1600 CID 2002:34 (15 June) Blot et al.

M A J O R A R T I C L E
Nosocomial Bacteremia Caused by Antibiotic-
Resistant Gram-Negative Bacteria in Critically
Ill Patients: Clinical Outcome and Length
of Hospitalization
Stijn Blot,
1
Koenraad Vandewoude,
1
Dirk De Bacquer,
2
and Francis Colardyn
1
1
Department of Intensive Care, Ghent University Hospital, and
2
Department of Public Health, Ghent University, Ghent, Belgium
Population characteristics and outcomes were retrospectively compared for critically ill patients with noso-
comial bacteremia caused by antibiotic-susceptible (AB-S; ) or antibiotic-resistant (AB-R; ) n p208 n p120
gram-negative bacteria. No signicant differences in severity of illness and comorbidity factors were seen
between groups. Patients with bacteremia caused by AB-R strains had a longer hospitalization before the onset
of the bacteremia. The in-hospital mortality for patients with bacteremia caused by AB-S strains was 41.8%;
for patients infected with AB-R strains, it was 45.0% ( ). A multivariate survival analysis demonstrated P p.576
that older age ( ), a high-risk source of bacteremia (abdominal and lower respiratory tract; P p.009 P p
), and a high acute physiology and chronic health evaluation IIrelated expected mortality ( ) were .031 P p.032
independently associated with in-hospital mortality ( ). Antibiotic resistance in nosocomial bacteremia P ! .05
caused by gram-negative bacteria does not adversely affect the outcome for critically ill patients.
The widespread use of broad-spectrum antibiotics is
the principal factor in the emergence of antibiotic re-
sistance. Consequently, in intensive care units (ICUs),
where the use of antibiotics is considerably greater than
in general wards [1], dealing with antibiotic-resistant
microorganisms is an almost daily challenge. This must
be considered a major problem, because the develop-
ment of newer and more-potent antibiotics cannot keep
up with the increase in antibiotic resistance.
In ICUs, gram-negative bacteria are responsible for a
Received 3 December 2001; revised 1 February 2002; electronically published
23 May 2002.
Presented in part: 14th Annual Congress of the European Society of Intensive
Care Medicine, Geneva, 30 September3 October 2001.
Financial support: Fund for Scientic Research, Flanders, Belgium (special
doctoral grant 1.9.205.02N00 to S.B.).
Reprints or correspondence: Dr. Stijn Blot, Dept. of Intensive Care, Ghent
University Hospital, De Pintelaan 185, B-9000 Ghent, Belgium (stijn.blot@rug
.ac.be).
Clinical Infectious Diseases 2002; 34:16006
2002 by the Infectious Diseases Society of America. All rights reserved.
1058-4838/2002/3412-0009$03.00
considerable percentage of all bloodstreaminfections [2].
In ICUs in the United States and Europe, patterns of
reduced susceptibility to antibiotics were found among
gram-negative bacteria [3, 4]. Despite the highprevalence
of antibiotic resistance among gram-negative bacteria
causing bacteremia, the clinical consequences of resis-
tance remain unclear. The main objective of our study
was to evaluate the relationship between antibiotic re-
sistance in gram-negative bacteria causing bacteremia
and the clinical outcomes for critically ill patients. Sec-
ondary objectives were to compare the length of the ICU
stay and hospitalization for patients with bacteremia
caused by antibiotic-susceptible (AB-S) or antibiotic-
resistant (AB-R) gram-negative strains.
METHODS
Setting. This study was performed in the ICU of the
1060-bed Ghent University Hospital in Ghent, Belgium.
The ICU has 54 beds and includes a medical and sur-
gical ICU, an ICU for cardiac surgery, and an ICU for

b
y

g
u
e
s
t

o
n

J
u
l
y

1
5
,

2
0
1
4
h
t
t
p
:
/
/
c
i
d
.
o
x
f
o
r
d
j
o
u
r
n
a
l
s
.
o
r
g
/
D
o
w
n
l
o
a
d
e
d

f
r
o
m

Bacteremia Caused by Resistant Microorganisms CID 2002:34 (15 June) 1601
severely burned patients. No signicant changes in mean age
of patients, length of the ICU stay, or acute physiology and
chronic health evaluation (APACHE) II score [5] were observed
during the study period.
Study design and data collection. We conducted a retro-
spective, observational cohort study that included ICU patients
with nosocomial, microbiologically documented bacteremia
caused by gram-negative bacteria. We compared data from pa-
tients with bacteremia caused by AB-S gram-negative bacteria
with data from patients with bacteremia caused by AB-R gram-
negative bacteria. In-hospital mortality (mortality rate for 3
evaluation points) was the principal outcome variable evalu-
ated. We also assessed secondary outcomes, including length
of stay in the ICU and in the hospital and prevalence of acute
organ failure.
The study included critically ill adult patients who were ad-
mitted to the ICU during a 9-year period (January 1992
December 2000). All microbiologically documented nosoco-
mial bloodstream infections are prospectively screened by the
center for infection control. This hospital-wide case-based sur-
veillance program was used to perform a retrospective search
for ICU patients with bacteremia caused by gram-negative bac-
teria. Every patient whose ICU stay was complicated by this
bloodstream infection was assessed in our analysis. For ICU
patients who developed 11 case of bacteremia caused by gram-
negative bacteria, only the rst episode was considered. Patients
with hemocultures that yielded 11 type of gram-negative bac-
teria of which at least 1 strain was AB-R were included in the
AB-R group.
Denitions. Bacteremia was considered to be nosocomial
when it was diagnosed at least 48 h after hospital admission.
Gram-negative bacteremia was dened as the presence of
gram-negative bacteria in the blood, documented by at least 1
positive hemoculture. Hemocultures were routinely performed
when the patients temperature was 138.4C or when infection
was suspected on clinical grounds; blood samples were pro-
cessed following the BacT/Alert (Organon Teknika) procedure.
A 10-mL blood culture inoculum was standard. Antibiotic
resistance was dened as in vitro resistance to ceftazidime. In
our hospital, ceftazidime resistance is considered to be an in-
dicator of epidemic extended-spectrumb-lactamaseproducing
strains or hyperproducers of b-lactamases, and, therefore, it is
a sign of infection with organisms that are resistant to multiple
drugs [6, 7]. Because susceptibility patterns for Pseudomonas
aeruginosa vary, such isolates were considered to be AB-R when
resistance to one of the following antipseudomonal antibiotics
was seen: piperacillin, ciprooxacin, ceftazidime, and imipen-
em [8].
Antibiotic resistance was determined according to methods
for disk-diffusion testing recommended by the National Com-
mittee for Clinical Laboratory Standards [9]. For the sake of
convenience, cases of bacteremia were designated AB-S bac-
teremia or AB-R bacteremia, depending on the antibiotic-
resistance status of the isolated organisms. During the study
period, no changes in microbiologic laboratory techniques were
seen. The source of the bacteremia was determined by intensive
care physicians and microbiologists, on the basis of the isolation
of gram-negative bacteria from the presumed portal of entry
and clinical evaluation. For the purpose of analysis, sources of
bacteremia were divided into 3 categories: low risk (associated
mortality, 30%), which were sinus, urinary tract, intravenous
catheter, and soft-tissue sources; intermediate risk (associated
mortality, 31%50%), which were primary sources of bacte-
remia; and high risk (associated mortality, 150%), which were
lower respiratory tract and abdominal sources. Patients with
11 possible source of bacteremia (e.g., 1 low-risk and 1 high-
risk source) were considered to have a high-risk source.
Antibiotic therapy was considered to be appropriate if the
drugs used had in vitro activity against the isolated strain. We
considered antibiotic therapy to be inadequate if the drugs
used did not have in vitro activity against the isolated strain
or if the patient did not receive antibiotic treatment. The delay
in the initiation of appropriate antibiotic treatment was cal-
culated from the day of onset of bacteremia. Acute respiratory
failure was dened as ventilator dependence; acute renal fail-
ure, as the need for renal replacement therapy; and hemo-
dynamic instability, as the need for vasopressive or inotropic
support during the ICU stay. For the comparison of outcomes,
survival status was evaluated at 14 and 28 days after the onset
of bacteremia and at the end of the hospital stay (all 3 evaluation
points are included in in-hospital mortality).
Statistical analysis. Continuous variables are given as
or as median (lowerupper quartile), depending meanSD
on the distribution. Comparative analyses were done with the
Mann-Whitney U test or the x
2
test, as appropriate. Survival
curves were prepared by means of the Kaplan-Meier method,
and univariate survival distributions were compared by the log-
rank test. To assess the relationship between in-hospital mor-
tality and a set of independent variables, a multivariate survival
analysis was used (Cox proportional hazard model); hazard
ratios and 95% CIs are reported. In this multivariate analysis,
continuous variables were handled continuously. Variables en-
tered in the Cox regression model were required to have a
plausible relationship with mortality, to avoid spurious asso-
ciations. Statistical analyses were performed using Statistica,
version 4.5 (StatSoft), and SPSS, version 9.0. All tests were 2
tailed; was considered to be statistically signicant. P ! .05
RESULTS
During the study period, 29,727 patients were admitted to the
ICU. Among these, 328 patients were identied as having gram-

b
y

g
u
e
s
t

o
n

J
u
l
y

1
5
,

2
0
1
4
h
t
t
p
:
/
/
c
i
d
.
o
x
f
o
r
d
j
o
u
r
n
a
l
s
.
o
r
g
/
D
o
w
n
l
o
a
d
e
d

f
r
o
m

1602 CID 2002:34 (15 June) Blot et al.
Table 1. Characteristics of hospitalized, critically ill patients who had nosocomial bacteremia caused by gram-negative bacteria.
Characteristic
Outcome of hospital stay Type of bacteremia
Death
(n p 141)
Survival
(n p 187) P
AB-S
(n p 208)
AB-R
(n p 120) P
Age, mean years SD 59 14.3 49 18.1 !.001 54 17.4 52 16.9 .303
APACHE II score, mean SD 26 9.4 22 8.0 !.001 23 9.0 23 8.6 .814
APACHE IIrelated expected mortality, mean %
SD 50 29.0 36 25.1 !.001 41 28.2 44 27.3 .305
Acute renal failure, % of patients 38.3 16.0 !.001 25.0 26.7 .431
Hemodynamic instability, % of patients 86.5 75.9 .017 77.9 85.0 .117
Acute respiratory failure, % of patients 95.7 92.0 .021 92.4 95.8 .227
Ventilator dependence, median days (lowerupper
quartile) 17 (628) 21 (834) .165 16 (527) 23 (1238) !.001
Length of ICU stay, median days
(lowerupper quartile)
Before onset of bacteremia 10 (318) 13 (624) .009 8 (3.517) 18 (933) !.001
After onset of bacteremia 8 (317) 12 (525) .002 10 (319) 11 (520) .321
Total 18 (934) 28 (1446) !.001 21 (940) 28.5 (1648) !.001
Length of hospitalization, median days
(lowerupper quartile)
Before onset of bacteremia 13 (527) 16 (727) .428 11 (522) 23 (1141) !.001
After onset of bacteremia 10 (321) 59 (28117) !.001 27 (963.5) 35 (1077) .333
Total 29 (1554) 76 (44144) !.001 47 (2285) 60 (30130) .007
Mortality at 14 days, % of patients 61.0 0 26.0 26.7 .550
Mortality at 28 days, % of patients 81.0 0 34.1 35.8 .756
In-hospital mortality, % of patients 100 0 41.8 45.0 .576
NOTE. AB-R, antibiotic resistant; AB-S, antibiotic susceptible; APACHE, acute physiology and chronic health evaluation; ICU, intensive care unit.
negative bacteremia and were included in the study cohort (a
prevalence of 11.0 cases of gram-negative bacteremia per 1000
ICU admissions). The mean age of the patients was 54
years. The mean APACHE II score was , and the 17.2 23 8.9
mean APACHE IIrelated expected mortality was 42%
. Fifty-four percent of the patients were admitted to the 27.8%
ICU after a surgical procedure; 75% of these patients had non-
elective surgery. Twenty percent of the patients were admitted
after experiencing trauma.
Among the 328 cases of bacteremia included in the analysis,
369 gram-negative isolates were identied. The most frequently
detected gram-negative microorganisms were Escherichia coli
( ), Enterobacter species ( ), P. aeruginosa ( n p71 n p68 n p
), Klebsiella species ( ), Acinetobacter species ( ), 62 n p52 n p51
and Serratia species ( ). In 36.6% (120 of 328) of the n p27
cases of bacteremia, the strain involved was AB-R. During the
study period, the yearly rate of AB-R bacteremia remained
stable ( ) (data not shown). P p.495
Outcome of hospital stay. Mortality rates at 14 days, at
28 days, and at the end of the hospital stay were, respectively,
26%, 35%, and 43%. The characteristics of the patients who
died in the hospital and the characteristics of those who sur-
vived are compared in table 1. Patients who died in the hospital
generally were older and had higher APACHE II scores and
APACHE IIrelated expected mortality, and the prevalence of
organ failure was also higher in this group. These patients had
a shorter ICU stay and a shorter hospitalization. Of all bac-
teremia-associated factors, only infection with P. aeruginosa
and high-risk sources of bacteremia were more prevalent among
nonsurvivors (table 2).
Patients with AB-S bacteremia versus those with AB-R bac-
teremia. The population characteristics of patients with AB-
S gram-negative bacteremia and those with AB-R gram-neg-
ative bacteremia are listed in tables 1 and 2. Polymicrobial
bloodstream infections were more likely to occur in the AB-
R group. Furthermore, no important differences in severity
of illness were found, but patients with AB-R bacteremia had
a longer stay in the ICU, as well as in the hospital. This seems
to be the consequence of a longer hospitalization before the
onset of the bacteremia; length of stay (both in ICU and
hospital) after the onset of the bacteremia was not different.
Antibiotic resistance was not associated with higher mor-
tality rates (table 1). Figure 1 shows the survival curves for
both groups from the onset of bacteremia to the end of hos-

b
y

g
u
e
s
t

o
n

J
u
l
y

1
5
,

2
0
1
4
h
t
t
p
:
/
/
c
i
d
.
o
x
f
o
r
d
j
o
u
r
n
a
l
s
.
o
r
g
/
D
o
w
n
l
o
a
d
e
d

f
r
o
m

Bacteremia Caused by Resistant Microorganisms CID 2002:34 (15 June) 1603
Table 2. Factors associated with bacteremia in a study of hospitalized, critically ill patients with nosocomial
bacteremia caused by gram-negative bacteria.
Factor
Outcome of hospital stay Type of bacteremia
Death
(n p 141)
Survival
(n p 187) P
AB-S
(n p 208)
AB-R
(n p 120) P
AB-R bacteremia 38.3 35.3 .576
Isolated microorganism
Escherichia coli 21.3 21.9 .888 30.8 5.8 !.001
Enterobacter species 17.0 23.5 .150 14.4 31.7 !.001
Pseudomonas aeruginosa 27.0 12.8 .001 18.8 19.2 .926
Klebsiella species 12.7 18.2 .184 15.9 15.8 .994
Acinetobacter species 12.1 17.6 .163 4.8 33.3 !.001
Serratia species 9.2 7.5 .572 9.6 5.8 .230
Polymicrobial bloodstream infection 33.3 26.2 .160 24.5 37.5 .013
Source of the bacteremia
Low risk 17.0 41.2 !.001 33.7 27.5 .247
Intermediate risk 22.7 25.1 .609 20.7 30.0 .060
High risk 60.3 32.6 !.001 45.7 42.5 .578
More than 1 possible source 9.2 10.0 .901 7.7 12.5 .152
Appropriate antibiotic therapy 89.3 94.6 .108 93.1 91.1 .547
Delay in antibiotic therapy, mean days SD 0.5 0.9 0.7 1.6 .296 0.5 0.8 0.8 2.0 .556
NOTE. Data are percentage of patients, unless otherwise indicated. AB-R, antibiotic resistant; AB-S, antibiotic susceptible.
Figure 1. Survival curves for patients in the intensive care unit who had bacteremia caused by antibiotic-susceptible (dashed line) or antibiotic-
resistant (solid line) gram-negative bacteria ( ; log-rank test). P p.319
pitalization. The mortality rates for the 2 groups were very
similar ( ). At the end of the hospitalization period, P p.319
mortality rates for patients with AB-S bacteremia and those
with AB-R bacteremia were, respectively, 41.8% and 45.0%
( ). To adjust for differences in clinical virulence, out- P p.576
come comparisons for the 2 groups of patients were per-
formed, according to the different types of gram-negative mi-
croorganisms involved. When patients were stratied by
most-prevalent bacteria, no statistically signicant difference
in mortality was seen between the 2 groups ( ; table 3). P 1 .05
Multivariate survival analysis. A multivariate survival
analysis demonstrated that older age, a high-risk source of
bacteremia, and a high APACHE IIrelated expected mortality
were independently associated with in-hospital mortality (ta-

b
y

g
u
e
s
t

o
n

J
u
l
y

1
5
,

2
0
1
4
h
t
t
p
:
/
/
c
i
d
.
o
x
f
o
r
d
j
o
u
r
n
a
l
s
.
o
r
g
/
D
o
w
n
l
o
a
d
e
d

f
r
o
m

1604 CID 2002:34 (15 June) Blot et al.
Table 3. Mortality rates among hospitalized, critically ill pa-
tients with bacteremia caused by gram-negative bacteria, strat-
ied by most frequently detected microorganism and distributed
by antibiotic-resistance status of the isolate.
Microorganism,
resistance status
No. of
isolates
Mortality, no. (%) of patients
At 14 days At 28 days Total
Escherichia coli
AB-S 64 14 (21.9) 20 (31.3) 25 (39.1)
AB-R 7 4 (57.1) 4 (57.1) 5 (71.4)
Enterobacter species
AB-S 31 7 (22.6) 9 (29.0) 11 (35.5)
AB-R 37 7 (18.9) 9 (24.3) 13 (35.1)
Pseudomonas
aeruginosa
AB-S 39 12 (30.8) 17 (43.6) 25 (64.1)
AB-R 23 9 (39.1) 11 (47.8) 13 (56.5)
Klebsiella species
AB-S 34 6 (17.6) 9 (26.5) 9 (26.5)
AB-R 18 5 (27.8) 8 (44.4) 9 (50.0)
Acinetobacter species
AB-S 10 0 1 (10.0) 1 (10.0)
AB-R 41 8 (19.5) 14 (34.1) 17 (41.5)
Serratia species
AB-S 21 7 (33.3) 10 (47.6) 10 (47.6)
AB-R 6 3 (50.0) 3 (50.0) 3 (50.0)
NOTE. Differences in mortality for patients with antibiotic-susceptible(AB-
S) and patients with antibiotic-resistant (AB-R) bacteremia did not reach sta-
tistical signicance at any time point ( ). P ! .05
Table 4. Factors associated with in-hospital mortality among
hospitalized, critically ill patients with bacteremia caused by
gram-negative bacteria.
Variable Hazard ratio (95% CI) P
Age 1.01 (1.001.02) .009
High-risk source of bacteremia 1.33 (1.031.72) .031
APACHE IIrelated expected
mortality 1.71 (1.052.79) .032
Acute renal failure 1.32 (0.981.76) .065
Pseudomonas aeruginosa
bacteremia 1.33 (0.961.85) .085
NOTE. Associations were assessed using multivariate survival analysis
(Cox proportional-hazard model). APACHE, acute physiology and chronic health
evaluation.
ble 4). Acute renal failure and bacteremia caused by P. aeru-
ginosa had an association with in-hospital mortality that was
of borderline signicance ( ). P ! .1
DISCUSSION
It is presumed that infections caused by AB-R bacteria result
in higher mortality, longer hospitalizations, and greater costs
than do infections caused by AB-S bacteria, although few data
support this intuitive concept [10]. The assumption that in-
fections caused by AB-R bacteria are associated with a higher
mortality rate may be based on the possibility that appropriate
antibiotic therapy will be initiated later for such infections than
for infections caused by AB-S bacteria. Although some studies
have described very high mortality rates in association with
infections caused by AB-R P. aeruginosa and Acinetobacter bau-
mannii [8, 11], the causal link between antibiotic resistance and
fatal bacteremia remains unclear.
Comparisons of the outcomes for patients with AB-R bac-
teremia and for patients with AB-S bacteremia can be difcult
to perform: important differences in demographic character-
istics between the populations are repeatedly observed in stud-
ies of bacteremia caused by AB-R gram-positive bacteria
[1214]. In the present study, no important differences in
severity of illness (as measured by the APACHE II score) or
prevalence of acute organ failure hampered the outcome com-
parison. Polymicrobial bloodstream infections were seen more
frequently among patients with AB-R bacteremia. Although
this is considered to be an ominous sign [15, 16], it has been
demonstrated elsewhere that the mortality associated with
polymicrobial bacteremia in ICU patients is not necessarily
higher than that associated with monomicrobial bacteremia
[17]. In our study, polymicrobial bacteremia was not asso-
ciated with higher in-hospital mortality on either univariate
or multivariate analysis.
Our data revealed that the mortality associated with AB-R
gram-negative bacteremia is not higher than that associated
with AB-S bacteremia. Although mortality is high in both
groups of patients, the observed mortality rates are in line with
the APACHE IIrelated expected mortality. In a retrospective
study, Menashe et al. [18] also found no signicant increase in
mortality among patients with bacteremia caused by extended-
spectrum b-lactamaseproducing Enterobacteriaceae (50%, vs.
38% among patients with bacteremia caused by nonextended-
spectrum b-lactamaseproducing isolates; ), but their P p.367
study cohort was small ( ). Harbarth et al. [16], in a n p55
large retrospective study, also failed to demonstrate that anti-
biotic resistance had a major clinical effect on outcome for
patients with gram-negative bacteremia. However, the hospital-
wide setting and large proportion of cases of community-
acquired bacteremia (61%) in that study hampers comparison
with the results we obtained.
Previous studies have identied important associations be-
tween antibiotic resistance and negative outcome for patients
with gram-negative bacteremia [1921]. However, causality re-
mains unclear, and differences in study populations impede
comparison with the present study.
It is a possible weakness in our study that the outcome
comparison between the 2 groups of patients might be con-

b
y

g
u
e
s
t

o
n

J
u
l
y

1
5
,

2
0
1
4
h
t
t
p
:
/
/
c
i
d
.
o
x
f
o
r
d
j
o
u
r
n
a
l
s
.
o
r
g
/
D
o
w
n
l
o
a
d
e
d

f
r
o
m

Bacteremia Caused by Resistant Microorganisms CID 2002:34 (15 June) 1605
founded by differences in the clinical virulence of the types of
gram-negative bacteria involved [16]. For example, 80% of Aci-
netobacter species, which are considered to be low-virulence
pathogens, were AB-R, whereas 37% of Pseudomonas isolates,
which are known to be extremely virulent, were AB-R. Re-
gardless, in the multivariate survival analysis, no individual
microorganism was recognized as being an independent pre-
dictor of mortality. Therefore, we assume that the confounding
effect of differences in clinical virulence of the different types
of bacteria seen in our study is of minor importance. However,
outcome evaluation after stratication for different bacteria can
be interesting.
In a prospective cohort study involving patients with bac-
teremia caused exclusively by Enterobacter species, Chow et al.
[21] describe greater morbidity and mortality among patients
with bacteremia caused by AB-R Enterobacter species. Our study
included 68 ICU patients with bacteremia caused by Entero-
bacter, 37 of whom were infected with AB-R strains. Although
these numbers are small and, therefore, are difcult to interpret,
the mortality rates in these subgroups were nearly equal (35.5%
and 35.1%, respectively), raising doubts about whether a worse
prognosis was associated with bacteremia caused by AB-R En-
terobacter species in our population. Also, when data were strat-
ied by other frequently detected gram-negative microorgan-
isms, no statistically signicant differences in the mortality rates
were found for the AB-S group and the AB-R group (table 3).
Nonetheless, the differences in the mortality rates for patients
with bacteremia caused by AB-S or AB-R strains of Klebsiella
species, E. coli, and, in particular, Acinetobacter species, which
reached borderline signicance ( ), are noteworthy. These P ! .1
results, however, must be interpreted cautiously because of the
small numbers of patients included.
It can be presumed that bacteremia caused by AB-R bacteria
may have a worse prognosis because of the delay in initiation
of appropriate antibiotic therapy [22, 23]. Especially whenhigh-
virulence microorganisms such as P. aeruginosa are involved,
early initiation of appropriate therapy is crucial to the survival
of the patient [24, 25]. In our study, patients with AB-R bac-
teremia were treated appropriately at a rate similar to that for
patients with AB-S bacteremia. The delay in the initiation of
appropriate treatment was longer in the AB-R group, but the
difference we found was of no signicance, either statistically
or clinically (table 2).
The high rate of administration of appropriate therapy that
was achieved in our ICU population can be explained by 3
factors: (1) consideration of previous colonization, as assessed
by regular site-specic surveillance cultures (3 times weekly),
(2) initial administration of broad-spectrum antibiotic therapy,
and (3) close interaction between the physician, clinical mi-
crobiologists, and clinical infectious diseases consultants. This
strategy appeared to contribute to a delay in administration of
appropriate antibiotic treatment that was short (!1 day) overall
in our study. Surveillance sampling was found to have a high
specic and negative predictive value, because colonizationpre-
ceded infection in almost all patients who had P. aeruginosa
infection in the ICU [26, 27]. On the other hand, performing
surveillance cultures routinely is expensive. Furthermore,
whether this policy has greater clinical benets than does ini-
tiation of blind therapy with broad-spectrum antibiotics is still
controversial. It is well understood that blind therapy is re-
stricted to a smaller spectrum, once microbiologic identication
is completed, yet it might be assumed that this promotes the
emergence of AB-R microorganisms, because there is more
exposure to broad-spectrum antibiotics. Our strategy is to re-
serve these agents for patients with sepsis whose condition is
deteriorating and for whom no isolate has been obtained from
the presumed septic focus and for patients who have micro-
biologically proved superinfection.
Patients with AB-R bacteremia had a signicantly longer stay
in the ICU, as well as in the hospital. However, when length
of ICU stay and length of hospitalization (from the day of the
onset of bacteremia to discharge or death) were compared, no
difference was found. We conclude from this that the excess
length of stay, both in the ICU and the hospital, is a result of
a longer ICU and hospital stay before the onset of the bacte-
remia. Consequently, it appears that cases of AB-R bacteremia
are not responsible for the increased need for hospital resources
among these patients but that risk factors for the acquisition
of AB-R infections are the major triggers of this problem.
In conclusion, we found that AB-R gram-negative bacteremia
is not associated with higher mortality than is AB-S gram-
negative bacteremia in critically ill patients. AB-R bacteremia
is associated with a longer stay in the ICU, as well as in the
hospital. This, however, is the consequence of a longer stay
before the onset of the bacteremia, whereas length of stay in
the ICU and length of stay in the hospital after onset of the
bacteremia in the 2 group of patients were not different.
References
1. Roder BL, Nielson SL, Magnussen P, et al. Antibiotic usage in an
intensive care unit in a Danish university hospital. J Antimicrob
Chemother 1993; 32:63342.
2. Vincent J-L, Bihari DJ, Suter PM, et al. The prevalence of nosocomial
infection in intensive care units in Europe: results of the European
Prevalence of Infection in Intensive Care (EPIC) study. JAMA 1995;
274:63944.
3. Burwen DR, Banerjee SN, Gaynes RP. Ceftazidime resistance among
selected nosocomial gram-negative bacilli in the United States. National
Nosocomial Infections Surveillance System. J Infect Dis 1994; 170:
16225.
4. Hanberger H, Garcia-Rodriguez J, Gobernado M, et al. Antibiotic sus-
ceptibility among aerobic gram-negative bacilli in intensive care units
in 5 European countries. JAMA 1999; 281:6771.
5. Knaus WA, Draper EA, Wagner DP, Zimmerman JE. APACHE II: a
severity of disease classication system. Crit Care Med 1985; 13:81829.

b
y

g
u
e
s
t

o
n

J
u
l
y

1
5
,

2
0
1
4
h
t
t
p
:
/
/
c
i
d
.
o
x
f
o
r
d
j
o
u
r
n
a
l
s
.
o
r
g
/
D
o
w
n
l
o
a
d
e
d

f
r
o
m

1606 CID 2002:34 (15 June) Blot et al.
6. Sirot D. Extended-spectrum plasmid-mediated beta-lactamases. J An-
timicrob Chemother 1995; 36(Suppl A):1934.
7. Colardyn F, Decruyenaere J, Verschraegen G, et al. Failure of selective
digestive decontamination (SDD) to control an epidemic of coloni-
zation by extended spectrum b-lactamaseproducing Klebsiella pneu-
moniae (CAZ) [abstract IM-9]. In: Program and abstracts of the 37th
Interscience Conference on Antimicrobial Agents and Chemotherapy
(Toronto). Washington, DC: American Society for Microbiology, 1997:
365.
8. Carmeli Y, Troillet N, Karchmer AW, Samore MH. Health and eco-
nomic outcomes of antibiotic resistance in Pseudomonas aeruginosa.
Arch Intern Med 1999; 159:112732.
9. National Committee for Clinical Laboratory Standards (NCCLS). Per-
formance standards for antimicrobial disk susceptibility tests. NCCLS
document M2A5. Villanova, PA: NCCLS, 1993.
10. Mehta RM, Niederman MS. Antibiotic resistance in the intensive care
unit. In: Vincent J-L, ed. Yearbook of intensive care 2001. Berlin:
Springer-Verlag, 2001:15161.
11. Rello J, Rue M, Jubert P, et al. Survival in patients with nosocomial
pneumonia: impact of the severity of illness and the etiologic agent.
Crit Care Med 1997; 25:18627.
12. Blot S, Vandewoude K, Colardyn F. Staphylococcus aureus infections
[letter]. N Engl J Med 1998; 339:20256.
13. Soriano A, Martinez J, Mensa J, et al. Pathogenic signicance of meth-
icillin resistance for patients with Staphylococcus aureus bacteremia.
Clin Infect Dis 2000; 30:36873.
14. Bhavnani SM, Drake JA, Forrest A, et al. A nationwide multicenter,
case-control study comparing risk factors, treatment, and outcome for
vancomycin-resistant and -susceptible enterococcal bacteremia. Diagn
Microbiol Infect Dis 2000; 36:14558.
15. Reuben AG, Musher DM, Hamill RJ, Brouke I. Polymicrobial bacte-
remia: clinical and microbiologic patterns. Rev Infect Dis 1989; 11:
16183.
16. Harbarth S, Rohner P, Auckenthaler R, Safran E, Sudre P, Pittet D.
Impact and pattern of gram-negative bacteraemia during 6 y at a large
university hospital. Scand J Infect Dis 1999; 31:1638.
17. Rello J, Quintana E, Mirelis B, Gurgui M, Net A, Prats G. Polymicrobial
bacteremia in critically ill patients. Intensive Care Med 1993; 19:225.
18. Menashe G, Borer A, Peled N, et al. Clinical signicance and impact
on mortality of ESBL-producing gram-negative isolates in nosocomial
bacteremia [abstract 1425]. In: Program and abstracts of the 39th In-
terscience Conference on Antimicrobial Agents and Chemotherapy
(San Francisco). Washington, DC: American Society for Microbiology,
1999:616.
19. Krcmery V Jr, Spanik S, Krupova I, et al. Bacteremia due to multi-
resistant gram-negative bacilli in neutropenic cancer patients: a case
controlled study. J Chemother 1998; 10:3205.
20. Mehlhaff DL, Briceland L, Tobin E, et al. Comparison of survival in
patients having bacteremia with and without extended-spectrum beta-
lactamase producing Enterobacteriaceae (ESBLPE) [abstract J98]. In:
Program and abstracts of the 36th Interscience Conference on Anti-
microbial Agents and Chemotherapy (New Orleans). Washington, DC:
American Society for Microbiology, 1996:237.
21. Chow JW, Fine MJ, Shlaes DM, et al. Enterobacter bacteremia: clinical
features and emergence of antibiotic resistance during therapy. Ann
Intern Med 1991; 115:58590.
22. Kollef MH, Sherman G, Ward S, Fraser VJ. Inadequate antimicrobial
treatment of infections: a risk factor for hospital mortality among
critically ill patients. Chest 1999; 115:46274.
23. Ibrahim EH, Sherman G, Ward S, Fraser VJ, Kollef MH. The inuence
of inadequate antimicrobial treatment of blood stream infections on
patient outcomes in the ICU setting. Chest 2000; 118:14655.
24. Bodey G, Jadeja L, Elting L. Pseudomonas bacteremia: retrospective
analysis of 410 episodes. Arch Intern Med 1985; 145:16219.
25. Vidal F, Mensa J, Almela M, et al. Epidemiology and outcome of
Pseudomonas aeruginosa bacteremia, with special emphasis on the in-
uence of antibiotic treatment. Arch Intern Med 1996; 156:21216.
26. Bertrand X, Thouverez M, Talon D, et al. Endemicity, molecular di-
versity and colonisation routes of Pseudomonas aeruginosa in intensive
care units. Intensive Care Med 2001; 27:12638.
27. Langer M, Carretto E, Haeusler EA. Infection control in ICU: back
(forward) to surveillance samples. Intensive Care Med 2001; 27:15613.

b
y

g
u
e
s
t

o
n

J
u
l
y

1
5
,

2
0
1
4
h
t
t
p
:
/
/
c
i
d
.
o
x
f
o
r
d
j
o
u
r
n
a
l
s
.
o
r
g
/
D
o
w
n
l
o
a
d
e
d

f
r
o
m