Inappropriate antibiotic therapy in Gram-negative sepsis increases

hospital length of stay*

Andrew F. Shorr, MD, MPH; Scott T. Micek, PharmD; Emily C. Welch, PharmD; Joshua A. Doherty, BA;
Richard M. Reichley, RPh; Marin H. Kollef, MD

Objectives: To describe the impact of initially inappropriate
antibiotic therapy on hospital length of stay in Gram-negative
severe sepsis and septic shock.
Design: Retrospective cohort.
Setting: Academic urban hospital.
Patients: Patients with Gram-negative bacteremia (primary or
secondary, nosocomial or non-nosocomial) and severe sepsis or
septic shock.
Interventions: None.
Measurements and Main Results: We defined initially inappro-
priate antibiotic therapy as occurring when the patient either was
not administered an antibiotic within 24 hrs of sepsis onset or
was treated with an antibiotic to which the culprit pathogen was
resistant in vitro. The cohort included 760 subjects (mean age
59.3 ⴞ 16.3 yrs, mean Acute Physiology and Chronic Health
Evaluation II score 23.7 ⴞ 6.7). More than half of infections were
nosocomial (55.1%), and Escherichia coli represented the most
common pathogen (n ⴝ 225). Pseudomonas species were isolated
in 17.4% of patients. Nearly one-third of patients (31.3%) received
initially inappropriate antibiotic therapy. Patients administered ini-
tially inappropriate antibiotic therapy were more likely to have a
nosocomial infection, to have underlying cancer or diabetes or
both, to require chronic hemodialysis, and to undergo mechanical
ventilation. Those administered initially inappropriate antibiotic
therapy also faced higher inhospital mortality. The unadjusted
median length of stay after sepsis onset in those administered
initially inappropriate antibiotic therapy was 11 days compared to
9 days in those treated appropriately (p ⴝ .028 by log-rank test).
In a Cox model controlling for the multiple confounders noted,
initially inappropriate antibiotic therapy independently correlated
with continued hospitalization (adjusted hazard ratio 1.19, 95%
confidence interval 1.01–1.40, p ⴝ .044). Adjusting for these
covariates indicated that initially inappropriate antibiotic therapy
independently increased the median attributable length of stay by
2 days.
Conclusions: Initially inappropriate antibiotic therapy occurs in
one-third of persons with severe sepsis and septic shock attrib-
utable to Gram-negative organisms. Beyond its impact on mor-
tality, initially inappropriate antibiotic therapy is significantly
associated with length of stay in this population. Efforts to de-
crease rates of initially inappropriate antibiotic therapy may serve
to improve hospital resource use by leading to shorter overall
hospital stays. (Crit Care Med 2011; 39:46 –51)
KEY WORDS: antibiotics; length of stay; outcomes; sepsis

S evere sepsis remains associated
with substantial morbidity,
and crude mortality rates ap-
proach 40% (1, 2). Survivors of
sepsis often also have prolonged hospital-
izations (1, 2). Reflecting this, direct hos-
pital costs may exceed $40,000 per epi-
sode (1). Indirect costs attributable to
lost productivity are excessive as well.
*See also p. 199.
From the Washington Hospital Center (AFS),
Washington, DC; Pharmacy Department (STM, ECW),
Barnes-Jewish Hospital, St. Louis, MO; Pulmonary
and Critical Care Division (MHK), Washington Uni-
versity School of Medicine, St. Louis, MO; and Hos-
pital Informatics Group (JAD, RMR), BJC Healthcare,
St. Louis, MO.
The authors have not disclosed any potential con-
flicts of interest.
For information regarding this article, E-mail:
afshorr@dnamail.com
Copyright © 2010 by the Society of Critical Care
Medicine and Lippincott Williams & Wilkins
DOI: 10.1097/CCM.0b013e3181fa41a7
Projected increases in the incidence of
severe sepsis attributable to the shifting
demographics of the population under-
score the growing economic burden of
sepsis (1, 2).
Efforts to improve outcomes in severe
sepsis focus on a multifaceted approach
emphasizing early patient identification,
resuscitation, consideration of corticoste-
roids, and antibiotic therapy (3, 4). Many
recommendations for the treatment of
severe sepsis are now applied as part of a
“bundle-of-care” approach. Unfortu-
nately, many specific recommendations
in these bundles and in formal guidelines
remain controversial. Additionally, broad
application of guideline recommenda-
tions appears to have only a modest im-
pact on survival (5).
The need for appropriate antibiotic
therapy, however, represents one aspect
of sepsis care for which there is broad
consensus (6). Multiple studies document
that initially inappropriate antibiotic
therapy (IIAT) leads to excess mortality
(7–9). Furthermore, protocols that en-
compass specific antibiotic recommenda-
tions that reflect local microbiology and
resistance patterns result in improved
outcomes in severe sepsis (10). A recent
systematic assessment of key components
of sepsis therapy indicates that initially
appropriate antibiotic therapy may be the
most consistent intervention physicians
can undertake in hopes of enhancing sur-
vival (6).
Ensuring initially appropriate antibi-
otic therapy is frequently challenging be-
cause it often necessitates the use of
broad-spectrum agents. This is particu-
larly true as the epidemiology and extent
of resistance in Gram-negative patho-
gens, which are major causes of severe
sepsis, continue to evolve. Broad-spec-
trum therapy helps guarantee that the
patient is treated with an antimicrobial to
which the causative agent is fully suscep-
tible, is active in vitro, and is a central

46 Crit Care Med 2011 Vol. 39, No. 1

aspect of a de-escalation approach to an-
timicrobial administration. Use of initial
broad-spectrum therapy has led to con-
cern regarding expense—agents with
more expansive antibacterial activity tend
to be more costly. Alternatively, ensuring
initially appropriate antibiotic therapy
might improve financial outcomes by fa-
cilitating a patient’s rapid recovery.
We sought to determine how initial
antibiotic therapy in Gram-negative se-
vere sepsis affected a patient’s length of
stay (LOS) after sepsis onset. We hypoth-
esized that IIAT would be associated with
a longer hospital LOS.
MATERIALS AND METHODS
Study Overview
We conducted a retrospective review of all
patients with Gram-negative severe sepsis or
septic shock at a single institution between
January 2002 and December 2007. The study
was conducted at Barnes-Jewish Hospital, a
1,200-bed academic urban teaching hospital.
This cohort was drawn from an earlier report
describing the microbiology of Gram-negative
sepsis (11). The institutional review board ap-
proved this study; informed consent was not
required because of the retrospective nature of
the analysis. Hospital LOS after the diagnosis
of severe sepsis served as our primary end
point.
Subjects
We included adult patients with severe sep-
sis based on the accepted American College of
Chest Physicians/Society of Critical Care Med-
icine consensus definition for this syndrome
(12). We also required evidence of a Gram-
negative pathogen based on blood cultures.
Identification of the primary source of infec-
tion was made by review of blood cultures,
along with cultures of the airways (both spu-
tum and lower airways), urine, intra-abdomi-
nal sites, and central venous catheters. We
enrolled patients with polymicrobial infec-
tions so long as both pathogens were Gram-
negative organisms. Patients were entered
into the analysis once persistent bacteremia
was defined as present only if the first negative
blood culture occurred ⬎5 days after the first
positive blood culture. If a patient experienced
multiple episodes of Gram-negative sepsis, then
we included only the first episode. We assessed
patients with community-acquired, healthcare-
associated, and nosocomial processes.
Definitions and Covariates
All definitions were chosen prospectively.
We defined IIAT as occurring if the patient did
Crit Care Med 2011 Vol. 39, No. 1
not receive an anti-infective active in vitro
against the culprit Gram-negative bacteria re-
covered. Additionally, an active antimicrobial
had to be administered within 24 hrs of the
eventually positive culture being drawn.
Hence, subjects either receiving delayed treat-
ment or having never been prescribed an ac-
tive antimicrobial were defined as being ad-
ministered IIAT. All in vitro testing was
performed in the microbiology laboratory via
Kirby-Bauer disk diffusion following the
guidelines of and the breakpoints established
by the Clinical Laboratory and Standards In-
stitute. All interpretations were made by
trained microbiology technicians.
We recorded information regarding demo-
graphics (e.g., age, gender, race), comorbidi-
ties, the infection itself, and acute severity of
illness. Specifically we determined if the pa-
tients had coronary artery disease, chronic
obstructive pulmonary disease, diabetes, end-
stage renal disease, immunosuppression, ma-
lignancy, and other conditions. We also noted
whether the patients came from a nursing
home and their Charlson comorbidity scores
(13). To assess the severity of illness, we re-
corded whether the subject required care in
the intensive care unit, mechanical ventila-
tion, or vasopressor administration, and we
determined the Acute Physiology and Chronic
Health Evaluation II score (14). Further, we
assessed, as others have, the number of organ
failures based on International Classification
of Diseases 9th edition discharge diagnoses
codes, and these were confirmed by chart re-
view to ensure that they coincided with the
episode of bacteremia (1). For therapy, we
recorded whether the patient received dro-
trecogin alfa (activated). For nosocomial in-
fections, we assessed the LOS before sepsis
onset, with day 0 defined as the day cultures
were obtained. For all infections, we ascer-
tained the primary site of infection. We finally
documented whether the patient survived the
index hospitalization.
Statistics
Categorical variables were compared with
Fisher’s exact test. We compared normally dis-
tributed continuous variables via Student’s t
test while using the Mann-Whitney U test for
nonparametric continuous factors. We con-
structed survival curves according to the
methods of Kaplan and Meier to examine the
time to discharge as a function of the appro-
priateness of antibiotic treatment. These
curves were compared via log-rank test. All
tests were two tailed and unpaired, whereas a
p value of ⬍.05 was assumed to represent
statistical significance.
To evaluate the independent impact of IIAT
on LOS, we created a Cox proportional-
hazards model. As an initial step, we deter-
mined that the proportional-hazards assump-
Figure 1. Pathogen distribution. E. coli, Esche-
richia coli; P. aeruginosa, Pseudomonas aerugi-
nosa; K. pneumoniae, Klebsiella pneumoniae.
tion was not violated in our data, and we
examined all variables were examined to assess
for possible colinearity. A priori, we placed in
this model factors that we felt were biologi-
cally likely to affect LOS, irrespective of IIAT.
Specifically, we controlled for age, gender, co-
morbidities, infection source, severity of ill-
ness, preinfection LOS (for nosocomial infec-
tions), and hospital mortality. We also
controlled for other variables that differed
based on the appropriateness of initial treat-
ment at the level of p ⬍ .15. IIAT was retained
in the model because this was the focus of our
analysis. We determined the LOS related to
IIAT by comparing the adjusted Kaplan-Meier
curves reflecting LOS for those administered
IIAT and those treated appropriately. Because
of potential interactions between preinfection
LOS and postinfection LOS, we conducted a
sensitivity analysis only in those patients with
nosocomial infection. Similarly, we performed
a sensitivity analysis in hospital survivors
only. All analyses were completed using SPSS
17.0 (SPSS, Chicago, IL).
RESULTS
The final cohort included 760 patients
(mean age 59.3 ⫾ 16.3 yrs, median Acute
Physiology and Chronic Health Evalua-
tion II score 24.0, interquartile range
18.0 –28.0), and the crude inhospital mo-
rality rate was 41.2%. Nearly one-third of
subjects (31.3%) received IIAT. Of per-
sons administered IIAT (n ⫽ 238), we
observed a delay in therapy as the reason
in 58.0%. In the remainder, the pathogen
was resistant in vitro to the antibiotics
administered. The distribution of patho-
gens recovered is shown in Figure 1. En-
terobacteriaceae species along with Kleb-
siella species comprised the majority of
isolates, whereas Pseudomonas aerugi-
nosa and Acinetobacter species ac-
47
Table 1. Patient characteristics home. There were no differences with
respect to either the prevalence of other
Inappropriate Appropriate
comorbidities or the Charlson score. Al-
Antibiotics Antibiotics
Characteristic (n ⫽ 238) (n ⫽ 522) p though there were no differences be-
tween the groups in terms of Acute Phys-
Demographics iology and Chronic Health Evaluation II
Age, mean ⫾ SD (yr) 57.7 ⫾ 15.8 59.9 ⫾ 16.5 .082 score, organ failures, and need for vaso-
Male 48.7% 54.2% .183
pressors, patients receiving IIAT more of-
Chronic health state
Immunosupressed 32.4% 34.3% .620 ten needed mechanical ventilation at sep-
Chronic dialysis 14.7% 7.1% .001 sis onset (62.6% vs. 51.5%, p ⫽ .005).
Nursing home resident 13.4% 18.2% .035 Individuals administered IIAT more of-
Coronary artery disease 11.7% 7.9% .123
Chronic obstructive pulmonary disease 21.6% 17.2% .179
ten experienced nosocomial infections
Congestive heart failure 21.6% 18.1% .308 and, reflecting this, had a longer me-
Malignancy 23.1% 34.1% .002 dian LOS in the hospital before the on-
Diabetes mellitus 27.5% 20.1% .034 set of severe sepsis (9 days vs. 1 day, p ⫽
Charlson score, mean ⫾ SD 4.8 ⫾ 3.7 4.8 ⫾ 3.7 .994
Disease severity .001). There was no correlation between
Acute Physiology and Chronic Health 23.2 ⫾ 6.6 23.9 ⫾ 6.7 .203 the distribution of primary sites of in-
Evaluation II, mean ⫾ SD fection and IIAT. IIAT correlated with
Need for mechanical ventilation 62.6% 51.5% .005 an increased risk of death (odds ratio
Need for vasopressors 59.2% 58.0% .812
Organ failures, mean ⫾ SD 2.3 ⫾ 1.0 2.2 ⫾ 1.1 .265
1.87, 95% CI 1.37–2.55). Persistent bac-
Treatment with drotrecogin alfa (activated) 3.8% 4.4% .846 teremia was more common in patients
Infection characteristics administered IIAT (odds ratio 3.56, 95%
Nosocomial 69.3% 48.7% .001 CI 1.55– 8.17).
Community-acquired 5.9% 11.1%
Healthcare-associated 24.8% 40.2% Figure 2 reveals the probability of re-
Additional factors maining hospitalized after the diagnosis
Length of stay before infection (mean ⫾ SD) 15.3 ⫹ 20.7 7.5 ⫹ 14.9 .001 of severe sepsis as a function of IIAT.
Length of stay before infection (median) 9 1 .001 Subjects treated with inappropriate ther-
Hospital mortality 51.7% 36.4% .001
apy had a median increase of 2 days in
hospital LOS (11 days vs. 8 days, p ⫽
.028). In the Cox model adjusting for
multiple covariates (Table 2), IIAT re-
mained linked with an increased proba-
bility of remaining hospitalized. The ad-
justed hazard ratio related to IIAT
equaled 1.18 (95% CI 1.02–1.39). Other
variables significantly correlated with re-
maining hospitalized included need for
mechanical ventilation, preinfection
LOS, and inhospital mortality. After con-
trolling for these and the other factors
displayed in Table 2, we estimate that
IIAT independently increased LOS by ap-
proximately 2 days. The adjusted LOS of
patients initially treated appropriately
was 8 days vs. 10 days in subjects receiv-
ing IIAT. Among those with nosocomial
infection (n ⫽ 403), the adjusted hazard
Figure 2. Impact of inappropriate therapy on length of stay. ratio associated with IAT equaled 1.12
(95% CI 1.01–1.65), whereas both prein-
fection LOS and hospital mortality re-
counted for approximately one-quarter of mented in Table 1. Subjects receiving mained associated with LOS. In the co-
infections. Most patients (78.9%) were IIAT were slightly younger than those hort of hospital survivors (n ⫽ 447), the
treated in the intensive care unit. The administered appropriate treatment. Per- adjusted hazard ratio for IIAT approached
urine was the most common site of pri- sons treated with IIAT were also more statistical significance (adjusted hazard
mary infection (51.1%), followed by the likely to be undergoing chronic dialysis, ratio 1.12, 95% CI 0.990 –1.41, p ⫽ .07),
lung (38.8%). In 9.9% of cases, the pri- to have underlying cancer, and to have and the need for mechanical ventilation,
mary source was unclear. diabetes mellitus. Persons administered preinfection LOS, nosocomial-type infec-
Baseline characteristics of persons IIAT were also 1.73 times (95% confi- tion, and Acute Physiology and Chronic
treated with IIAT as compared to appro- dence interval [CI] 1.07–2.81) more likely Health Evaluation II score were indepen-
priate antimicrobial therapy are docu- to have been admitted from a nursing dently linked to LOS.

48 Crit Care Med 2011 Vol. 39, No. 1

Table 2. Independent factors associated with length of staya sequently, longer durations of hospital-
ization (17). That the correlation between
Hazard 95% Confidence IIAT and longer LOS exists in multiple
Variable Ratio Interval p
types of infections suggests that the
Mechanical ventilation 1.27 1.09–1-49 .003 theme here is consistent across a spec-
Inappropriate antibiotic therapy 1.18 1.01–1.39 .044 trum of patients. In other words, our
Preculture length of stay (days) 0.98 0.97–0.99 .001 observations are not unique to individu-
Inhospital mortality 0.64 0.54–0.74 .001 als with severe sepsis or septic shock and
therefore do not reflect a chance finding.
Hazard ratios of ⬎1 indicate that the factor is associated with a greater probability of remaining
hospitalized. Hence, both the need for mechanical ventilation and inappropriate antibiotic therapy
In septic shock, one potential reason that
correlate with a greater chance of remaining in the hospital. For hazard ratios of ⬍1, those variables some studies indicate that bundled ap-
correlate with a greater chance of no longer being hospitalized (i.e., being discharged sooner). For proaches to care have only a modest im-
example, patients who died while hospitalized are more likely to be discharged. Similarly, the shorter pact on outcomes is perhaps because in
the hospital length of stay before sepsis onset, the more likely the patient is to be no longer many of the reports describing sepsis
hospitalized after the suspicion of infection onset. bundles and protocols, rates of IIAT
a
In creating the Cox proportional-hazards model, variables were selected for entry in the model change minimally or are not even de-
based on one of two criteria. First, we entered into the model factors we felt were a priori biologically scribed. In reports detailing the experi-
linked to length of stay (e.g., severity of illness, hospital mortality). Second, we included in the model ence with protocols and bundles for sep-
other variables that differed in univariate analysis at p ⬍ .15.
sis in which there has been a larger,
positive effect on outcomes, particularly
DISCUSSION Expanding on their work, the same inves- economic ones, these bundles have
tigators explored survival in nearly 5,700 clearly emphasized reducing rates of IIAT
This retrospective analysis of a large persons with septic shock (15). They (10). In that same vein, clinicians should
group of patients with Gram-negative se- noted a clear adverse impact of IIAT determine ways to minimize IIAT. Possi-
vere sepsis and septic shock documents across a range of primary infection and ble approaches for achieving this include
that IIAT is prevalent and leads to excess the creation of antibiotic prescribing pro-
pathogen types.
morbidity. Specifically, IIAT results in an tocols, routine discussion of antibiotic
Our results add to these earlier analy-
approximately 2-day increase in hospital management strategies in multidisci-
ses by focusing on an end point other
LOS. This negative effect on LOS is inde- plinary meetings, and earlier consulta-
than mortality. Failure to appreciate how
pendent of multiple confounders, includ- tion with experts in infectious diseases.
one’s approach to sepsis management af-
ing both comorbid diseases and severity Why does appropriate antibiotic ther-
fects resource use and morbidity may re-
of illness. Given the costs associated with apy result in more-rapid discharge? The-
sult in myopic resource allocation. For
a day in the hospital and the escalating oretically, the application of an antibiotic
example, through an appreciation of the
prevalence of severe sepsis and septic that kills the culprit pathogen early in the
shock, our results suggest that IIAT likely potential economic costs of IIAT, physi- subject’s disease may facilitate recovery.
has major economic implications. cians can make a strong case to hospital Through eliminating the pathogen that
Previous studies have examined the administrators and national policy mak- has unleashed the cascades that result in
significance of IIAT in severe infections ers about the potential value to be found severe sepsis and possibly shock, homeosta-
generally and in sepsis specifically. In a in addressing rates of IIAT. In addition, sis can be more quickly achieved. Further-
cohort of approximately 655 critically ill via emphasizing the economic sequelae more, in organisms that release toxins
patients with a variety of infections, Ibra- of management decisions made early in a that either initiate or propagate sepsis
him et al (7) determined that one-quarter patient’s hospital course, one can rein- physiology, rapid bacterial eradication
of such individuals received IIAT. Crude force the fact that the approach to the prevents the pathogen from manufactur-
hospital mortality rates were doubled in septic patient must be multidisciplinary ing or releasing (or both) more toxin. If
those treated inappropriately and, after and aggressive. In infectious syndromes the patients recover more quickly and
adjusting for potential covariates, these other than severe sepsis and septic shock, progression to advanced organ failure
investigators estimated that IIAT more decreasing rates of IIAT have become a ceases, then they can be more rapidly
than tripled a patient’s risk for death. focus of efforts to improve the quality of transitioned out of the hospital. A shorter
Among patients with fungemia, which care. In part, motivating such initiatives overall LOS also protects the patient from
may arise as a complication of intensive is recognition that poor antibiotic many of the later insults that can in-
care unit care, delays in administration of choices can cost the institution and soci- crease the risk of death in the septic pa-
antifungal therapy may more than double ety economically. tient, including nosocomial infection,
the risk of hospital mortality (8). This The present findings are consistent Clostridium difficile colitis, and venous
relationship persists even after adjusting with earlier studies addressing the mor- thromboembolism.
for severity of illness, comorbid condi- bidity cost associated with IIAT. Shorr et As a practical matter, the relationship
tions, and catheter management. More al (16) documented that IIAT in methicil- between IIAT and both mortality and
recently, Kumar et al (9) examined out- lin-resistant Staphylococcus aureus in- morbidity underscores the need for a par-
comes in 2,500 subjects with septic fections led to excessive hospital LOS. adigm for antibiotic utilization that em-
shock. They concluded that each 1-hr de- Similarly, failure to follow national phasizes initial broad-spectrum coverage.
lay in the administration of appropriate guidelines for antibiotic prescribing in Thus, physicians should adopt a de-escala-
antibiotics was associated with a ⬎10% community-acquired pneumonia results tion approach to antibiotic prescribing. Be-
increase in the probability of death (9). in more patients receiving IIAT and, con- cause physicians cannot successfully antic-

Crit Care Med 2011 Vol. 39, No. 1 49

ipate the culprit pathogen based purely on
history, it becomes crucial for clinicians to
treat with a broad-spectrum antibiotic reg-
imen designed to cover the most common
resistant pathogens at play in the infectious
syndrome in question. This imperative sug-
gests an urgent need for clinicians to be
aware of their local microbiology. As a cor-
ollary, intensivists, hospitalists, emergency
department physicians, and infectious dis-
ease practitioners must collaborate with
those in microbiology and infection control
to collect data that reliably allow one to
make an informed antibiotic selection. Fur-
thermore, the present results suggest that
technologies that facilitate rapid patho-
gen identification and, in the future, may
allow one to know antimicrobial suscep-
tibilities quickly are both urgently needed
and likely to be cost effective. Short of
such diagnostic methods, paradigms that
arbitrarily restrict access to antibiotics
may be short-sighted because they expose
patients to the risk of IIAT and the insti-
tutional costs related to it.
Our study has several important limi-
tations. First, the retrospective design
highlights that our study may be prone to
various forms of bias. To try to minimize
bias, we chose variable definitions pro-
spectively. Second, the definitions of se-
vere sepsis and septic shock are not com-
pletely objective. Thus, we may have
failed to recognize some patients with
true sepsis while enrolling other individ-
uals who simply had an uncomplicated
infection. This fact may further have been
compounded by the fact that we required
positive blood cultures in all subjects. An
added concern regarding definitions is
how we elected to describe IIAT. There is
no formally endorsed definition of IIAT,
and we utilized one that is necessarily
simple. Proper administration of antibi-
otics required more than timeliness and
in vitro activity. Issues of dosing and
route logically should be considered as
well. We did not explore these specific
issues because there is no consensus re-
garding the optimal dosing for many anti-
infectives; however, at the same time, the
definition of IIAT we relied on has been
applied by others in multiple analyses (7–
11). Third, the data are derived from a
single center. This necessarily limits the
generalizability of our conclusions to
other settings and institutions. Similarly,
we addressed only Gram-negative patho-
gens and therefore cannot immediately
extrapolate our findings to subjects with
Gram-positive infections. Fourth, we may
have failed to account for other impor-
50
tant confounders that might contribute
to LOS. For instance, we lacked data on
the duration of residence in a nursing
home. Any statistical model that ad-
dresses imbalances between groups has
limitations. We believe that we identified
the major factors that could have driven
outcomes, including often overlooked
but nonetheless significant issues such as
previous duration of hospitalization.
Fifth, we did not collect information about
whether a pathogen developed resistance
during what was classified as initially ap-
propriate therapy. This certainly may
have had an effect on outcomes. Finally,
we do not mean for readers to interpret
our study as a formal cost-effectiveness
analysis. A dedicated cost-effectiveness
analysis would necessitate an evaluation
of many other factors, including the im-
pact of antibiotic therapy on superinfec-
tions, the emergence of resistance, and
quality of life. Our findings, however, can
serve as potential inputs for standard
cost-effectiveness modeling.
In conclusion, IIAT in Gram-negative
severe sepsis and septic shock is nega-
tively correlated with LOS. IIAT appears
to be associated with a lengthening of the
hospital stay by approximately 2 days.
Identifying means to reduce rates of IIAT
may limit the morbidity associated with
severe sepsis and septic shock.
REFERENCES
1. Angus DC, Linde-Zwirble WT, Lidicker J, et
al: Epidemiology of severe sepsis in the
United States: Analysis of incidence, out-
come, and associated costs of care. Crit Care
Med 2001; 29:1303–1310
2. Martin GS, Mannino DM, Eaton S, et al: The
epidemiology of sepsis in the United States
from 1979 through 2000. N Engl J Med 2003;
348:1546 –1554
3. Dellinger RP, Levy MM, Carlet JM, et al:
Surviving Sepsis Campaign: International
guidelines for management of severe sepsis
and septic shock: 2008. Crit Care Med 2008;
36:296 –327
4. Ferrer R, Artigas A, Levy MM, et al: Im-
provement in process of care and outcome
after a multicenter severe sepsis educa-
tional program in Spain. JAMA 2008; 299:
2294 –2303
5. Levy MM, Dellinger RP, Townsend SR, et
al: The Surviving Sepsis Campaign: Results
of an international guideline-based perfor-
mance improvement program targeting
severe sepsis. Crit Care Med 2010; 38:
367–374
6. Barochia AV, Cui X, Vitberg D, et al: Bundled
care for septic shock: An analysis of clinical
trials. Crit Care Med 2010; 38:668 – 678
7. Ibrahim EH, Sherman G, Ward S, et al: The
influence of inadequate antimicrobial treat-
ment of bloodstream infections on patient
outcomes in the ICU setting. Chest 2000;
118:146 –155
8. Labelle AJ, Micek ST, Roubinian N, Kollef
MH: Treatment-related risk factors for hos-
pital mortality in Candida bloodstream infec-
tions. Crit Care Med 2008; 36:2967–2972
9. Kumar A, Roberts D, Wood KE, et al: Dura-
tion of hypotension before initiation of effec-
tive antimicrobial therapy is the critical de-
terminant of survival in human septic shock.
Crit Care Med 2006; 34:1589 –1596
10. Micek ST, Roubinian N, Heuring T, et al:
Before-after study of a standardized hospital
order set for the management of septic
shock. Crit Care Med 2006; 34:2707–2713
11. Micek ST, Welch EC, Khan J, et al: Empiric
combination antibiotic therapy is associated
with improved outcome against sepsis due to
Gram-negative bacteria: A retrospective anal-
ysis. Antimicrob Agents Chemother 2010; 54:
1742–1748.
12. Bone RC, Balk RA, Cerra FB, et al: Defini-
tions for sepsis and organ failure and guide-
lines for the use of innovative therapies in
sepsis The ACCP/SCCM Consensus Confer-
ence Committee American College of Chest
Physicians/Society of Critical Care Medicine
1992. Chest 2009; 136(Suppl 5):e28
13. Schneeweiss S, Maclure M: Use of comorbid-
ity scores for control of confounding in stud-
ies using administrative databases. Int J Epi-
demiol 2000; 29:891– 898
14. Knaus WA, Draper EA, Wagner DP, et al:
APACHE II: A severity of disease classification
system. Crit Care Med 1985; 13:818 – 829
15. Kumar A, Ellis P, Arabi Y, et al: Initiation of
inappropriate antimicrobial therapy results
in a fivefold reduction of survival in human
septic shock. Chest 2009; 136:1237–1248
16. Shorr AF, Micek ST, Kollef MH: Inappropri-
ate therapy for methicillin-resistant Staphy-
lococcus aureus: Resource utilization and
cost implications. Crit Care Med 2008; 36:
2335–2340
17. Shorr AF, Bodi M, Rodriguez A, et al: Impact
of antibiotic guideline compliance on dura-
tion of mechanical ventilation in critically ill
patients with community-acquired pneumo-
nia. Chest 2006; 130:93–100
APPENDIX
During the time of this study, the antimi-
crobial control program at Barnes-Jewish
Hospital allowed unrestricted use of cefepime
and gentamicin. Initiation of intravenous cip-
rofloxacin, imipenem/cilastatin, meropenem,
or piperacillin/tazobactam was restricted and
required preauthorization from either a clini-
cal pharmacist or an infectious disease physi-
cian. Each intensive care unit had a clinical
pharmacist who reviewed all antibiotic orders
to ensure that the dosing and interval of an-
tibiotic administration were adequate for indi-
Crit Care Med 2011 Vol. 39, No. 1
vidual patients based on body size, renal func-
tion, and the resuscitation status of the
patient. Beginning in June 2005, a sepsis order
set was implemented in the emergency depart-
ment, general medical wards, and intensive
Crit Care Med 2011 Vol. 39, No. 1
care units with the intent of standardizing
empirical antibiotic selection for patients
with sepsis based on the infection type (i.e.,
community-acquired pneumonia, health-
care-associated pneumonia, intra-abdominal
infection, etc.) and the local antibiogram.
However, antimicrobial selection, dosing,
and de-escalation of therapy were still opti-
mized by clinical pharmacists in these clin-
ical areas.
51