IV 2 2

Clostridium sp.: 12 GI 5 1 6Un 1 1 2S 0 2 2

S. pneumoniae: 12 R 11 1 12Enterococcus sp.: 11 GI 4 4

U 2 1 3Un 3 3

Nonfermentinggram-negativebacilli: 10

R 1 1 1 3Un 1 1 1 3S 2 2

Candida spp.: 8 IV 3 3GI 2 1 3

Enterobacter sp.: 6 Un 1 2 3U 2 2

GI: gastrointestinal; U: urinary; R: respiratory; S: skin and soft tissues; Un: unknown;IV: intravascular: osteoarticular; CNS: central nervous system.

a Microorganisms responsible for single cases are not included in the Table.

543J.M. Ruiz-Giardin et al. / European Journal of Internal Medicine 24 (2013) 541–545

negative bacilli, 60%; and, in particular, coagulase-negative staphylo-cocci, 52.1%.

Definitive source

0

5

10

15

20

25

30

GI U R S Un iv

Accurate diagnosis

All other diagnosis

Nu

mb

er o

f p

atie

nts

GI: gastrointestinal; U: urinary; R: respiratory; S: skin and soft tissues; Un: unknown;

IV: intravascular

Fig. 2. Diagnostic accuracy of bacteremia sources in 76 nosocomial bacteremias.

3.1.2.2. Acquisition.– Community acquired: A total of 236 cases of bacteremia were

community acquired. The urinary tract was the most common de-finitive source (41.5% of cases), followed by the gastrointestinaltract (26.6% of cases) (Fig. 1). The most common isolated microor-ganisms were E. coli, 111 cases (47%); S. aureus, 10 cases (4.2%);and S. pneumoniae, 11 cases (4.7%).

– Nosocomial: Nosocomial infections accounted for 73 cases ofbacteremia. An intravascular line was the most common definitiveorigin with 27 cases (39.1%), followed by the urinary tract, 16cases (23.2%); the gastrointestinal tract; 11 cases (15.9%); and un-known sources, 9 cases (13%) (Fig. 2). The most common isolatedmicroorganisms included coagulase-negative staphylococci, 16cases (23.2%); Klebsiella spp., 11 cases (15.9%); and S. aureus, 9cases (13%).

Globally, of the 312 bacteremia episodes for which the place ofacquisition was known, the source was misdiagnosed in 19.7% ofcases (95% CI: 10.7%–28.6%; p b 0.001).

There was a non-significant 1% difference in the diagnostic accuracybetween CA and N bacteremia (95% CI: −9%–11%; p = 0.83).

3.1.2.3. Source. The global diagnostic accuracy of physicians washigher for gastrointestinal sources of bacteremia, 88.8% (95% CI:79%–84%); respiratory origins, 93.9% (95% CI 95%: 79%–99%); andurinary sources, 83.6% (95% CI: 75%–89%). Conversely, diagnosticaccuracy was lower for skin and soft tissue sources, 77.2% (95% CI:54%–92%); and, notably, intravascular sources, 56% (95% CI: 39%–71%). It was significantly more difficult for physicians to accurately di-agnose the source of intravascular bacteremia compared with allother sources.

The sensitivity (S), specificity (E), and positive (PPV) and negative(NPV) predictive values (percentage) for each source were as follows:

Respiratory: S 93.5%, E 96.6%, PPV 74.3%, NPV 99.3%.Gastrointestinal: S 88.8%, E 93.7%, PPV 83.3%, NPV 95.7%.Urinary: S 83.6%, E 96.6%, PPV 93.2%, NPV 91.4%.Skin and soft tissues: S 77.2%, E 99.7%, PPV 93.7%, NPV 99.7%.Unknown: S 78.5%, E 90.6%, PPV 93.75%, NPV 99.7%.Intravascular: S 56%, E 100%, PPV 100%, NPV 94.06%.

Table 2Mortality associated with accurate and inaccurate bacteremia source predictions.

Bacteremia source suspicion Mortality

Definite sourcea Accurate Inaccurate Accurate source suspicionn (%)

Inaccurate source suspicionn (%)

OR (CI95%) p

GI 72 9 11 (15.2) 2 (22.2) 0.59U 94 18 5 (5.3) 0 0.31R 31 2 7 (22.5) 0 0.44S 15 5 4 (26.7) 0 0.19Un 22 4 4 (18.2) 1 (25) 0.75IV 23 17 2 (8.7) 6 (35.2) 5.7 (0.98–33) 0.03OA 2CNS 1Total 257 58 31 (12) 9 (15.5) 1.33 (0.59–2.99) 0.47

a GI: Gastrointestinal; U: Urinary, R: Respiratory; S: Skin and soft tissue; Un: Unknown; IV: intravascular; CNS: central nervous system.

544 J.M. Ruiz-Giardin et al. / European Journal of Internal Medicine 24 (2013) 541–545

3.1.2.4. Suspected bacteremia source and mortality. Information on thebacteremia source and patient mortality was available in 315 bacter-emia episodes. Mortality in the accurately suspected bacteremiasource group was 12% (31/257) vs. 15.5% (9/58) in the misdiagnosedsource group. The absolute difference in mortality was 3.45% (95% CI:−0.6%–13.5%; p = 0.47) and the odds ratio (OR) was 1.33 (95% CI:0.59–2.99).

In the case of intravascular bacteremias, when the source was mis-diagnosed, there was a 26% increase in mortality (95% CI: 1%52%;p = 0.03) compared with a correct source diagnosis. The OR was5.7 (95% CI: 0.98–33; p = 0.05). Otherwise, there were no significantdifferences in mortality when comparing accurately presumed andmisdiagnosed sources for the remaining definitive sources (Table 2).

When the bacteremia source was accurately diagnosed, therewere 8% more appropriately prescribed empirical treatments thanwhen the bacteremia source was inaccurately presumed, thoughthis difference was non-significant (95% CI: −5%–21%; p = 0.2).

Data on empirical antimicrobial treatment and mortality was avail-able for 297 bacteremia episodes (Table 3). The mortality in the groupreceiving appropriate empirical antimicrobial treatment was 10.6% (24out of 226 bacteremias). The mortality in the group receiving inappro-priate empirical antimicrobial treatment was 19.7% (14 out of 71bacteremias). The OR was 2 (95% CI: 1.01–4.25).

4. Discussion

This study demonstrates that, overall, clinicians' initial diagnosticaccuracy of the source of bacteremia is high, although in 20% of bac-teremia episodes, the source of infection is misdiagnosed. Diagnosticaccuracy is consistently approximately 80% for both strict anaerobicbacteremias and other aetiologies and both for CA and nosocomialbacteremias. In addition, clinicians' diagnostic accuracy is higherwhen the bacteremia source is gastrointestinal, respiratory or urinary

Table 3Mortality associated with appropriate and inappropriate empirical antimicrobialtherapy.

Empirical antimicrobialtherapy

Mortality

Definitesourcea

Appropriate Inappropriate Appropriaten (%)

Inappropriaten (%)

OR (CI95%)

GI 56 16 8 (14.2) 3 (18.8) 0.77 (0.16–3.11)U 87 23 4 (4.6) 1 (4.3) 1.06 (0.11–9.97)R 26 5 5 (19.2) 0 –

S 15 4 2 (13.3) 2 (50) 0.15 (0.01–1.80)Un 21 4 3 (14.3) 2 (50) 0.16 (0.01–1.67)IV 20 19 2 (10) 6 (31.5) 0.24 (0.04–1.38)CNS 1 1Total 226 71 24 (10.6) 14 (19.7) 2 (1.01–4.25)

a GI: Gastrointestinal; U: Urinary, R: Respiratory; S: Skin and soft tissue; Un:Unknown; IV: intravascular; CNS: central nervous system.

and lower when the bacteremia source is skin and soft tissue or, no-tably, intravascular, with congruencies of approximately 56%.

However, the diagnostic accuracy of the bacteremia source (withthe exception of intravascular bacteremia) is unlikely to influencemortality, most likely because it does not influence the appropriatenessof empirical antimicrobial treatment. Empirical antimicrobial treat-ment is the factor than can most influence mortality due to bacteremia.The importance of appropriate empirical treatment was confirmed bystudies such as the one by Paul M et al., where appropriate empiricalantibiotic treatment had a significant survival benefit in MRSA bacter-emia [9].

With respect to the association between mortality and the earlydiagnosis of the source of bacteremia, previous studies suggest thatinappropriate empirical antibiotic therapy is among the predictivefactors of mortality and that the antibiotic choice may be determinedby the suspected source of bacteremia [10–13].

Nevertheless, as Darvey PG and Marwick C et al. proposed, therelationship between inappropriate empirical antibiotic therapy andmortality should be carefully qualified by considering the timeelapsed from the outset of infection to the time of empirical antibiotictreatment to adjust mortality adequately by empirical treatment.Data on the time from infection to the beginning of treatment wasnot available for this study. In addition, the authors suggest thatalthough inappropriate therapy undoubtedly has an important influ-ence on outcomes, therapy needs to be considered in the context ofother patient risk-factors, such as co-morbid conditions, severityscore measures, and functional status [14].

Other studies have also reported that mortality rates varydepending on the source of bacteremia (gastrointestinal, lower respira-tory tract or unknown sources exhibit higher mortality [15,16]), theclinicians' accuracy in diagnosing intravascular bacteremias and thedelay to appropriate treatment [17]. Nevertheless Roberts andcolleagues [18] reported that intravascular bacteremias, with theexception of endocarditis, are not associated with a higher initial mor-tality. However, mortality increases progressively from day 5 onwards[18]. Brun-Buisson et al. comment on certain types of microorganismsassociated with non-severe sepsis, such as coagulase-negative staphy-lococci [19]. Bacteremias from intraabdominal, respiratory and menin-geal sources are all associated with severe sepsis in Brun-Buisson'sstudy. Moreover, the risk of death secondary to bacteremia increaseswith age, rapidly or ultimately fatal underlying disease, the presenceof severe sepsis, shock and infection caused by gram-positive microor-ganisms other than coagulase-negative staphylococci. The authors con-clude that the prognosis and severity of sepsis are influenced bysources of infection and that gram-negative bacilli and coagulase-negative staphylococci possess lower severity risks than other microor-ganisms [19].

In our study, we identified an association between inaccuratesource diagnosis in cases of intravascular bacteremia and mortality,possibly because of the use of inappropriate empirical antibiotic

545J.M. Ruiz-Giardin et al. / European Journal of Internal Medicine 24 (2013) 541–545

therapy, although the differences were not statistically significant.Other risk factors such as nosocomial bacteremia, resistant microor-ganisms (fungi, S. aureus, non-fermenting gram-negative bacilli),longer hospital stays, complications and severe comorbid conditionswere not analysed in the present study. In addition, it is likely thatcertain types of bacteremias are tolerant of inappropriate antimicro-bial treatments that delay appropriate therapy without increasingmortality.

To analyse mortality associated with appropriate and inappropriateempirical antimicrobial therapy for every source, we would need alarger number of patients exhibiting every source of bacteremia toidentify significant differences.

Therefore, the absence of differences in mortality secondary to theinitial diagnostic accuracy of the source of bacteremia, with theexception of intravascular sources, observed in our study could bedue to the following: first, accuracy in initial source suspicion; andsecond, appropriate use of empirical antimicrobial treatment, evenwhen the initial source suspicion was incorrect.

We believe that there is a need for improving the diagnostic accuracyof bacteremia sources when first assessing a septic patient, as improvedaccuracy would likely promote appropriate empirical treatments andshorter lengths of stay. Further, when any source of bacteremia isconfirmed, intravascular bacteremia should be excluded. However, thelatter guideline should be carefully adapted to individual health centres.

5. Summary

Clinicians' initial diagnostic accuracy of bacteremia sources is highfor all but intravascular sources. Diagnostic accuracy remains high forboth strict anaerobic bacteremias and other aetiologies and for bothcommunity and nosocomial bacteremias. The misdiagnosis of infec-tions of intravascular origins and inappropriate empirical antimicrobialtreatments are related to a higher mortality.

Learning points

• Clinicians' initial diagnostic accuracy of bacteremia sources is highfor all but intravascular sources.

• Diagnostic accuracy of bacteremia source doesn't like to have an in-fluence on mortality. The misdiagnosis of infections of intravascularorigins and inappropriate empirical antimicrobial treatments arerelated to a higher mortality.

Conflict of interests

All authors declare no conflicts of interest.

References

[1] Bosly A, Glupczynski Y. A 62-month retrospective epidemiological survey ofanaerobic bacteraemia in a university hospital. Clin Microbiol Infect 2006;12:527–32.

[2] Haddy RT, Nadkarni DD, Mann BL, Little DR, Domers TD. Clostridial bacteremia inthe community hospital. Scand J Infect Dis 2000;32:27–30.

[3] Redondo MC, Arbo MD, Glindlinger J, Snydman DR. Attributable mortality ofbacteremia associated with the Bacteroides fragilis group. Clin Infect Dis1995;20:1492–6.

[4] Ramos JM, García-Corbeira P, Fernández-Roblas Soriano F. Bacteriemia poranaerobios: análisis de 131 episodios. Enferm Infecc Microbiol Clin 1994;12:9–16.

[5] Wenzel RP, Edmond MB. The impact of hospital-acquired bloodstream infections.Emerg Infect Dis 2001;7:174–7.

[6] Vallés J, León C, Álvarez-Lerma F. Nosocomial bacteremia in critically ill patients: amulticenter study evaluating epidemiology and prognosis. Clin Infect Dis1997;24:387–95.

[7] Bouza E, Reig M, Garcia de la Torre M, Rodríguez-Créixems M, Romero J,Cercenado E, et al. Retrospective analysis of two hundred and twelve cases ofbacteremia due to anaerobic microorganisms. Eur J Clin Microbiol 1985;4:262–7.

[8] Siegman-Igra Y, Fourer B, Orni-Wasserlauf R, Golan Y, Noy A, Schwartz D, et al.Reappraisal of community-acquired bacteremia: a proposal of a new classificationfor the spectrum of acquisition of bacteremia. Clin Infect Dis 2002;34:1431–9.

[9] Paul M, Kariv G, Goldberg E, Raskin M, Shaked H, Hazzan R, et al. Importance ofappropriate empirical antibiotic therapy for methicillin-resistant Staphylococcusaureus bacteraemia. J Antimicrob Chemother 2010;65:2658–65.

[10] Gómez J, García-Vázquez E, Baños R, Canteras M, Ruiz J, Baños V, et al. Predictorsof mortality in patients with methicillin-resistant Staphylococcus aureus (MRSA)bacteraemia: the role of empiric antibiotic therapy. Eur J Clin Microbiol InfectDis 2007;26:239–45.

[11] Lombardi DP, Engleberg C. Anaerobic bacteremia: incidence, patient characteristics,and clinical significance. Am J Med 1992;92:53–60.

[12] Arpi M, Renneberg J, Andersen HK, Nielsen B, Larsen SO. Bacteremia at a Danishuniversity hospital during a twenty five-year period (1968–1992). Scand J InfectDis 1995;27:245–51.

[13] Salonen JH, Eerola E, Meurman O. Clinical significance and outcome of anaerobicbacteremia. Clin Infect Dis 1998;26:1413–7.

[14] Davey PG, Marwick C. Appropriate vs. inappropriate antimicrobial therapy. ClinMicrobiol Infect 2008;14:15–21.

[15] Angus DC, Wax RS. Epidemiology of sepsis: an update. Crit Care Med 2001;29:S109–16.

[16] Rello J, Ricart M, Mirelis B, Quintana E, Gurgui M, Net A, et al. Nosocomial bacter-emia in a medical–surgical intensive care unit: epidemiologic characteristics andfactors influencing mortality in 111 episodes. Intensive Care Med 1994;20:94–8.

[17] Lai CH, Chi CY, Chen HP, Chen TL, Lai CJ, Fung CP, et al. Clinical characteristics andprognostic factors of patients with Stenotrophomonas maltophilia bacteremia. JMicrobiol Immunol Infect 2004;37:350–8.

[18] Roberts FJ, Geere IW, Coldman A. A three-year study of positive blood cultures,with emphasis on prognosis. Rev Infect Dis 1991;13:34–46.

[19] Brun-Buisson C, Doyon F, Carlet J. Bacteremia and severe sepsis in adults: a multi-center prospective survey in ICUs and wards of 24 hospitals. French bacteremia-sepsis study group. Am J Respir Crit Care Med 1996;154:617–24.