Intensive Care Medicine

, Volume 38, Issue 12, pp 1930–1945

Characteristics and determinants of outcome of hospital-acquired bloodstream infections in intensive care units: the EUROBACT International Cohort Study

Authors

  • Alexis Tabah
    • Medical ICUAlbert Michallon University Hospital, Université Grenoble 1
    • Albert Bonniot Institute, Team 11: Outcome of Mechanically Ventilated Patients and Respiratory CancersUniversité Grenoble 1, U 823
    • Outcomerea Organization
  • Despoina Koulenti
    • Department of Critical CareUniversity Hospital ATTIKON, Medical School University of Athens
  • Kevin Laupland
    • Albert Bonniot Institute, Team 11: Outcome of Mechanically Ventilated Patients and Respiratory CancersUniversité Grenoble 1, U 823
    • Peter Lougheed Centre, Department of Critical Care MedicineUniversity of Calgary and Alberta Health Services
  • Benoit Misset
    • Paris Sorbonne Cité, Medical Surgical ICU, Groupe Hospitalier Paris Saint-JosephUniversité Paris Descartes
  • Jordi Valles
    • Critical Care CenterHospital Sabadell
  • Frederico Bruzzi de Carvalho
    • Centro De Terapia IntensivaHospital Mater Dei
  • José Artur Paiva
    • Emergency and Intensive Care UnitHospital de S. Joao
  • Nahit Çakar
    • Department of Anaesthesiology and Intensive CareIstanbul University and Istanbul Medical School
  • Xiaochun Ma
    • Department of Critical Care MedicineThe First Affiliated Hospital of China Medical University
  • Philippe Eggimann
    • Adult Critical Care Medicine and Burn CentreCentre Hospitalier Universitaire Vaudois
  • Massimo Antonelli
    • Department of Intensive Care and AnaesthesiologyPoliclinico Universitario A
  • Marc J. M. Bonten
    • Department of Medical Microbiology, Julius Center for Health Sciences and Primary CareUniversity Medical Center Utrecht
  • Akos Csomos
    • Surgical Intensive CareSemmelweis University
  • Wolfgang A. Krueger
    • Anaesthesiology and Intensive Care MedicineClinics of Constance
  • Adam Mikstacki
    • Department of Anaesthesiology and Intensive TherapyPoznan University of Medical Sciences Regional Hospital
  • Jeffrey Lipman
    • Department of Intensive Care MedicineRoyal Brisbane and Women’s Hospital
  • Pieter Depuydt
    • Department of Intensive CareGhent University Hospital
  • Aurélien Vesin
    • Albert Bonniot Institute, Team 11: Outcome of Mechanically Ventilated Patients and Respiratory CancersUniversité Grenoble 1, U 823
    • Outcomerea Organization
  • Maité Garrouste-Orgeas
    • Albert Bonniot Institute, Team 11: Outcome of Mechanically Ventilated Patients and Respiratory CancersUniversité Grenoble 1, U 823
    • Outcomerea Organization
    • Paris Sorbonne Cité, Medical Surgical ICU, Groupe Hospitalier Paris Saint-JosephUniversité Paris Descartes
  • Jean-Ralph Zahar
    • Outcomerea Organization
    • Infection Control UnitNecker University Hospital
  • Stijn Blot
    • General Internal Medicine and Infectious DiseasesGhent University Hospital
  • Jean Carlet
    • Consultant for WHO
  • Christian Brun-Buisson
    • Medical Intensive Care UnitHenri Mondor Teaching Hospital
  • Claude Martin
    • Department of Anesthesia and Intensive Care and Trauma CenterMarseilles University Hospital
  • Jordi Rello
    • Critical Care DepartmentVall d’Hebron University Hospital
  • Georges Dimopoulos
    • Department of Critical CareUniversity Hospital ATTIKON, Medical School University of Athens
    • Medical ICUAlbert Michallon University Hospital, Université Grenoble 1
    • Albert Bonniot Institute, Team 11: Outcome of Mechanically Ventilated Patients and Respiratory CancersUniversité Grenoble 1, U 823
    • Outcomerea Organization
Special Article

DOI: 10.1007/s00134-012-2695-9

Cite this article as:
Tabah, A., Koulenti, D., Laupland, K. et al. Intensive Care Med (2012) 38: 1930. doi:10.1007/s00134-012-2695-9

Abstract

Purpose

The recent increase in drug-resistant micro-organisms complicates the management of hospital-acquired bloodstream infections (HA-BSIs). We investigated the epidemiology of HA-BSI and evaluated the impact of drug resistance on outcomes of critically ill patients, controlling for patient characteristics and infection management.

Methods

A prospective, multicentre non-representative cohort study was conducted in 162 intensive care units (ICUs) in 24 countries.

Results

We included 1,156 patients [mean ± standard deviation (SD) age, 59.5 ± 17.7 years; 65 % males; mean ± SD Simplified Acute Physiology Score (SAPS) II score, 50 ± 17] with HA-BSIs, of which 76 % were ICU-acquired. Median time to diagnosis was 14 [interquartile range (IQR), 7–26] days after hospital admission. Polymicrobial infections accounted for 12 % of cases. Among monomicrobial infections, 58.3 % were gram-negative, 32.8 % gram-positive, 7.8 % fungal and 1.2 % due to strict anaerobes. Overall, 629 (47.8 %) isolates were multidrug-resistant (MDR), including 270 (20.5 %) extensively resistant (XDR), and 5 (0.4 %) pan-drug-resistant (PDR). Micro-organism distribution and MDR occurrence varied significantly (p < 0.001) by country. The 28-day all-cause fatality rate was 36 %. In the multivariable model including micro-organism, patient and centre variables, independent predictors of 28-day mortality included MDR isolate [odds ratio (OR), 1.49; 95 % confidence interval (95 %CI), 1.07–2.06], uncontrolled infection source (OR, 5.86; 95 %CI, 2.5–13.9) and timing to adequate treatment (before day 6 since blood culture collection versus never, OR, 0.38; 95 %CI, 0.23–0.63; since day 6 versus never, OR, 0.20; 95 %CI, 0.08–0.47).

Conclusions

MDR and XDR bacteria (especially gram-negative) are common in HA-BSIs in critically ill patients and are associated with increased 28-day mortality. Intensified efforts to prevent HA-BSIs and to optimize their management through adequate source control and antibiotic therapy are needed to improve outcomes.

Keywords

Hospital acquired bloodstream infections Critically ill patients Antibiotic therapy Prognosis Multilevel models Extensively resistant bacterias

Introduction

Bloodstream infection (BSI) is an important cause of severe sepsis and septic shock that is associated with high resource utilisation, morbidity and mortality [14]. Hospital-acquired bloodstream infection (HA-BSI) is recognised as a major patient-safety concern and a marker of quality of care [5]. Patients admitted to intensive care units (ICUs) have multiple risk factors for HA-BSIs including severe acute illness, co-morbidities and frequent use of invasive devices [6]. Many changes have occurred in the epidemiology of BSI in recent years, particularly with the emergence of drug-resistant organisms, which has increased the treatment-failure rate and the risk of adverse patient outcomes [2, 6]. Although hospital-acquired infections in critically ill patients have been the focus of numerous reports worldwide [28], there is a paucity of contemporary multinational data on the epidemiology and outcome determinants of HA-BSIs in ICU patients [2, 4, 9].

The objective of this study is to describe the epidemiology of HA-BSIs treated in the ICU and to identify determinants of treatment failure and outcomes in Europe and internationally.

Methods

This study is reported in accordance with the STrengthening the Reporting of OBservational studies in Epidemiology (STROBE) guidelines [10].

A prospective observational multicentre international cohort design was used. All participating ICUs obtained approval from their local ethics committees. The pre-defined goal was to include 1,100 ICU patients with HA-BSI.

Study protocol and definitions

Patients were enrolled if they had a new diagnosis of HA-BSI and were admitted to an ICU. The study focussed on the first episode of HA-BSI. Detailed definitions of HA-BSI are provided in the Electronic Supplement.

Data collected for each patient included the dates and times of collection and positivity of the first positive blood culture; source of infection; presence of sepsis; severity of illness; co-morbidities; and management including source control, antimicrobial drugs and adjunctive treatments. All study data were obtained from patient files, and no additional tests were performed for the purpose of the study. Severity of illness was defined at ICU admission and at HA-BSI diagnosis using SAPS II and Sequential organ failure assessment (SOFA) scores [11], respectively. Co-morbidities were assessed using the Charlson index and the five markers of the Chronic Health Evaluation of the APACHE II score reported by Knaus et al. [12, 13].

Clinical variables and relapses or new episodes of HA-BSI were recorded until ICU discharge, and all-cause mortality within 28 days since first positive blood culture was ascertained.

We recorded information on each ICU including type of hospital and ICU, number of beds and patients, and mortality rate in the previous year (2008). We also recorded factors possibly associated with antimicrobial use such as the availability of an infectious diseases specialist, procedures for antibiotic use and infection control protocols.

The organisms causing HA-BSI and their antimicrobial susceptibility test results were recorded according to local policies. Detailed definitions of adequate antimicrobial therapy are given in the Electronic Supplement.

Drug-resistant organisms were classified as susceptible (SUS), multidrug resistant (MDR), extensively resistant (XDR) and pan-drug resistant (PDR) according to the European Society of Clinical Microbiology and Infectious Diseases (ESCMID) [14]. Each category was included within the previous category: all PDR organisms were XDR, and all XDR organisms were MDR. Definitions and worksheets can be downloaded from the European Centre for Disease Prevention and Control (ECDC) website [15]. According to this classification scheme, bacteria other than enterococci, Staphylococcus aureus, Enterobacteriaceae and non-fermenting gram-negative bacteria were considered susceptible. We arbitrarily classified yeasts among MDR organisms on the basis of their non-sensitivity to first-line empirical antimicrobial therapies.

Data management and statistical analysis

In each study ICU, data were entered into a password-protected and secured web-based server. Online data were managed in MySQL 5.1.41 (Oracle, Redwood Shores, CA, USA) and downloaded to SAS 9.2 (SAS Institute, Cary, NC, USA) for offline management and analysis. The electronic case-report form was developed locally using open-source software (PHP/MySQL) with the primary goal of ensuring easy and consistent data entry, data verification and easy communication between the coordinating centre (Albert Bonniot Institute, University Joseph Fourier, Grenoble, France) and each study ICU. At the coordinating centre, two investigators (A.T., D.K.) routinely checked the data for completeness and for consistency in definition use. All missing, extreme or implausible values were sent back to the study-ICU investigators for review. Where data could not be confirmed or remained questionable, the primary author (A.T.) made a final adjudication about study inclusion, in agreement with the main investigator (J.-F.T.). Doubtful cases were reviewed by two other investigators (J.-R.Z. and K.L.). Missing data were replaced by the median value, and missing times of sampling and of blood culture positivity by 12:00 a.m.

The statistical analyses considered only the first episode of HA-BSI, as information was fully collected only for these occurrences.

Means with standard deviation (SD) were used to describe normally or near-normally distributed continuous data and were compared using the t test. Medians with interquartile range (IQR) were computed for skewed data and were compared using the Mann–Whitney test. Fisher’s exact test or the chi-square test was performed to compare categorical data.

In multivariable analyses, variables were organised into three tiers: country, ICU and patient. To identify factors associated with 28-day mortality, we built a three-tiered hierarchical logistic mixed model using the GLIMMIX procedure of the SAS software. The influence of country-based and ICU-based variables on the outcome was included through both fixed and random effects. Multilevel modelling takes into account the hierarchical structure of the data, which may manifest as intra-class correlations. To obtain a conservative estimate of the standard error, a separate random-error term should be specified for each level of the analysis [16]. Therefore, to avoid overestimating the significance of risk factors for death by day 28, we took intra-class correlations into account, and we specified a separate random-error term for each tier. Variables potentially associated with 28-day mortality (p values less than 0.10 on univariate analysis) were introduced into the multivariable model and selected using a backward approach. Two-way clinically relevant interactions were tested in the final model. In all analyses, two-sided p values less than 0.05 were deemed statistically significant. No correction for multiple testing was performed.

Results

We enrolled 1,156 patients from 162 ICUs in 120 cities in 24 countries (Fig. 1). Sixty-three ICUs accepted the study but did not participate, and 16 ICUs (30 patients) were excluded because of incomplete data. Of the 1,227 patients included initially, 71 were excluded prior to data analysis (34 incomplete files, 8 patients with community-acquired BSI and 29 patients with cultures positive for skin contaminants), leaving 1,156 patients for the study.
https://static-content.springer.com/image/art%3A10.1007%2Fs00134-012-2695-9/MediaObjects/134_2012_2695_Fig1_HTML.gif
Fig. 1

Geographic distribution of included patient and antimicrobial resistance categories

The study patients were located in many different geographic regions, although they predominated in European countries, particularly France (17.8 %) and Greece (24.9 %) (Fig. 1). No country-related factors were significantly associated with 28-day mortality (Table E1, Electronic Supplement).

Most of the ICUs (64 %) were university or university-affiliated. An infectious disease specialist was available in 86 % of the ICUs. Written antibiotic procedures were available in 71 % of ICUs and followed strictly in 23 %. Further details on ICU-related prognosis factors are presented in Table 1.
Table 1

Associations between ICU characteristics and 28-day mortality

Variable

Characteristics

All ICUs (N = 162)

All patients (N = 1,156)

Alive on D28 (N = 743)

Dead on D28 (N = 413)

p-Valuea

Academic status of hospital

University

74 (45.7)

594 (51.4)

386 (49.5)

208 (50.4)

0.9

University affiliated

30 (18.5)

199 (17.2)

128 (17.2)

71 (17.2)

 

General

58 (35.8)

363 (31.4)

229 (30.8)

134 (32.4)

 

Funding of hospital

Public

145 (89.5)

1,042 (90.1)

670 (90.2)

372 (90)

1

Private

17 (10.5)

114 (9.9)

73 (9.8)

41 (9.9)

 

ICU recruitment

Medical

30 (18.5)

212 (18.3)

137 (18.4)

75 (18.1)

0.4

Surgical

16 (9.9)

78 (6.7)

60 (8)

18 (4.3)

 

Medical + surgical

107 (66)

807 (69.8)

509 (68.5)

298 (72.1)

 

Cardiac

3 (1.9)

14 (1.2)

9 (1.2)

5 (1.2)

 

Other

6 (3.7)

45 (3.9)

28 (3.8)

17 (4.1)

 

Type of ICU

Open

49 (30.2)

308 (26.6)

201 (27)

107 (25.9)

0.7

Closed

113 (69.8)

848 (73.4)

542 (72.9)

306 (74.1)

 

Number of ICU beds

 

12 [8]

14 [10]

14 [10]

14 [10]

0.5

Number of admissions/year

 

597 [458]

600 [748]

600 [750]

618 [737]

1

Number of admissions/bed/year

≤40

80 (49.4)

551 (47.7)

367 (49.4)

184 (44.6)

0.14

>40

82 (50.6)

605 (52.3)

376 (50.6)

229 (55.4)

 

Number of intensivists

 

6 [6]

7 [6]

7 [6]

7 [6]

0.5

Patient-to-nurse ratio

 

2 [1]

2.4 [1]

2.3 [1]

2.5 [1]

0.3

Overall ICU mortality rate in the past year (2008)

 

20 [13]

20 [13.4]

20 [13]

20.3 [12]

0.045

Availability of an infectious diseases specialist

No

22 (13.6)

139 (12)

84 (11.3)

55 (13.3)

0.3

Consultant only

87 (53.7)

587 (50.8)

369 (49.7)

218 (52.8)

 

Yes

53 (32.7)

430 (37.2)

290 (39)

140 (33.9)

 

Availability of an antibiotic committee

No

31 (19.1)

233 (20.2)

160 (21.5)

73 (17.7)

0.18

Yes

131 (80.9)

923 (79.8)

583 (78.5)

340 (82.3)

 

Written antibiotic procedures

No

47 (29)

307 (26.6)

195 (26.2)

112 (27.1)

0.17

Yes, not strict

78 (48.1)

594 (51.4)

399 (53.7)

195 (47.2)

 

Yes, strict

37 (22.8)

255 (22.1)

149 (20)

106 (25.7)

 

Infection prevention/control protocols

No

25 (15.4)

146 (12.6)

94 (12.6)

52 (12.6)

1.0

Yes, not strict

64 (39.5)

495 (42.8)

318 (42.8)

177 (42.9)

 

Yes, strict

73 (45.1)

515 (44.6)

331 (44.5)

184 (44.6)

 

Microbiological lab

24 h, in hospital

105 (64.8)

811 (70.2)

518 (69.7)

293 (70.9)

0.6

24 h, outside hospital

12 (7.4)

61 (5.3)

42 (5.6)

19 (4.6)

 

Day only, in hospital

43 (26.5)

268 (23.2)

175 (23.5)

93 (22.5)

 

No

2 (1.2)

16 (1.4)

8 (1)

8 (1.9)

 

Catheter tip culture

No

10 (6.2)

60 (5.2)

38 (5.1)

22 (5.3)

0.4

Yes always

82 (50.6)

600 (51.9)

403 (54.2)

197 (47.7)

 

Yes, unless deceased

11 (6.8)

86 (7.4)

54 (7.2)

32 (7.7)

 

Yes, only for sepsis

59 (36.4)

410 (35.5)

248 (33.4)

162 (39.2)

 

Results are shown as n (%), median [IQR] where applicable

aHierarchical logistic regression with random effect for centre and country

A total of 1,317 bloodstream isolates were cultured from the 1,156 study patients. A single organism was found in 1,016 (88 %) patients, two organisms in 120 (10 %) patients, three organisms in 19 (2 %) patients and four organisms in 1 (<1 %) patient.

HA-BSIs were diagnosed a median of 14 (IQR, 7–26) days after hospital admission. Among them, 76 % (877/1,156) were diagnosed in the ICU, with a median time from ICU admission to diagnosis of 8 (IQR, 3–16) days. Timing of acquisition had no significant influence on mortality. Patient characteristics are reported in Table 2 and details on the infections and treatment in Table 3.
Table 2

Associations between baseline (admission to the ICU) patient characteristics and 28-day mortality

Variable

Characteristics

All patients (n = 1,156)

Alive on D28 (n = 743)

Dead on D28 (n = 413)

p-Valuea

Age, years

 

59.5 ± 17.7

57.2 ± 18.3

63.7 ± 15.8

<0.0001

SAPS II

 

50 ± 17

47 ± 15

56 ± 17

<0.0001

Obesity (BMI >30 kg/m2)

 

251 (22.1)

175 (69.7)

76 (30.3)

0.08

Malnutrition (BMI <18.5 kg/m2)

 

43 (3.8)

24 (55.8)

19 (44.2)

0.18

Male

 

756 (65.4)

503 (66.5)

253 (33.5)

0.03

Charlson co-morbidity index

0

386 (33.4)

296 (76.7)

90 (23.3)

<0.0001

 

1–2

404 (34.9)

263 (65.1)

141 (34.9)

 
 

3+

366 (31.7)

184 (50.3)

182 (49.7)

 

Chronic illnesses

 Respiratory

 

98 (8.5)

50 (51)

48 (49)

0.008

 Cardiovascular

 

117 (10.1)

71 (60.7)

46 (39.3)

0.5

 Renal

 

61 (5.3)

34 (55.7)

27 (44.3)

0.19

 Hepatic

 

44 (3.8)

21 (47.7)

23 (52.3)

0.017

 Immunosuppression

 

151 (13.1)

68 (45)

83 (55)

<0.0001

 Medical admission

 

672 (58.1)

398 (59.2)

274 (40.8)

0.0003

Organ dysfunctions at admission

 Neurological

 

354 (30.6)

236 (66.7)

118 (33.3)

0.5

 Haemodynamic

 

591 (51.1)

342 (57.9)

249 (42.1)

<0.0001

 Respiratory

 

937 (81.1)

586 (62.5)

351 (37.5)

0.009

 Renal

 

155 (13.4)

86 (55.5)

69 (44.5)

0.01

 No organ failure on admission

 

126 (10.9)

95 (75.4)

31 (24.6)

0.006

 Septic shock at admission

 

267 (23.1)

139 (52.1)

128 (47.9)

<0.0001

BMI body mass index. Results are shown as n (%), mean ± SD where applicable

aHierarchical logistic regression with random effect for centre and country

Table 3

Patient characteristics at the diagnosis of hospital-acquired bloodstream infection

Variable

Characteristics

All patients (N = 1,156)

Alive on D28 (N = 743)

Dead on D28 (N = 413)

p-Valuea

Time from ICU admission to HA-BSI

Hospital-acquired, before ICU admission

279 (24.1)

168 (60.2)

111 (39.8)

0.01

ICU-acquired, ≤7 days

281 (24.3)

186 (66.2)

95 (33.8)

ICU-acquired, >7 days

596 (51.6)

389 (65.3)

207 (34.7)

Sepsis syndrome

SIRS/SEPSIS

150 (13)

110 (73.3)

40 (26.7)

<0.0001

Severe sepsis

476 (41.2)

347 (72.9)

129 (27.1)

Septic shock

530 (45.8)

286 (54)

244 (46)

SIRS score

0

24 (2.1)

15 (62.5)

9 (37.5)

0.7

1

88 (7.6)

62 (70.5)

26 (29.5)

2

272 (23.5)

173 (63.6)

99 (36.4)

3

426 (36.9)

276 (64.8)

150 (35.2)

4

346 (29.9)

217 (62.7)

129 (37.3)

Fever (>38.3 °C or 101 °F)

Yes

724 (62.6)

486 (67.1)

238 (32.9)

0.02

SOFA score

 

7 [6]

4 [5]

10 [6]

<0.0001

Mechanical ventilation

 

1,030 (89.1)

654 (63.5)

376 (36.5)

0.006

Hypotension

 

583 (50.4)

317 (54.4)

266 (45.6)

<0.0001

Gram-negative bacteriad

 

705 (61)

458 (65)

247 (35)

0.4

Gram-positive bacteriad

 

416 (36)

277 (66.6)

139 (33.4)

0.3

Fungusd

 

96 (8.3)

57 (59.4)

39 (40.6)

0.2

Anaerobesd

 

19 (1.6)

9 (47.4)

10 (52.6)

0.13

Susceptible organisms

 

570 (49.3)

401 (70.4)

169 (29.6)

<0.0001

MDR

 

586 (50.7)

342 (58.4)

244 (41.6)

<0.0001

XDR

 

254 (22)

151 (59.4)

103 (40.6)

0.09

Infection source

No clear source

274 (23.7)

181 (66.1)

93 (33.9)

0.0003

Catheter-related

247 (21.4)

171 (69.2)

76 (30.8)

Intra-abdominal

134 (11.6)

64 (47.8)

70 (52.2)

Respiratory tract

244 (21.1)

145 (59.4)

99 (40.6)

Urinary tract

47 (4.1)

35 (74.5)

12 (25.5)

Other

73 (6.3)

55 (75.3)

18 (24.7)

Multiple sources

149 (12.9)

92 (67.2)

45 (32.8)

Antimicrobial therapyb

Inadequate

154 (13.3)

79 (51.3)

75 (48.7)

0.0007

Adequate

1,002 (86.7)

664 (66.3)

338 (33.7)

Time to adequate treatment

<4 h

608 (52.6)

394 (64.8)

214 (35.2)

0.003

24–48 h

166 (14.4)

109 (65.7)

57 (34.3)

Day 2–day 5

228 (19.7)

161 (70.6)

67 (29.4)

>Day 5 or never

154 (13.3)

79 (51.3)

75 (48.7)

Source control

Not required

644 (55.7)

409 (63.5)

235 (36.5)

<0.0001

Done

467 (40.4)

324 (69.4)

143 (30.6)

Required, not done

45 (3.9)

10 (22.2)

35 (77.8)

Time to source control (days)

 

1.0 [2.5]

1 [2.8]

0.8 [2.0]

0.01

Data are n (%) or median [IQR]

SIRS Systemic inflammatory response syndrome

aHierarchical logistic regression including random effects for country and centres

bDenotes no adequate anti-microbial treatment within 5 after collection of the first positive blood culture

cPolymicrobial episodes are counted in more than one category

The most common organisms causing HA-BSI and their resistance patterns are presented in Table 4. Of monomicrobial infections (n = 1,016), 592 (58.3 %) were gram-negative, 333 (32.8 %) gram-positive, 79 (7.8 %) fungal and 12 (1.2 %) due to strict anaerobes.
Table 4

Isolates found in hospital-acquired bloodstream infections in patients in intensive care units

 

Susceptible, n (%)

MDR,a n (%)

XDR,a n (%)

PDR,a n (%)

Total

28-day mortality

Gram-negative

    

759 (57.6 %)

264 (34.8 %)

 Acinetobacter spp.

13 (8.1 %)

147 (91.9 %)

114 (71.3 %)

1 (0.6 %)

160 (12.2 %)

55 (34.4 %)

 Klebsiella spp.

46 (29.5 %)

110 (70.5 %)

76 (48.7 %)

3 (1.9 %)

156 (11.9 %)

52 (33.3 %)

 Pseudomonas spp.

95 (63.3 %)

55 (36.7 %)

41 (27.3 %)

1 (0.7 %)

150 (11.4 %)

60 (40 %)

 Escherichia coli

57 (58.2 %)

41 (41.8 %)

5 (5.1 %)

0 (0 %)

98 (7.4 %)

34 (34.7 %)

 Enterobacter spp.

48 (54.6 %)

40 (45.5 %)

17 (19.3 %)

0 (0 %)

88 (6.7 %)

29 (33 %)

 Other gram-negative

69 (64.5 %)

38 (35.5 %)

15 (14.0 %)

0 (0 %)

107 (8.1 %)

34 (31.8 %)

Gram-positive

    

440 (33.4 %)

149 (33.9 %)

 Enterococcus spp

103 (71.5 %)

41 (28.5 %)

2 (1.4 %)

0 (0 %)

144 (10.9 %)

61 (42.4 %)

 Coagulase-negative staphylococci and other staphylococci

141 (100 %)

0 (0 %)

0 (0 %)

0 (0 %)

141 (10.7 %)

41 (29.1 %)

 Staphylococcus aureus

60 (50.4 %)

59 (49.6 %)

0 (0 %)

0 (0 %)

119 (9 %)

37 (31.1 %)

 Other gram-positive

36 (100 %)

0 (0 %)

0 (0 %)

0 (0 %)

36 (2.7 %)

10 (27.8 %)

Anaerobes

    

20 (1.5 %)

10 (50 %)

 Bacteroides spp.

13 (100 %)

0 (0 %)

0 (0 %)

0 (0 %)

13 (1 %)

6 (46.2 %)

 Other anaerobes

7 (100 %)

0 (0 %)

0 (0 %)

0 (0 %)

7 (0.5 %)

4 (57.1 %)

Fungi

    

98 (7.4 %)

40 (40.8 %)

 Candida albicans

0 (0 %)

56 (100 %)

0 (0 %)

0 (0 %)

56 (4.3 %)

23 (41.1 %)

 Candida non-albicans

0 (0 %)

39 (100 %)

0 (0 %)

0 (0 %)

39 (3 %)

14 (35.9 %)

 Other

0 (0 %)

3 (100 %)

0 (0 %)

0 (0 %)

3 (0.2 %)

3 (100 %)

 Total (patient)b

570 (49.3 %)

586 (50.7 %)

254 (22 %)

5 (0.43 %)

1,156

413 (35.7 %)

 Total (micro-organisms)

688 (52.2 %)

629 (47.8 %)

270 (20.5 %)

5 (0.38 %)

1,317

 

Percentages of SUS susceptible, MDR multidrug-resistant, XDR extensively drug resistant and PDR pan-drug-resistant strains of each pathogen are shown. The “Total” column shows the percentage of each pathogen in the cohort

aEach category is included within the previous category: all PDR organisms are XDR, and all XDR organisms are MDR

bIn case of BSI due to more than one micro-organism, the most resistant one was taken into account to classify the patient

Carbapenem resistance was found in 110/166 (69 %) Acinetobacter spp., 59/156 (38 %) Klebsiella pneumoniae, 56/150 (37 %) Pseudomonas spp., 5/88 (5.7 %) Enterobacter spp. and 1/98 (1 %) Escherichia coli. Of the 119 S. aureus isolates, 57 (48 %) were methicillin-resistant; and of the 70 Enterococcus faecium isolates, 16 (23 %) were vancomycin-resistant (VRE). Isolates acquired in the ICU were more often drug resistant compared with other hospital-acquired isolates (413/992, 42 % versus 82/325, 25 %; p < 0.0001). The distribution of organisms differed significantly (p < 0.001) by country (Table E2, Electronic Supplement).

Among the 1,156 study patients, 608 (52.6 %) received adequate antibiotic therapy before or within 24 h following collection of the first positive blood culture, whereas 154 (13.3 %) did not receive adequate antibiotic therapy within 5 days or before ICU discharge or death. Patients who received adequate antimicrobials only after the fifth day had lower SOFA scores. Late adequate therapy was also associated with a longer time to blood-culture positivity (median [IQR], 100 h [23–144]). Resistance rates showed a significant positive association with failure to receive adequate antibiotic therapy. As shown in Fig. 2, time to adequate antibiotic therapy increased with antimicrobial resistance (chi-square for trends, p < 0.001).
https://static-content.springer.com/image/art%3A10.1007%2Fs00134-012-2695-9/MediaObjects/134_2012_2695_Fig2_HTML.gif
Fig. 2

Cumulative percentage of patients receiving at least one adequate antimicrobial, on each calendar day before and after the date of collection of the first positive blood culture, shown by antimicrobial resistance status (trend chi-square, p < 0.001). The calendar day (without details on time) was available in 46 patients for blood collection and 67 patients for treatment initiation; a time of 12:00 a.m. was assigned in these patients

The most frequently prescribed empirical antimicrobials and their adequacy against the causative organism are shown in Fig. 3a, and the first adequate treatment is shown in Fig. 3b. Carbapenems were the antimicrobials most frequently given in the first 24 h (n = 212; 19 %), followed by glycopeptides (n = 172; 15 %) and piperacillin-tazobactam (n = 152; 13 %). Treatment was adequate in the first 24 h with monotherapy in 339 (29 %) patients and combined therapy in 269 (23 %) patients. The 548 (47 %) remaining patients did not receive adequate antimicrobials in the first 24 h.
https://static-content.springer.com/image/art%3A10.1007%2Fs00134-012-2695-9/MediaObjects/134_2012_2695_Fig3_HTML.gif
Fig. 3

Antibiotic use. a Most frequent empiric antimicrobials, reported as the number of patients receiving the antimicrobial in the 12 h before, to 24 h after collection of the first positive blood culture. b Antimicrobials most frequently used in the 5 days following HA-BSI diagnosis, reported as the number of patients receiving the treatment. “Adequate” and “inadequate” show the percentage of cases in which the organism recovered from blood cultures was susceptible to each molecule

The 28-day all-cause mortality rate was 413/1,156 (35.7 %), with 381/413 (92 %) deaths occurring in the ICU. On day 28, 295/743 (40 %) survivors were still in the ICU. Among the 1,156 patients, 194 (17 %) had second and 44 (4 %) had third episodes of HA-BSI identified within the 28-day follow-up period. Patient- and ICU-related factors included in the univariate analysis are presented in Tables 13.

On multivariable analysis (Table 5), factors significantly associated with higher 28-day mortality were turn-over above the median value, higher mean ICU mortality rate in the previous year (2008) and several patient-related factors (Table 5). 28-day mortality was not significantly influenced by factors that may reflect antimicrobial stewardship (availability of an infectious diseases specialist, microbiological laboratory, written antibiotic protocols and/or infection-control protocols). 28-day mortality was significantly higher in older patients and in patients with chronic respiratory disease or immune deficiency. Intensity of the host response and organ dysfunctions as reflected by septic shock or a higher SOFA score at HA-BSI onset was an independent risk factor for 28-day mortality. The source of infection significantly affected 28-day mortality when introduced as a multiple-class variable (catheter, intra-abdominal, respiratory, urinary, multiple and other focus) (p = 0.02, data not shown). Only abdominal source of infection was associated with 28-day mortality at the final step of variable selection. 28-day mortality was significantly higher in the event of inadequate antibiotic therapy or failure to control the infection source if required (absence of catheter removal, n = 12; surgical treatment for abdominal source of infection, n = 22; other sources, n = 6; multiple possible foci, n = 5).
Table 5

Hierarchical logistic regression model of the effect of patient- and centre-related variables on 28-day mortality following hospital-acquired bloodstream infection

Variable

Estimate (SE)

OR [95 %CI]

p-Valuea

Centres

   

 ICU mortality in 2008 (per percentage point)

0.02 (0.01)

1.02 [1.00–1.04]

0.03

 Turn-over >40 admissions/bed/year (median ICU turn-over)

0.49 (0.21)

1.64 [1.08–2.49]

0.02

Patients

   

 Female

0.29 (0.15)

1.34 [1.01–1.79]

0.046

 Medical patient

0.34 (0.15)

1.40 [1.04–1.90]

0.03

 Age (per year)

0.02 (0.005)

1.02 [1.01–1.03]

0.0002

 SAPS II (per point)b

0.01 (0.005)

1.01 [1.00–1.02]

0.01

 Chronic respiratory disease

0.56 (0.24)

1.75 [1.09–2.80]

0.02

 Chronic immunological disease

0.75 (0.21)

2.11 [1.40–3.19]

0.0004

 SOFA without cardiovascular points (per point)c

0.18 (0.03)

1.20 [1.14–1.26]

<0.0001

 Septic shock

0.38 (0.15)

1.46 [1.09–1.96]

0.01

 Intra-abdominal source

0.48 (0.21)

1.61 [1.07–2.44]

0.023

 Organism resistancec

  

0.043

 Non-resistant organisms

Ref.

  

 MDR

0.40 (0.17)

1.49 [1.07–2.06]

 

 XDR/PDR

0.35 (0.20)

1.42 [0.95–2.11]

 

 Timing to adequate treatment

  

<0.0001

 Never

Ref.

  

 Before day 6d

−0.97 (0.25)

0.38 [0.23–0.63]

 

 Since day 6d

−1.63 (0.45)

0.20 [0.08–0.47]

 

 Source control

  

0.0003

 Achieved

Ref

1

 

 Not required

0.12 (0.15)

1.12 [0.83–1.52]

 

 Required, not achieved

1.77 (0.44)

5.86 [2.5–13.9]

 

Co-variance parameters

   

 Country

0 (–)e

 

 Centre

0.53 (0.16)

 

0.001

SE standard error, OR odds ratio, 95 %CI 95 % confidence interval

MDR multidrug resistant, XDR extensively drug resistant, PDR pan-drug resistant (see Electronic Supplement for definitions). In patients with multiple organisms, the worst resistance status was considered

aHierarchical logistic model with random effects for both country and centre

bSAPS II score at ICU admission

cSOFA score at the onset of hospital-acquired bloodstream infection

dThe reference day if the day of collection of the first positive blood culture

eCo-variance parameter is zero

Presence of an MDR/XDR/PDR organism was associated with a longer time to adequate antimicrobial therapy and with an increase in 28-day mortality. XDR or PDR resistance levels were not associated with higher 28-day mortality when compared with MDR levels. The results remained similar after exclusion of fungal HA-BSIs from the analysis (Table E3, Electronic Supplement) and when considering patients with a single episode of HA-BSI (Table E4, Electronic Supplement). The impact of inadequate antimicrobial treatment was not significantly different between episodes with and without MDR strains (p = 0.54).

Discussion

EUROBACT provides a contemporary analysis of the prognostic factors of HA-BSI among patients admitted to ICUs internationally. There was a predominance of patients included in France, Greece and southern Europe. Drug resistance rates were very high overall, and MDR was particularly common among gram-negative pathogens. Piperacillin + tazobactam, carbapenems, and glycopeptides were extensively used both as initial empirical drugs and once the culture results were available. Mortality was higher in ICUs with higher turn-over rates. Inadequate antibiotic treatment and failure to control the source of infection were both associated with 28-day mortality, independently from age, chronic co-morbidities, severity of acute illness, shock and organ dysfunctions. Antimicrobial resistance was associated with a significantly longer time to adequate antimicrobial treatment and with a higher risk of death, even after controlling for adequacy of antimicrobial treatment.

The epidemiology of BSIs in ICU patients has changed over time. Gram-positive bacteria and yeasts have become major causes of BSI in the last two decades [1, 17]. MDR gram-negative bacteria are re-emerging [18], as confirmed by the present report. The most frequent pathogens in EUROBACT were Acinetobacter, Klebsiella and Pseudomonas spp., followed by enterococci, coagulase-negative staphylococci and S. aureus. The high prevalence of Acinetobacter in our study is probably related to the overrepresentation of southern European countries and may not reflect the situation worldwide. We observed significant variability in the distributions of organism groups and resistance patterns across countries, in accordance with European surveillance reports [15]. MDR organisms were ubiquitous, and three-quarters of the countries reported at least one XDR organism. These data are consistent with earlier reports [15]. Thus, XDR Acinetobacter baumanii and P. aeruginosa and carbapenemase-producing K. pneumoniae have been reported in southern Europe, as well as in South America and Asia, and have shown a tendency to spread rapidly throughout the world [18]. Moreover, with the expansion of international travel [19], no country is exempt from the risk of a major XDR outbreak.

The relationship between promptness of adequate therapy and prognosis is complex. Physicians prescribe early extended-spectrum antimicrobial therapy to patients with severe clinical presentations. In contrast, antimicrobial therapy is often delayed in patients without organ dysfunctions or with long times to blood-culture positivity. Finally, patients who die very early never receive antimicrobial therapy. These considerations explain why treatment adequacy was unrelated to outcomes in recent cohort studies [9, 20, 21]. Observational studies cannot provide proof of a causal relationship between treatment adequacy and outcome. In the present study, we found an association between time to adequate treatment and mortality even after adjustment for acute-illness severity at BSI onset. Further multistate models taking into account adequate antimicrobial therapy as a non-absorbing state and severity of the BSI are needed to definitely estimate the impact of delayed therapy. Although fluid resuscitation and adequate treatment of organ dysfunctions are key components of critical care, prompt adequate antimicrobial treatment remains a cornerstone of BSI management in the ICU [22].

Many studies, including the present one, have found that bacterial resistance decreased the chance of early adequate therapy [23, 24]. Bacterial resistance was also associated with mortality in recent, large, well-conducted epidemiological studies [2, 25]. However, when both bacterial resistance and adequacy of treatment are taken into account, treatment inadequacy is seen to make a larger contribution to the mortality increase than bacterial resistance [9, 26, 27]. One major finding from our study is the significant impact of MDR infections on patient outcome, even after adjustment on antimicrobial treatment adequacy, source control and all other ICU- and patient-related determinants of 28-day mortality. Interestingly, XDR infection did not have a greater influence on 28-day mortality than MDR infection, despite a further increase in time to adequate therapy. This finding is consistent with experimental studies suggesting that resistance to antimicrobial agents may be associated with decreases in bacterial fitness, metabolic activity [28] or virulence [29].

The strong relationship between absence of source control and mortality in our study further supports the widely recognised importance of source control in patients with infection or sepsis [22]. To the best of our knowledge, although this importance is widely accepted by clinicians, it has not been demonstrated in previous studies, except in necrotising fasciitis [30].

In the present study, an intra-abdominal source was an independent risk factor for mortality, in keeping with previously reported evidence that intra-abdominal sources of bacteraemia were associated with higher mortality [26, 31].

The influence of centre-related characteristics on patient outcomes deserves some commentary. In ICUs with more than 40 admissions/bed/year, mortality rates in patients with HA-BSI were higher. This finding may reflect delayed diagnosis, inappropriate symptomatic or aetiological treatment, unidentified factors or chance alone. Among the ICU-related factors that might influence antimicrobial use, availability of an infectious diseases specialist, antibiotic committee or written antibiotic therapy procedures did not significantly influence 28-day mortality in our study. The effectiveness of antibiotic stewardship programs in ICUs remains debated [32]. In our studies, the large proportion of ICUs having at least one physician with special training in infectious diseases may have limited our ability to detect a positive impact of antibiotic stewardship programs on patient outcomes. Studies have established that reducing antimicrobial pressure via antimicrobial stewardship programs improves the antimicrobial susceptibility of pathogens. Although such programs did not significantly influence mortality in our study, they are likely to diminish the risk for drug resistance and should be further promoted.

Several limitations of the EUROBACT study merit discussion. First, the distribution of the participating ICUs is not representative of the populations or healthcare systems in the 24 participating countries. In some countries, the number of included patients was very small. As a result, the distributions of pathogens and resistance rates reported here should be interpreted with caution. Second, each participating ICU performed laboratory tests according to their own local protocols, as opposed to sending the isolates to a central laboratory for standardised susceptibility testing. Furthermore, molecular testing of strain relatedness or confirmation of specific resistance mechanisms was not feasible in this large multinational study. Third, the study data were abstracted and entered by investigators at each ICU, with more than 100 individuals entering data in all, raising the possibility of inconsistencies. However, we attempted to minimize inconsistencies through the use of standardised definitions and of direct data entry into a web-based server. In addition, three of us (A.T., D.K., J.-F.T.) reviewed each included case for inconsistencies.

This study provides contemporary information on HA-BSI outcomes in critically ill patients within the context of increasing rates of antimicrobial resistance, particularly among gram-negative pathogens. Both MDR pathogens and failure to administer adequate antimicrobials were associated with 28-day mortality. Furthermore, the results confirm the importance of source control in severe infections. Our data underline the importance of enhanced measures to prevent HA-BSI and to control the dissemination of resistant micro-organisms. They also indicate a need for developing new antimicrobial agents for MDR gram-negative infections. The high resistance rates found in our study should encourage health authorities to preserve one of our most important resources, namely antimicrobials [33].

Acknowledgments

This research project received the Clinical Research Award with a €20,000 research grant from the European Critical Care Research Network (ECCRN). The authors thank Steven Tassel for assistance in graphics design and Benjamin Keen for his open-source form management software (http://www.formtools.org/).

Supplementary material

134_2012_2695_MOESM1_ESM.docx (32 kb)
Supplementary material 1 (DOCX 31 kb)

Copyright information

© Copyright jointly held by Springer and ESICM 2012