Pseudomonas aeruginosa Bacteremia: Risk Factors for Mortality and Influence of Delayed Receipt of Effective Antimicrobial Therapy on Clinical Outcome
Received 28 October 2002; accepted 25 April 2003;
electronically published 23 August 2003. Presented in part: 42nd Interscience Conference on Antimicrobial Agents and Chemotherapy
San Diego, CA
September 2002 (abstract K-85). Seoul National University College of Medicine
, 28 Yongon-dong Chongno-gu, Seoul, 110-744,
Republic of Korea
Laboratory Medicine, Seoul National University College of Medicine
Republic of Korea
Among the nosocomial pathogens, Pseudomonas aeruginosa is recognized as a major cause of morbidity and mortality. Data on 136 patients with P. aeruginosa bacteremia were retrospectively analyzed to evaluate risk factors for mortality. The median age of the patients was 55 years (range, 15-85 years), 78.7% of the cases were hospital-acquired, and the 30-day mortality rate was 39% (53 of 136 patients). Multivariate analysis demonstrated that risk factors for mortality included severe sepsis, pneumonia, delay in starting effective antimicrobial therapy, and an increasing APACHE II score (all P values !.05). In 123 of the 136 patients (excluding 13 patients treated with inadequate definitive antibiotics), 30-day mortality was 27.7% (13 of 47 patients) in the group of patients who received initially effective empirical antimicrobial therapy, and 43.4% (33 of 76) in the group of patients who received delayed effective antimicrobial therapy (P p .079). There was a trend toward higher mortality as the length of delay increased. Delay in starting effective antimicrobial therapy for P. aeruginosa bacteremia tended to be associated with higher mortality.
Pseudomonas aeruginosa is an important nosocomial
pathogen, especially in individuals with neutropenia
and those who are immunocompromised . During
the 1960s, when P. aeruginosa first emerged as a
common cause of gram-negative bacteremia and effective
antipseudomonal antibiotics were unavailable, the
mortality rate was 90% [4, 5]. As antipseudomonal
antibiotics were introduced, treatment outcomes in cases
of P. aeruginosa bacteremia improved. However, P.
aeruginosa continues to be a serious cause of infection,
associated with a high rate of morbidity and a mortality
rate ranging from 18% to 61% .
Pseudomonas infection is clinically indistinguishable
from other forms of gram-negative bacterial infection.
For this reason, patients with Pseudomonas infection
might receive empirical antibiotics that are inactive
against Pseudomonas, especially before antibiotic
susceptibility results become available . It has been
well documented that inappropriate antimicrobial
therapy is associated with adverse outcome , but little
information exists about whether ineffective empirical
antimicrobial therapy given during the first 4872 h,
when results of microbiological testing are unavailable,
affects the outcome adversely. We aimed to evaluate the
risk factors for mortality and to determine the influence
of delayed effective antimicrobial therapy on the clinical
outcome of patients with P. aeruginosa bacteremia.
PATIENTS AND METHODS
Patients. A review was conducted of the clinical
microbiological laboratory results and medical records of
150 individuals diagnosed from January 1998 to
December 2001 at Seoul National University Hospital (Seoul,
Republic of Korea), a 1500-bed tertiary care university hospital
and referral center. This review identified 145 patients with
clinically significant P. aeruginosa bacteremia. Only the first
bacteremic episode for each patient was included in the analysis.
Nine patients with cases of polymicrobial bacteremia were
excluded, and, therefore, 136 patients were enrolled in this study.
Microbiological testing. Identification of P. aeruginosa was
performed with a Vitek-GNI Card (bioMerieux), and antibiotic
susceptibility testing was performed using the disk diffusion
method, following the recommendations of the NCCLS [15, 16].
Study design and data collection. A retrospective cohort
study was conducted to evaluate the risk factors for mortality
and the influence of a delay in receiving effective antimicrobial
therapy on patient outcome. Thirty-day mortality was the main
outcome measurement. We reviewed the medical records of
patients retrospectively. Data collected included: age; sex;
underlying disease; source of bacteremia; severity of illness (as
calculated by the APACHE II score ); duration of hospital
stay before bacteremia; antimicrobial therapy regimen; the time
of administration of antibiotics and blood culture sample
acquisition; and antimicrobial therapy during the 30 days prior
to bacteremia. The presence of the following comorbid
conditions was documented: neutropenia; presentation with septic
shock; receipt of intensive care unit (ICU) care; use of an
immunosuppressive agent within 30 days prior to bacteremia;
history of prior corticosteroid use; being in a postoperative
state; and history of an invasive procedure within 72 h prior
to bacteremia. In addition, the presence of a central venous
catheter, urinary catheter, or mechanical ventilation was noted.
The time at which antibiotics were administered and blood
samples for culture were obtained was evaluated.
Antimicrobial therapy regimens. Patients with cancer who
were febrile and neutropenic were initially treated with an
empirical regimen that consisted of an antipseudomonal penicillin
and tobramycin. After 4872 h, the antibiotic regimen was
modified, depending on the clinical response of the patient
and/or the susceptibility of the isolated organism. If there was
a favorable response, the initial antibiotic regimen was
continued for most of the patients. However, if the response was
unfavorable, it was recommended that empirical antibiotics be
switched to broad-spectrum cephalosporin and amikacin.
The patients with P. aeruginosa bacteremia received
antimicrobial therapy for 12 weeks. The patients who had
secondary complications, such as abscess, received prolonged
therapy for 14 weeks and, if needed, surgical drainage. This study
was retrospective; therefore, the patients physicians, not the
researchers, decided the antimicrobial therapy regimens.
Definitions. The presence of P. aeruginosa bacteremia was
defined as the identification of P. aeruginosa in a blood culture
. Clinically significant P. aeruginosa bacteremia was defined
as 1 positive blood culture and clinical features compatible
with sepsis. Empirical administration of antimicrobial therapy
was classified as effective initial antimicrobial therapy or
delayed effective antimicrobial therapy according to the initial
antimicrobial therapy regimens that were administered within
24 h after blood culture samples were obtained. Effective initial
antimicrobial therapy was defined as therapy administered
within 24 h after blood culture samples were obtained and
consisting of an initial empirical regimen containing
antipseudomonal antibiotics (such as antipseudomonal penicillin,
ceftazidime, carbapenem, or fluoroquinolone) that were later
proved to be active in vitro against blood isolates of P.
aeruginosa. Delay in effective antimicrobial therapy was defined
as the administration of empirical antibiotics ineffective against
the P. aeruginosa isolate prior to the availability of the results
of antibiotic susceptibility testing, with a delay of 124 h after
blood culture samples were obtained. Regardless of the
empirical therapy, effective definitive antimicrobial therapy was
defined as treatmentafter determining the antibiotic
susceptibility of the blood isolateswith an antibiotic regimen
containing antipseudomonal antibiotics active in vitro against the
blood isolates. The length of the delay of effective antimicrobial
therapy was defined as the interval between the time the blood
culture samples were obtained and the time effective antibiotics
Nosocomially acquired infection was defined as the
following: infection that occurred 148 h after hospital admission;
infection that occurred !48 h after admission to hospital in
patients who had been hospitalized within the 2 weeks prior
to admission; or infection that occurred 148 h after hospital
admission in patients who had been transferred from an outside
hospital or nursing home . Nosocomial bloodstream
infections (and other nosocomial infections) were defined
according to the criteria proposed by the Centers for Disease
Control and Prevention . Neutropenia was defined as an
absolute neutrophil count of !500 cells/mm3. Septic shock was
defined as sepsis associated with evidence of organ
hypoperfusion and either a systolic blood pressure of !90 or 130 mm
Hg less than the baseline value or a requirement for the use
of vasopressor to maintain blood pressure .
Statistical analysis. The x2 test, Fishers exact test, or
linear-by-linear association were used to compare categorical
variables, as needed. To determine independent risk factors for
mortality, a multiple logistic regression model was used to
control for the effects of confounding variables. The results of
logistic regression analyses were reported as adjusted ORs with
95% CIs. All P values were 2-tailed, and P ! .05 was considered
to indicate statistical significance. The SPSS for Windows,
version 10.0 (SPSS), software package was used for this analysis.
Demographic and clinical characteristics. Data obtained
from a total of 136 patients with P. aeruginosa bacteremia were
analyzed. Of these patients, 66.2% were men, and the median
age for all patients was 55 (range, 1585) years. The percentage
of P. aeruginosa infections among all bloodstream infections
diagnosed during the study period was 3.8%. Demographic
data, underlying diseases, and risk factors for infection are
shown in table 1. The most common underlying disease was
solid tumor (58 [42.6%] of 136) and the most common primary
site of infection, excluding patients for whom the primary site
of infection was unknown (42 [30.9%] of 136), was the
pancreaticobiliary tract (32 [23.5%] of 136); 28.7% of the patients
had neutropenia; and 78.7% of the infections were nosocomial
Of 29 patients with community-acquired bloodstream
infection, 15 (51.7%) had a solid tumor and 6 (20.7%) had a
hematologic malignancy. Of 21 patients with a solid tumor or
hematologic malignancy, 11 (52.4%) were febrile neutropenic
Of 32 patients with pancreaticobiliary tract infection, 23
(71.9%) had a transhepatic biliary drainage device and 6
(18.8%) had undergone endoscopic retrograde
cholangiopancreatography (ERCP) within 72 h prior to the onset of
Pseudomonas bacteremia. However, 3 of the 32 patients with
pancreaticobiliary tract infection had no drainage devices and had
not undergone ERCP.
On the basis of antimicrobial susceptibility, the date of
bacteremia, and the wards of acquisition of organisms, there was
no evidence of clonal spread.
Clinical outcomes and risk factors for mortality. The
30day mortality rate among all patients was 39% (53 of 136
patients). Among nonneutropenic patients, the mortality rate
was 39.2% (38 of 97 patients), and, among neutropenic
patients, it was 38.5% (15 of 39). Among the febrile neutropenic
patients, the 30-day mortality rate was 15.8% among leukemia
patients (3 of 19 patients) and 60% among nonleukemic
patients (12 of 20).
Forty-three patients received monotherapy and 93 patients
received combination therapy. There was no significant
difference between the mortality rate among patients receiving
monotherapy and the rate among those receiving combination
therapy (34.9% vs. 40.9%; P p .51). The distribution of
antipseudomonal drugs used as definitive antimicrobial therapy in
this study population is as follows: piperacillin, 21.3% of
patients; ceftazidime, 34.6%; ciprofloxacin, 18.4%; and imipenem,
The 30-day mortality rate among the patients with
pancreaticobiliary tract infection was 22.2% (7 of 32 patients), which
was relatively low. Of these 7 patients, 24% of those in the
bTable 1. Demographic and clinical characteristics of patients
with Pseudomonas aeruginosa bacteremia.
55 (15 85)
NOTE. Values are reported as no. (%) of patients unless otherwise
a The site of infection was either documented or presumed on the basis
of clinical findings.
lactam therapy group (6 of 25) and 14.3% of those in the
quinolone therapy group (1 of 7) died. There was no significant
difference in the mortality rates of these 2 groups (P p 1.00).
By univariate analysis, variables significantly associated with
mortality included receipt of ineffective definitive antimicrobial
therapy, receipt of ineffective empirical antimicrobial therapy,
presentation with septic shock, receipt of ICU care, having
pneumonia, and having an increasing APACHE II score (all
P ! .05). No significant difference was observed in the
mortalities associated with community-acquired infections and
nosocomial infections (P p .897). Neutropenia was not associated
with a higher mortality (P p .938) (table 2).
Multivariate analysis using a logistic regression model
demonstrated that the independent risk factors for 30-day mortality
were: presentation with septic shock (OR, 45.37; 95% CI, 10.19
201.93; P ! .001), having pneumonia (OR, 11.43; 95% CI, 2.60
50.19; P p .001), receipt of ineffective definitive antimicrobial
therapy (OR, 11.68; 95% CI, 2.5154.38; P p .002), receipt of
ineffective empirical antimicrobial therapy (OR, 4.61; 95% CI,
1.1818.09; P p .028), and having an increasing APACHE II
score (1-point increments; OR, 1.31; 95% CI, 1.151.50; P !
.001) (table 3).
Influence of delayed effective antimicrobial therapy on
mortality. To evaluate the influence of delayed effective
antimicrobial therapy on mortality, data from 123 of 136 patients
(excluding 13 patients who received inappropriate definitive
antimicrobial therapy) were analyzed. The mean duration (
SD) of the delay in effective antimicrobial therapy was 3.5
1.28 days. The 30-day mortality rate in the delayed effective
treatment group was 43.4% (33 of 76 patients), whereas that
in the effective empirical treatment group was 27.7% (13 of
47). The former group tended to have a higher mortality rate
than the latter group (P p .079).
Before performing a subgroup analysis, we excluded a patient
Table 2. Thirty-day mortality rate among 136 patients with Pseudomonas
No. of deaths/
no. of patients (%)
OR (95% CI)
a This group includes 13 patients who received inadequate definitive antibiotics and 15 patients
who received ineffective empirical antibiotics and died before the pathogen was identified.
b Excluding use of corticosteroids.
Table 3. Independent risk factors for mortality for 136 patients
with Pseudomonas aeruginosa bacteremia.
Ineffective definitive antibiotic
Ineffective empirical antibiotic
Presentation with septic shock
Increasing APACHE II scorea
OR (95% CI)
NOTE. Multivariate analysis using logistic regression model.
a Per 1 point increase in score.
who had been treated with inappropriate definitive antibiotics
from the group of 39 neutropenic patients. Of the 38 patients
who received appropriate definitive antimicrobial therapy, 26
(68.4%) had received appropriate initial empirical
antimicrobial therapy against P. aeruginosa and 12 (31.6%) had received
inappropriate initial empirical antimicrobial therapy. Of 26
patients who received appropriate initial empirical antimicrobial
therapy, 7 (26.9%) died, and, of 12 patients who received
inappropriate initial empirical antimicrobial therapy, 7 (58.3%)
died (P p .081) (table 4).
Of the 97 nonneutropenic patients, 12 patients who had been
treated with inappropriate definitive antibiotics were excluded.
Of the 85 nonneutropenic patients with appropriate definitive
antimicrobial therapy, 21 (24.7%) had received appropriate
initial empirical antimicrobial therapy against P. aeruginosa and
64 (75.3%) had received inappropriate initial empirical
antimicrobial therapy. Of the 21 patients who had received
appropriate initial empirical antimicrobial therapy, 6 (28.6%) died,
and, of the 64 patients who had received inappropriate initial
empirical antimicrobial therapy, 26 (40.6%) died (P p .323)
No significant difference in mortality was found between the
delayed therapy and nondelayed therapy groups among patients
who had pancreaticobiliary tract infection (5 [21.7%] of 23
patients vs. 1 [25%] of 4 patients; P p 1.00) or among those
who had soft-tissue infection (3 [50%] of 6 vs. 3 [60%] of 5;
P p 1.00) (table 4).
The 30-day mortality rates, stratified according to the
duration of ineffective antimicrobial therapy, are presented in
figure 1. A trend toward higher mortality with increasing length
of delay was demonstrated (P p .020, using linear-by-linear
association). The percentage of patients with delayed initiation
of effective therapy whose therapy was delayed because the
organism was resistant to the antipseudomonal drug initially
administered was 10.1%.
Our study shows that the overall mortality rate of patients
infected with Pseudomonas bacteremia was 39%. No significant
differences in mortality were evident between neutropenic and
nonneutropenic patients. However, among the febrile
neutropenic patients, the 30-day mortality rate in patients without
leukemia was found to be higher than that in patients with
Previous studies have demonstrated that the presence of
severe underlying disease, pneumonia, septic shock, neutropenia,
and surgery are associated with a poor prognosis in cases of
infection with P. aeruginosa bacteremia [2, 3, 911]. In the
present study, delay in starting effective antimicrobial therapy
and inadequate definitive antibiotics were also associated with
a poor prognosis. However, subgroup analysis showed that a
delay in starting effective antimicrobial therapy was not a
significant cause of adverse outcome in nonneutropenic patients.
To evaluate the influence of delayed effective antimicrobial
therapy on mortality, we analyzed 123 of 136 patients, excluding
those that received inappropriate definitive antimicrobial
therapy. The delayed treatment group showed a trend toward higher
mortality than the effective empirical antimicrobial therapy
group, although this difference was not statistically significant
(43.4% vs. 27.7%; P p .079).
Vidal et al.  found no relationship between the adequacy
of empirical antimicrobial therapy and mortality. However, in
their institute, the length of delay in starting effective
antimicrobial therapy was 13 days. They suggested that, if the
delay was greater than that observed in their study, the
administration of ineffective empirical antibiotics might be
associated with a greater risk of death. In the present study, the
mean duration of delay in starting effective antimicrobial
therapy was 3.5 days, and this longer delay period may have
adversely affected clinical outcome. Indeed, a trend toward higher
mortality with increasing length of delay was demonstrated in
our study. In the report by Fluckiger et al. , patients for
whom an infectious diseases service was provided received
appropriate treatment more often and experienced significantly
fewer complications than those for whom one was not provided.
Our study demonstrates that, among neutropenic patients
with Pseudomonas bacteremia, the mortality rate is higher
among those treated with ineffective empirical antibiotics than
among those treated with effective antibiotics (7 [26.9%] of 26
vs. 7 [58.3%] of 12; P p .081). Vidal et al.  also suggested
that mortality among neutropenic patients increases if
empirical antibiotics are inactive in vitro against P. aeruginosa. High
mortality was found among patients with cases of pneumonia,
regardless of the adequacy of empirical antibiotics. P. aeruginosa
pneumonia is a fulminant disease and difficult to treat [22, 23],
and it may be reasonable to propose that the empirical
NOTE. Thirteen of the 136 patients in our study group were excluded from this analysis because they
received inadequate definitive antimicrobial therapy.
Deaths, n/N (%), by antibiotic regimen received
biotic regimen for nosocomial pneumonia be used for cases of
P. aeruginosa infection.
In the present study, the pancreaticobiliary tract was one of
the most common primary sites of Pseudomonas infection.
Twenty-three (71.9%) of 32 patients with pancreaticobiliary
tract infection caused by P. aeruginosa had a transhepatic biliary
drainage device and 6 (18.8%) of 32 had undergone ERCP
within the previous 72 h. These findings are consistent with
previous studies, which reported that ERCP and transhepatic
biliary drainage devices were associated with Pseudomonas
infection . However, no significant mortality differences
were found between patients with pancreaticobiliary tract
infection who were in the effective empirical antimicrobial
therapy group and those in the delayed treatment group. Our data
suggest that, if definitive antimicrobial therapy is appropriate,
delay in receiving effective antimicrobial therapy might not lead
to adverse outcome in patients with pancreaticobiliary tract
infection caused by P. aeruginosa.
In this study, we demonstrated that delay in receiving
appropriate antipseudomonal therapy had an adverse influence
on clinical outcome in patients with bloodstream infection
caused by P. aeruginosa. Therefore, we suggest that
antipseudomonal antibiotics should be considered for the empirical
antibiotic regimen, especially if P. aeruginosa infection is
prevalent. However, empirical antibiotics for bloodstream infections
should be recommended on the basis of the distribution of
pathogens in the institution where the regimen is administered.
We could not define the degree of prevalence of P. aeruginosa
that should trigger a change in the empiric antibiotic treatment.
Also, using these results to help derive recommendations for
empiric use of antipseudomonal therapy requires knowledge of
the likelihood that P. aeruginosa bacteremia is a cause of a
particular infection syndrome. Furthermore, the effect of a
delay in administering effective antimicrobial therapy might
depend on the severity of the underlying diseases and comorbid
conditions of the patients.
In conclusion, receiving inappropriate antimicrobial therapy,
experiencing septic shock, having pneumonia, and having a
severe underlying disease were the independent risk factors for
mortality. Delay in receiving effective antimicrobial therapy for
P. aeruginosa bacteremia tended to be associated with higher
mortality, and the mortality rate was higher with an increase
in the length of the delay in receiving antimicrobial therapy. It
is warranted that the delay of effective antimicrobial therapy
should be prevented in cases of bloodstream infection caused
by P. aeruginosa.