antibiotic resistance in intensive care units · therefore, emergence and spread of antimicrobial...
TRANSCRIPT
Antibiotic Resistance in Intensive Care Units
Dynamics of Colonization
Saskia Nijssen
Antibiotic Resistance in Intensive Care Units
Dynamics of Colonization
Antibiotica-resistentie op intensive care-afdelingen
Dynamiek van kolonisatie
(met een samenvatting in het Nederlands)
Proefschrift
ter verkrijging van de graad van doctor aan de Universiteit Utrecht
op gezag van de rector magnificus, prof. dr. W.H. Gispen,
ingevolge het besluit van het college voor promoties in het
openbaar te verdedigen op vrijdag 1 september 2006 des ochtends
te 10.30 uur
door
Saskia Nijssen
geboren op 9 september 1976 te Maarssen
Promotoren: Prof. dr. M.J.M. Bonten
Prof. dr. I.M. Hoepelman
Co-promotor: Dr. A.C. Fluit
Cover: Skyline Chicago
Lay-out: Alarick Verhoef
Printed by: Ponsen & Looijen B.V. Wageningen
ISBN-10: 90-393-4317-9
ISBN-13: 978-90-393-4317-3
Copyright: “Bacteria with suitcase” is derived from an issue of the Economist
2000. Permission to use this picture for this thesis has been requested on
June 13th 2006.
© 2006 S. Nijssen, Utrecht, The Netherlands
Dit proefschrift is tot stand gekomen door financiële ondersteuning van de
vakgroep Acute Geneeskunde en Infectieziekten van het UMC Utrecht en
ZonMw, projectnummer 2100.0051
Voor mijn ouders en Evelyne
Contents
Introduction to the thesis, objectives and questions 11
Chapter 1 Potential confounding in evaluating infection 19
control interventions in hospital settings: changing
antibiotic prescription
Chapter 2 Are active microbiological surveillance and 47
subsequent isolation needed to prevent the
spread of methicillin-resistant Staphylococcus aureus?
Chapter 3 The relative risk of physicians and nurses to transmit 63
pathogens in a medical intensive care unit
Chapter 4 Unnoticed spread of integron-carrying 71
Enterobacteriaceae in intensive care units
Chapter 5 Determining the relative importance of 95
bacterial transmission routes in hospital settings
Chapter 6 Comparison of E-tests and double disk diffusion 115
tests for the detection of Extended Spectrum
Beta-Lactamases (ESBLs)
Chapter 7 β-Lactam susceptibilities and prevalence of ESBL- 125
producing isolates among more than 5000 European
Enterobacteriaceae isolates
Chapter 8 A step-wise reduction of β-lactam exposure, with 145
control of all relevant confounders, failed to reduce
acquisition of third-generation cephalosporin-resistant
Enterobacteriaceae in two intensive care units
Chapter 9 General Discussion 176
Nederlandse Samenvatting 193
Dankwoord 203
Curriculum Vitae 211
List of Publications 213
Introduction to the thesis, objectives and questions
Introduction to the thesis
In 1945 Fleming received the Nobel Prize for his discovery of the
antimicrobial effects of penicillin. In his acceptance speech, he already
warned for the danger of antibiotic-resistance. In the past sixty years, many
classes of antimicrobials have been developed, but duration of benefit
appeared to be limited: resistance has emerged to every antimicrobial class.
Antimicrobial resistance seriously hampers treatment of nosocomial
infections and leads to increased morbidity, mortality and healthcare costs.
This is especially true for patients admitted to intensive care units: these
patients are severely ill, and thus vulnerable to infection. Furthermore,
consumption of antimicrobials is usually higher here than in other wards.
Therefore, emergence and spread of antimicrobial resistance is, apart from
being an ongoing threat in itself, an even bigger threat to treatment and
outcome of intensive care patients. Antimicrobial resistance epidemiology is
characterized by complex interactions between pathogen, host, other
microbial flora, and the environment. So, control of emergence and spread of
resistance is complicated. Knowledge about local epidemiology and
colonization dynamics of key pathogens is indispensable for the hospital’s
hygiene department in order to recognize changes in time and to design
control strategies.
12
Objectives and questions In this thesis, we will focus on the colonization dynamics of resistant
pathogens in intensive care units and investigate how different measures will
influence these.
In chapter 1, we describe the different determinants of colonization, how
these interact and the methodological consequences resulting from these
interactions for studies evaluating the effects of interventions on colonization
dynamics. In addition, we review to what extent the different aspects of this
methodology are addressed in studies evaluating the effects of changes in
antimicrobial prescription.
In chapters 2-5, we will focus on the roles of microbiological surveillance
(including molecular typing), monitoring of infection control and
mathematical modeling in determining variables important in colonization
dynamics and their interactions.
In chapter 2, the need of microbiological surveillance and subsequent
isolation in the prevention of spread of both methicillin-sensitive and
methicillin-resistant Staphylococcus aureus (MSSA and MRSA) was investigated
in a medical intensive care unit where these pathogens are endemic.
Contact rates, cohorting and adherence to hand hygiene are important
determinants of colonization. As these variables all interact with each other
and the magnitude varies between different types of healthcare workers, we
investigate how these affect the relative risk for physicians and nurses to
transmit pathogens in chapter 3.
Resistance can spread by different modes: by transmission of complete
bacteria and by horizontal gene transfer, i.e. transmission of genetic elements
between bacteria of the same species and different species. Horizontal gene
13
transfer in Enterobacteriaceae is mediated by conjugative plasmids and
transposons. Integrons, which are mobile elements incorporated in
chromosomal DNA, plasmids or transposons are associated with multi-
resistance in Enterobacteriaceae. In chapter 4, we determine the prevalence
of integrons in different species of Enterobacteriaceae with reduced
susceptibility to cephalosporins (ERSC) and the individual contributions of
bacterial transmission and horizontal gene transfer to the spread of resistance
in two intensive care units, with integrons as markers for resistance.
Molecular typing of bacterial isolates is labour-intensive, time-consuming and
expensive and, therefore, not performed routinely. As an alternative to these
techniques, mathematical modeling has been proposed recently. Using
longitudinal surveillance data on culture results at regular intervals between
admission and discharge, mean endemic prevalence of pathogens and the
relative contributions of endogenous and exogenous colonization routes
(with 95% confidence intervals) can be estimated without the need of
genotyping. In chapter 5, we determine the accuracy of a Markov model as a
tool to determine the relative importance of bacterial transmission routes in
the intensive care by validation of model predictions using longitudinal
surveillance and genotyping data from two intensive care units.
In the second part of this thesis, we focus on β-lactam resistance in
Enterobacteriaceae, chapter 6, 7, 8. ESBL-production in Enterobacteriaceae,
collected from 25 European hospitals, was determined by E-test ESBL and
the double disk diffusion test (DDT) and a comparison to determine the
individual performance of both tests is described in chapter 6. The
perspective of β-lactam resistance in European Enterobacteriaceae is
subsequently described in chapter 7. In chapter 8, colonization dynamics of
third-generation cephalosporin-resistant Enterobacteriaceae (CRE) in two
14
intensive care units of the University Medical Centre Utrecht were
determined. Prevalence and incidence of colonization, dominant route of
acquisition, risk factors for acquisition, antimicrobial use, contact rates,
cohorting levels and adherence to hand hygiene were determined during a so-
called baseline period of 8 months. Based on the findings during baseline, an
intervention to reduce acquisition rates of CRE was designed and the effects
were evaluated, using methods identical to baseline and with measurement of
all other variables. The intervention consisted of two three-month
antimicrobial regimens implemented in a crossover design: a heterogeneous
regimen consisting of weekly cycling of empirical antimicrobial therapy (three
different classes) and a homogeneous regimen during which empirical therapy
with a single antimicrobial class was used.
To summarize, the main questions to be answered in this thesis are:
1. What are the determinants of colonization dynamics, how do these
interact with each other and what are the methodological
consequences of these interactions for the evaluation of intervention
studies? Chapter 1.
2. To what extent are the aspects of this methodology addressed in
studies evaluating changes in antimicrobial prescription? Chapter 1.
3. Are active microbiological surveillance and subsequent isolation
needed to prevent the spread of methicillin-resistant Staphylococcus
aureus? Chapter 2.
15
4. To what extent do contact rates, cohorting levels and adherence to
hand hygiene affect the risk of pathogen transmission for nurses and
physicians? Chapter 3.
5. To what extent do cross-transmission of bacteria and horizontal gene
transfer (i.e. integrons) contribute to the spread of resistance?
Chapter 4.
6. Can mathematical modeling be used as a tool to accurately estimate
the relative importance of endogenous and exogenous colonization
routes in intensive care units? Chapter 5.
7. Are E-test ESBL and double disk diffusion test suitable for the
detection of ESBLs in Enterobacteriaceae? Chapter 6.
8. What is the perspective for β-lactam resistance in European
Enterobacteriaceae isolates, in particular for third-generation
cephalosporins and ESBL-production? Chapter 7.
9. What are the prevalence and incidence of colonization with, risk
factors for and predominant acquisition routes of CRE in two
intensive care units? Chapter 8.
10. What are the effects of a heterogeneous and a homogeneous
antimicrobial regimen to reduce β-lactam exposure on the acquisition
rates of CRE in these intensive care units? Chapter 8.
16
Chapter
Potential confounding in evaluating infection control interventions in hospital settings: changing antibiotic prescription
Nijssen S, Bootsma M, Bonten M
Clinical Infectious Diseases; Accepted for publication on September 1st 2006
Abstract
Colonization dynamics of antibiotic-resistant pathogens in hospital settings
are complex, due to multiple interacting variables. Determination and control
of these variables (introduction of resistance, infection control practices and
antibiotic use) is indispensable in the evaluation of intervention studies, as
these are potential confounders. The objective of this review is to describe
the complexity of colonization dynamics and to evaluate to what extent
confounding has been controlled for in interventions aimed at modifying
antibiotic prescription.
Potential confounders (introduction of resistance, infection control practices)
were either not measured or changed during the study period and, it remains
uncertain whether observed changes in antibiotic resistance prevalence after
intervention were causally related to the intervention. Choosing an
appropriate study design (randomized controlled trial vs. before-after study)
and primary end-point (colonization rather than infection rates), determining
colonization routes and controlling potential confounders will increase
validity of conclusions drawn from intervention studies.
20
Introduction
Nosocomial infections, especially those caused by antibiotic-resistant
pathogens, are a serious complication in critically ill patients. Infections are
usually preceded by colonization and the number of colonized patients
exceeds, by far, the number of prominent infections (tip of the iceberg
phenomenon). In this review we describe the complexity of colonization
dynamics in hospital settings and how different processes interact. As a result
of these interactions, confounding may complicate evaluation of
interventions. Subsequently, we evaluate to what extent confounding has
been controlled for in intervention trials on modification of antibiotic use in
intensive care units (ICUs). The endemic prevalence, usually expressed as the
average daily proportion of patients colonized with a pathogen of interest, in
a hospital ward can change due to three processes: a) admission and discharge
of colonized and non-colonized patients, b) de novo resistance development or
eradication of susceptible flora and selection of pre-existent resistant flora
and c) patient-to-patient transmission of resistant strains or resistance
determinants [1] (Figure 1).
Figure 1. Processes changing endemic prevalence of antibiotic-resistant pathogens.
Introduction of resistant pathogens
Admission/transfer of
colonized patients
Endogenous colonization
De novo resistance
Selection of pre-existent resistant flora
Changes in the prevalence of
antibiotic-resistant pathogens
Exogenous colonization
Failures in infection control
21
De novo resistance development implies processes in which susceptible
bacteria become resistant to antibiotics through mutations or horizontal gene
transfer (conjugation, transduction or transformation). Subsequently, selective
antibiotic pressure may facilitate overgrowth of these resistant strains.
Transmission of pathogens from patient to patient usually occurs via the
hands of healthcare workers or through use of contaminated equipment. The
success of this route depends on colonization pressure, patient’s bacterial
load, cohorting levels, contact rates, adherence to hand hygiene of staff, and
the susceptibility of non-colonized patients for pathogen acquisition [1,2].
Selective antibiotic pressure enhances the risk of transmission by increasing a
patient’s bacterial load (by selection of pre-existent resistant flora) with
subsequent risk of hand contamination in healthcare workers, and by creating
new ecological niches for resistant flora after eradication of susceptible flora
in other patients [3].
Due to these different colonization processes, the relatively small numbers of
patients in a ward (typically 10-20) and the characteristic rapid patient
turnover (with average stay of only several days), prevalence levels of
colonized patients within a hospital ward continuously fluctuate due to events
occurring just by chance [1,4,5]. Understanding the dynamics of colonization
is indispensable to design appropriate and targeted infection control strategies.
Measures to decrease prevalence of colonization with antibiotic-resistant
bacteria target the three processes depicted in Figure 1 and include barrier
precautions, improving hand hygiene, cohorting of patients, reducing
overcrowding and understaffing, and changing antimicrobial prescription
(restriction, rotation or cycling of antimicrobial agents).
22
Confounding in intervention studies
Strategies to reduce the nosocomial prevalence of antibiotic resistance have
been evaluated in both outbreak and non-outbreak periods, and almost
always include the implementation of a combination of measures, which
hampers evaluation of the effect of individual interventions. Moreover, such
interventions are almost always evaluated in quasi-experimental study designs
(i.e. before-after studies), which increase the risk of confounding.
Confounders are those variables that may affect the outcome in the same
matter as the variable subject to intervention. The optimal approach to
evaluate the efficacy of a single intervention is to minimize confounding
either by performing a randomized-controlled trial or by quantifying potential
confounders with subsequent adjustments in statistical analysis. Considering
the dynamics of antibiotic resistance epidemiology, introduction of resistance,
antibiotic pressure, and infection control measures all are potential
confounders in intervention studies.
Introduction of resistance
Introduction of resistance can be determined by obtaining cultures on
admission.
Ideally, patients should be cultured as soon as possible, as acquisition of
resistant pathogens occurs within several hours. Nosocomial infections have
been defined as those diagnosed more than 48 hours after hospital admission
[6,7]. In many studies, a similar time window has been used to define
nosocomial acquisition of colonization. Yet, the optimal timing of culturing
to distinguish between introduction and nosocomial acquisition remains to be
23
determined. Screening within 48 hours of admission probably yields the
majority of patients introducing resistant pathogens into an ICU.
For pathogens that persist until ICU-discharge screening on admission and at
discharge would yield all information necessary to determine acquisition rates.
More culture moments per patient are needed when colonization can be
eradicated or when daily endemic prevalence is used as end-point.
Antibiotic use
Different unities are used to express antibiotic usage: proportion of patients
receiving a specific (class of) antibiotic, antibiotics used as a proportion of
total antibiotic use, the amount of antibiotics given (e.g. in grams or defined
daily dosage (DDD)) and days of antibiotic use. Integration of time,
preferably in a patient-specific manner like the number of patient-days,
provides more information. To maximize comparison of patient exposure,
the World Health Organization proposed and defined the defined daily
dosage methodology [8].
Cross-transmission
Dissection of the process leading to cross-transmission, typically through
vectors (such as contaminated hands), identifies a multi-step process with
complex interaction between all determinants. The first step is that a
healthcare worker must contact a colonized patient. Then, this contact must
lead to contamination of the healthcare worker’s hands for some time. And
finally, the healthcare worker with contaminated hands must contact another,
non-colonized patient, before disinfection of the hands has occurred. The
24
success of this chain of events depends on cohorting level of nursing staff,
contact rates and adherence to hand hygiene [1].
Cohorting has been expressed as the likelihood that, after a first patient
contact, the next patient contact will be with the same patient [1,9].
Transmission is not possible with complete cohorting, as this indicates a
patient/nurse ratio of 1 and no other patients than the assigned are contacted
(assuming that environmental contamination is not relevant). Thus, cohorting
levels are, to a large extent, influenced by patient-staffing ratios.
Understaffing of nursing personnel has been associated with higher contact
rates for nurses [10,11], higher nursing workload, lower cohorting levels and
lower hand hygiene adherence [9,11-15]. The likelihood of hand
contamination is further influenced by duration of patient care, types of body
secretions handled [10], skin diseases, such as eczema [16], or wearing rings
or artificial nails [17]. In fact, understaffing and/or overcrowding has been
associated with increased risk of catheter-related bloodstream infections [18]
and spread of methicillin-resistant Staphylococcus aureus, methicillin-sensitive
Staphylococcus aureus [4,14,15,19] and Enterobacter cloacae[13].
Cohorting levels and adherence to hand hygiene are important variables in
cross-transmission. Lapses in adherence to hand hygiene can be compensated
by an increase in the cohorting level and vice versa [9]. Therefore, cohorting
levels and adherence to hand hygiene, as potential confounders, should be
determined in intervention studies. Cohorting levels and hand hygiene can be
measured by observational studies [4,9,12], which are, however, labour-
intensive and always include the risk of unintentional change in behaviour of
healthcare workers due to the observation (the so-called Hawthorne effect)
[20]. Instead, staffing-patient ratios and contact rates are relatively easy to
record and can be used as surrogate markers for cohorting [4,9]. Intuitively,
25
workload assessment could be a surrogate marker for contact rates and hand
hygiene adherence, but quantifying this relationship is complex. Several
nursing workload measurement systems are available [21-24]. Yet, there is
only little evidence that changes in workload correlate to frequencies of cross-
transmission [25]. In our own experience, the Therapeutic Intervention
Severity Score_28 (TISS_28) workload assessment method was poorly
associated with contact rates and adherence to hand hygiene [26].
Endogenous and exogenous acquisition
Optimally, the relative importance of exogenous and endogenous acquisition
of colonization should be determined before implementing interventions. For
instance, enforcing adherence to hand hygiene might be of limited value
when the vast majority of acquisitions result from endogenous selection or
when admission of colonized patients is the dominant variable determining
prevalence. Currently, genotyping of bacterial isolates from different patients
is the most accurate method to quantify cross-transmission rates. Yet, it is a
time-consuming, labour-intensive and costly method and the diagnostic delay
- inherent to conventional culturing and genotyping methods - precludes real-
time determination of resistance epidemiology. Rapid diagnostic and
genotyping methodology and application of mathematical algorithms may
facilitate real-time monitoring of the relative importance of different
transmission routes in the future, which would improve our ability to design
targeted infection control strategies [27,28].
26
Statistical evaluation
Finally, accurate evaluation of interventions needs application of appropriate
statistical methods. The relevance of colonization pressure in antibiotic
resistance epidemiology has been unequivocally demonstrated [2,29,30].
Colonization pressure reflects to patient-dependency, a fundamental aspect of
infectious diseases. Cross-transmission, depends, amongst other variables, on
the number of other patients being colonized: with high endemic prevalence,
risk of cross-transmission will be higher than with low prevalence and vice
versa (i.e., autocorrelation). Moreover, a period of high endemic prevalence,
with a high incidence of cross-transmission, is likely to be followed by a
period with lower prevalence, just because of chance events (i.e. regression to
the mean). In contrast, endogenous selection of resistance is not influenced
by the colonization status of other patients, making patient-dependency
irrelevant and autocorrelation and regression to the mean less important. As a
consequence, patients cannot be considered to be independent from each
other when cross-transmission is relevant. Yet, most statistical analyses
explicitly assume independency of observations (e.g. Student’s T-test, Mann-
Whitney U test, Fisher’s exact test and χ2-test). As a matter of fact, the use of
these statistical tests may lead to erroneous interpretations. As an example we
have performed 100.000 Monte Carlo simulations of the dynamics of a
resistant pathogen in a 10-bed ICU, where 80% of acquisitions occur through
cross-transmission, during a one-year period (see Figure 2 for more details).
If we assume an intervention implemented after six months, without any
effect on cross-transmission though, up to 30% of statistical comparisons
(using χ2-test or Student’s T-test) would reveal a statistical significant
difference in antibiotic resistance acquisition and/or prevalence between both
periods.
27
The likelihood of false statistical interpretation decreases with a diminishing
relative importance of endogenous acquisition (lowering the relevance of
patient-dependency) [31].
Figure 2. Fraction of 100,000 simulations with a statistical significant result using χ2-test with p-value of 0.05 as a function of the relative importance of cross-transmission. The red and blue lines correspond with two-sided and one-sided test respectively.
28
Evaluation of potential confounding in studies changing
antibiotic prescription
In order to evaluate to what extent confounding has been controlled for in
intervention trials, we systematically reviewed studies in which the effects of
changes in antibiotic prescription on the prevalence and incidence of
colonization and/or infection with antibiotic-resistant bacteria in ICUs were
evaluated.
PubMed was searched with the following search terms, first as individual
term, subsequently in a combined approach: antibiotic or antimicrobial,
resistance, ICU, intensive care, colonization, infection. Selection criteria
included: antibiotic-resistant bacteria of any kind, interventions targeting
antibiotic use to reduce prevalence and/or incidence of antibiotic-resistant
bacteria (restriction of specific classes or individual antibiotics, rotation
and/or cycling) and non-outbreak settings. Reports in languages other than
English or without abstract, and reviews were excluded.
This search yielded 1017 articles, which were first screened by title and
abstract to determine if selection criteria were indeed met. Studies only
published as abstracts or without abstracts were excluded. Ultimately, 19
studies, performed between 1984 and 2006, which met the criteria mentioned,
were reviewed (Table 1a-1e).
These articles were reviewed for the following items: pathogen of interest,
type of intervention, study design, presence of a baseline period if applicable,
chosen end-points, determination of the route of acquisition, antibiotic use,
determination of potential confounders (introduction of resistance and
infection control practices) and interpretation of results (Table 1a-1e).
29
Study design, baseline and intervention period
Fourteen studies focussed on gram-negative bacteria (GNB) [32-44], two
studies addressed VRE [30-45] and three studies evaluated all types of
antibiotic-resistant pathogens [46-48].
The interventions tested were antimicrobial cycling with recurrence of the
initial regimen (A-B-A) (n=2) [39,49], antimicrobial rotation with different
antibiotic regimens (A-B-C-D) (n=4) [35,37,38,47], antimicrobial rotation per
month vs. per consecutive patient (n=1) [44] and antimicrobial restriction or
substitution (n=12) [30,32-34,36,40-43,46,48].
Sixteen studies had a prospective cohort design executed within a single unit,
two had a prospective crossover design in two wards and one had a
controlled design in two neonatal ICU populations. In twelve out of sixteen
studies that did not have a simultaneously studied control group, end-points
and potential confounders were first determined during a baseline period and
then, with identical methodology, during the intervention period (before-after
studies). Cohort studies with different antibiotic policies (B-C-D) compared
to a standard policy (A) were considered as before-after studies if separate
analyses between cycles B, C and D were not performed.
End-points of analysis
Colonization rates were the primary end-point in ten studies [30,32,34-
36,39,40,42-44], six studies used antimicrobial susceptibility rates of clinical
isolates, and, therefore, determined infection rates [33,37,41,46-48] and both
colonization and infection rates were used as end-points in four studies
[30,38,44,49].
30
Relative importance of acquisition routes
The relative importance of different acquisition routes was not determined in
any study, though some information about transmission dynamics was
provided by genotyping of selected isolates in four studies. This revealed
polyclonal presence of cefuroxime-resistant GNB (n=1) [42], monoclonal
epidemics of ceftazidime-resistant GNB (n=1) [34], low cross-transmission
rates of ceftazidime-resistant GNB (n=1) [35] and monoclonal spread of
fluoroquinolone resistant pathogens in combination with monoclonal and
polyclonal spread of β-lactam resistant pathogens (n=1) [39].
Antibiotic use
Antibiotic use, the subject of intervention in each study, was analysed in 16 of
18 studies. There is, however, no uniform measure of antibiotic use and time
components were also variably used. Six of the reviewed studies expressed
use of a specific antibiotic as percentage of total antibiotic use (n=1) [49], as
percentage of patients receiving antibiotics (n=2) [30,33,38], as number of
courses/100 enrolled patients/ICU admissions (n=2) [46,47], or as total
grams used (n=1) [43], thus, not using time in the denominator. Eleven
studies integrated time in the expression of antibiotic use, either as number of
dosages or grams/month (n=2) [34,41]), or as antibiotic-days (n=4)
[30,35,37,40], or as percentages of ICU-days (n=2) [36,48], or as DDD/1000
patient-days (n=3) [39,44,45]. Two studies expressed antibiotic use with and
without time components [30,44].
31
Control of potential confounders
Introduction of resistance, defined as carriage of resistant bacteria within the
first 48 hours of admission, was measured in ten studies [30,32,34,36,38-
40,43,44,49]. Neither adherence to hand disinfection, nor cohorting levels,
staffing levels or workload were determined in any of these 19 studies. In five
studies, infection control strategies were even implemented or enhanced
while interventions were ongoing [38-40,44,47].
Interpretation of efficacy
In twelve studies, authors concluded that the intervention was effective in
reducing the prevalence of resistance, either with or without increased
resistance in other pathogens [32,36,37,40-48]. Others concluded that their
intervention had no effect (n=6) [30,34,35,38,39,49] and/or an opposite
effect (n=1) [33] on antibiotic resistance. All studies used standard statistics
(i.e., Student’s T-test, χ2-test) for data analysis.
32
Tab
le 1a
. Stu
dies
on
effe
cts o
f ant
ibio
tic p
resc
riptio
n ch
ange
.
a A
bbre
viat
ions
: GN
B: g
ram
-neg
ativ
e ba
cter
ia; G
PB: g
ram
-pos
itive
bac
teria
; GE
N: g
enta
mici
n; A
MK
: am
ikac
in; N
ET
: net
ilmyc
in; T
OB:
tobr
amyc
in; C
EP:
ce
phalo
spor
ins C
AZ: c
efta
zidi
me;
FEP:
cef
epim
e; C
TX
: cef
otax
ime;
CXM
: cef
urox
ime;
CPI:c
efpi
rom
; CR
O: c
eftri
axon
e; T
ZP: p
iper
acill
in/t
azob
acta
m; T
IM:
ticar
cillin
/clav
ulan
ic ac
id; A
MX
: am
oxic
illin
; AM
C: a
mox
icill
in/c
lavul
anic
acid
; AM
P: a
mpi
cillin
; Pen
: pen
icilli
n; C
LI: c
linda
myc
in; F
Q: f
luor
oqui
nolo
nes;
CIP
: ci
prof
loxa
cin;
LVX
: lev
oflo
xaci
n; S
XT
: trim
etho
prim
/sul
fam
etho
xazo
le; M
EM
: mer
open
em; I
PM: i
mip
enem
Scor
ed it
em
Raz
et a
l. (3
2)
Infe
ctio
n 19
87
Ham
mon
d et
al.
(33)
Cr
it Ca
re M
ed
1990
Kale
nic
et a
l. (4
2)
J Hos
p In
fect
19
93
Toltz
is et
al.
(34)
Cr
it Ca
re M
ed
1998
St
udy
desi
gn
Befo
re-a
fter s
tudy
Be
fore
-afte
r stu
dy
Retro
spec
tive
and
pros
pect
ive
data
ana
lyse
s Be
fore
-afte
r stu
dy
Path
ogen
of in
terest
G
NBa
G
NB
GN
B G
NB
Inter
ventio
n Su
bstit
utio
n:
GE
N b
y A
MK
Su
bstit
utio
n:
GE
N/T
OB/
NE
T by
AM
K
Subs
titut
ion:
A
MP/
GE
N b
y CX
M/G
EN
Re
stric
tion:
CA
Z
Basel
ine
Yes
N
o Y
es, a
lthou
gh re
trosp
ectiv
e Y
es
End
-poi
nt o
f ana
lysi
s Co
loni
zatio
n ra
tes
Infe
ctio
n ra
tes
Colo
niza
tion
& In
fect
ion
rate
s Co
loni
zatio
n ra
tes
Gen
otypin
g of i
solate
s N
ot p
erfo
rmed
N
ot p
erfo
rmed
Y
es
Yes
Antib
iotic
use
N
ot a
naly
zed
% P
atie
nts r
eceiv
ing
antib
iotic
s N
ot a
naly
zed
n do
ses/
mon
th
Con
trol o
f con
foun
ders
Intro
ducti
on of
ant
ibioti
c resi
stanc
e A
nalyz
ed
Not
ana
lyzed
A
nalyz
ed
Ana
lyzed
Infec
tion
contro
l pra
ctices
N
ot a
naly
zed
N
ot a
naly
zed
Not
ana
lyze
d N
ot a
naly
zed
Con
clus
ion
auth
ors
Redu
ctio
n of
GE
N-re
sista
nt G
NB.
N
o em
erge
nce
of A
MK
-resis
tant
G
NB
Incr
ease
in re
sista
nce
to
AM
K/N
ET/
TOB
Redu
ctio
n in
AM
P- a
nd C
XM
-re
sista
nt G
NB
No
redu
ctio
n in
an
tibio
tic-re
sista
nt G
NB
33
Tab
le 1b
. Stu
dies
on
effe
cts o
f ant
ibio
tic p
resc
riptio
n ch
ange
.
a A
bbre
viat
ions
: GN
B: g
ram
-neg
ativ
e ba
cter
ia; G
PB: g
ram
-pos
itive
bac
teria
; GE
N: g
enta
mici
n; A
MK
: am
ikac
in; N
ET
: net
ilmyc
in; T
OB:
tobr
amyc
in; C
EP:
ce
phalo
spor
ins C
AZ: c
efta
zidi
me;
FEP:
cef
epim
e; C
TX
: cef
otax
ime;
CXM
: cef
urox
ime;
CPI:
cefp
irom
; CR
O: c
eftri
axon
e; T
ZP: p
iper
acill
in/t
azob
acta
m; T
IM:
ticar
cillin
/clav
ulan
ic ac
id; A
MX
: am
oxic
illin
; AM
C: a
mox
icill
in/c
lavul
anic
acid
; AM
P: a
mpi
cillin
; PE
N: p
enici
llin;
CLI
: clin
dam
ycin
; FQ
: flu
oroq
uino
lone
s; C
IP:
cipr
oflo
xaci
n; L
VX: l
evof
loxa
cin;
SX
T: t
rimet
hopr
im/s
ulfa
met
hoxa
zole;
ME
M: m
erop
enem
; IPM
: im
ipen
em.
b Su
rveil
lance
team
and
alco
hol d
ispen
sers
impl
emen
ted
Scor
ed it
em
De
Man
et a
l. (3
6)
Lanc
et
2000
Gru
son
et a
l. (3
7)
Am
J Re
sp C
rit C
are
Med
20
00
Lan
et a
l. (4
3)
Shoc
k 20
00
Raym
ond
et a
l. (4
7)
Crit
Care
Med
20
01
Stud
y de
sign
Pr
ospe
ctiv
e cr
oss-
over
stud
y Be
fore
-afte
r stu
dy
Befo
re-a
fter s
tudy
Be
fore
-afte
r stu
dy
Path
ogen
of in
terest
G
NBa
G
NB
GN
B G
NB
and
GPB
In
terven
tion
Com
paris
on:
PEN
/TO
B vs
AM
X/C
TX
Rota
tion:
FE
P; T
ZP;
IPM
; TIM
Su
bstit
utio
n:
CAZ
by
TZP
Rota
tion:
non
-pr
otoc
oliz
ed v
s pr
otoc
oliz
ed e
mpi
rical
ther
apy
Basel
ine
Not
app
licab
le Y
es
Yes
Y
es
End
-poi
nt o
f ana
lysi
s Co
loni
zatio
n ra
tes
Infe
ctio
n ra
tes
Colo
niza
tion
rate
s In
fect
ion
rate
s
Gen
otypin
g of i
solate
s N
ot p
erfo
rmed
N
ot p
erfo
rmed
N
ot p
erfo
rmed
N
ot p
erfo
rmed
Antib
iotic
use
%
Of d
ays a
ntib
iotic
use
A
ntib
iotic
-day
s G
ram
s use
d A
ntib
iotic
cou
rses
/100
IC
U a
dmiss
ions
C
ontro
l of c
onfo
unde
rs
Intro
ducti
on of
ant
ibioti
c resi
stanc
e N
ot a
naly
zed
Not
ana
lyze
d A
naly
zed
Ana
lyse
d
Infec
tion
contro
l pra
ctices
N
ot a
naly
zed
N
ot a
naly
zed
Not
ana
lyze
d N
ot a
naly
zed
b
Con
clus
ion
auth
ors
Redu
ctio
n of
resis
tanc
e ra
tes G
NB
to 3
rd-g
ener
atio
n CE
P Re
duct
ion
of M
RSA
and
CRO
-re
sista
nt G
NB
Redu
ctio
n in
ant
ibio
tic-re
sista
nt
GN
B In
crea
se re
sista
nce
durin
g LV
X a
nd T
ZP
34
Tab
le 1c
. Stu
dies
on
effe
cts o
f ant
ibio
tic p
resc
riptio
n ch
ange
.
a A
bbre
viat
ions
: GN
B: g
ram
-neg
ativ
e ba
cter
ia; G
PB: g
ram
-pos
itive
bac
teria
; GE
N: g
enta
mici
n; A
MK
: am
ikac
in; N
ET
: net
ilmyc
in; T
OB:
tobr
amyc
in; C
EP:
ce
phalo
spor
ins C
AZ: c
efta
zidi
me;
FEP:
cef
epim
e; C
TX
: cef
otax
ime;
CXM
: cef
urox
ime;
CPI:c
efpi
rom
; CR
O: c
eftri
axon
e; T
ZP: p
iper
acill
in/t
azob
acta
m; T
IM:
ticar
cillin
/clav
ulan
ic ac
id; A
MX
: am
oxic
illin
; AM
C: a
mox
icill
in/c
lavul
anic
acid
; AM
P: a
mpi
cillin
; PE
N: p
enici
llin;
CLI
: clin
dam
ycin
; FQ
: flu
oroq
uino
lone
s; C
IP:
cipr
oflo
xaci
n; L
VX: l
evof
loxa
cin;
SX
T: t
rimet
hopr
im/s
ulfa
met
hoxa
zole;
ME
M: m
erop
enem
; IPM
: im
ipen
em.
Scor
ed it
em
Puzn
iak e
t al.
(30)
Cl
in In
fect
Dis
20
01
Alle
gran
zi e
t al.
(46)
J H
osp
Infe
ct
2002
Mos
s et a
l. (4
9)
Crit
Care
Med
20
02
Toltz
is et
al.
(35)
Pe
diat
rics
2002
St
udy
desi
gn
Befo
re-a
fter s
tudy
Be
fore
-afte
r stu
dy
Befo
re-a
fter s
tudy
Co
ntro
lled
stud
y of
2
NIC
U p
opul
atio
ns
Path
ogen
of in
terest
V
REa
GN
B an
d G
PB
GN
B
GN
B In
terven
tion
Subs
titut
ion:
CA
Z b
y CI
P Su
bstit
utio
n:
AM
C by
SX
T &
TZ
P by
IPM
Cycli
ng:
IPM
; TZ
P; C
AZ
&CL
I (lat
er
chan
ged
to F
EP)
Rota
tion:
G
EN
; TZ
P; C
AZ
vs.
unre
stric
ted
use
Basel
ine
Yes
Y
es
No
Not
app
licab
le
End
-poi
nt o
f ana
lysi
s Co
loni
zatio
n ra
tes
Infe
ctio
n ra
tes
Colo
niza
tion
& in
fect
ion
rate
s Co
loni
zatio
n ra
tes
Gen
otypin
g of i
solate
s N
ot p
erfo
rmed
N
ot p
erfo
rmed
N
ot p
erfo
rmed
Y
es
Antib
iotic
use
%
Of p
atien
ts re
ceiv
ing
spec
ific
antib
iotic
; mea
n du
ratio
n of
an
tibio
tic th
erap
y
Ant
ibio
tic c
ours
es/1
00 e
nrol
led
patie
nts
% O
f tot
al an
tibio
tic u
se
Ant
ibio
tic-d
ays
Con
trol o
f con
foun
ders
Intro
ducti
on of
ant
ibioti
c resi
stanc
e A
naly
zed
N
ot a
naly
zed
Ana
lyze
d N
ot a
naly
zed
Infec
tion
contro
l pra
ctices
N
ot a
naly
zed
Not
ana
lyze
d N
ot a
naly
zed
Not
ana
lyze
d
Con
clus
ion
auth
ors
No
effe
ct V
RE a
cqui
sitio
n Re
duct
ion
MRS
A o
f TZ
P-re
sista
nt
Pseu
domo
nas.
Incr
ease
IPM
-resis
tant
Ps
eudo
mona
s
No
effe
ct o
n re
sista
nce
rate
s N
o ef
fect
on
antib
iotic
-re
sista
nt G
NB
35
Tab
le 1d
. Stu
dies
on
effe
cts o
f ant
ibio
tic p
resc
riptio
n ch
ange
.
a A
bbre
viat
ions
: GN
B: g
ram
-neg
ativ
e ba
cter
ia; G
PB: g
ram
-pos
itive
bac
teria
; GE
N: g
enta
mici
n; A
MK
: am
ikac
in; N
ET
: net
ilmyc
in; T
OB:
tobr
amyc
in; C
EP:
ce
phalo
spor
ins C
AZ: c
efta
zidi
me;
FEP:
cef
epim
e; C
TX
: cef
otax
ime;
CXM
: cef
urox
ime;
CPI:c
efpi
rom
; CR
O: c
eftri
axon
e; T
ZP: p
iper
acill
in/t
azob
acta
m; T
IM:
ticar
cillin
/clav
ulan
ic ac
id; A
MX
: am
oxici
llin;
AM
C: a
mox
icilli
n/cla
vulan
ic a
cid; A
MP:
am
pici
llin;
Pen
: pen
icilli
n; C
LI: c
linda
myc
in; F
Q: f
luor
oqui
nolo
nes;
CIP
: ci
prof
loxa
cin;
LVX
: lev
oflo
xaci
n; S
XT
: trim
etho
prim
/sul
fam
etho
xazo
le; M
EM
: mer
open
em; I
PM: i
mip
enem
.
c Ba
rrier
pre
caut
ions
impl
emen
ted
beca
use
of re
sista
nce
to p
roto
coliz
ed d
rugs
.
Scor
ed it
em
Du
et a
l. (4
1)
Crit
Care
Med
20
03
Gei
ssler
et a
l. (4
8)
Inte
nsiv
e Ca
re M
ed
2003
Toltz
is et
al.
(40)
Pe
diat
r Inf
ect D
is J
2003
Van
Loo
n et
al.
(39)
A
m J
Resp
Crit
Car
e
2004
St
udy
desi
gn
Befo
re-a
fter s
tudy
Be
fore
-afte
r stu
dy
Befo
re-a
fter s
tudy
Pr
ospe
ctiv
e co
hort
stud
y
Path
ogen
of in
terest
G
NBa
M
RSA
and
GN
B G
NB
GN
B In
terven
tion
Rest
rictio
n:
3rd -g
ener
atio
n CE
P Su
bstit
utio
n:
Non
-pro
toco
lized
vs.
prot
ocol
ized
an
tibio
tics
Subs
titut
ion:
N
on-p
roto
coliz
ed v
s. FE
P Cy
cling
: LV
X; C
PI; L
VX
; TZ
P
Basel
ine
Yes
Y
es
Yes
N
o
End
-poi
nt o
f ana
lysi
s In
fect
ion
rate
s In
fect
ion
rate
s Co
loni
zatio
n ra
tes
Colo
niza
tion
rate
s
Gen
otypin
g of i
solate
s N
ot p
erfo
rmed
N
ot p
erfo
rmed
N
ot p
erfo
rmed
Y
es
Antib
iotic
use
G
ram
s/m
onth
A
ntib
iotic
-day
s/10
00 d
ays I
CU
pres
ence
A
ntib
iotic
-day
s D
DD
/100
0 pa
tient
-day
s
Con
trol o
f con
foun
ders
Intro
ducti
on of
ant
ibioti
c resi
stanc
e N
ot a
naly
zed
Not
ana
lyze
d A
naly
sed
Ana
lyse
d
Infec
tion
contro
l pra
ctices
N
ot a
naly
zed
Not
ana
lyze
d N
ot a
naly
zed
Not
ana
lyze
d c
Con
clus
ion
auth
ors
Redu
ctio
n of
resis
tanc
e ra
tes G
NB
to 3
rd-g
ener
atio
n CE
P Re
duct
ion
of M
RSA
and
CRO
-re
sista
nt G
NB
Redu
ctio
n in
ant
ibio
tic-re
sista
nt
GN
B In
crea
se re
sista
nce
durin
g LV
X a
nd T
ZP
36
Tab
le 1e
. Stu
dies
on
effe
cts o
f ant
ibio
tic p
resc
riptio
n ch
ange
. Sc
ored
item
W
arre
n et
al.
(38)
Cr
it Ca
re M
ed
2004
Win
ston
et a
l. (4
5)
Am
J In
fect
Con
trol
2004
Mar
tinez
et a
l. (4
4)
Crit
Care
Med
20
06
Stud
y de
sign
Pr
ospe
ctiv
e co
hort
stud
y Be
fore
-afte
r stu
dy
Pros
pect
ive
cros
s-ov
er st
udy
Path
ogen
of in
terest
G
NB
VRE
G
NB
Inter
ventio
n Ro
tatio
n:
FEP;
FQ
; IPM
; TZ
P Su
bstit
utio
n:
TIM
by
TZP
Rota
tion
per m
onth
vs.
rota
tion
per c
onse
cutiv
e pa
tient
of F
EP/
CAZ
; CIP
; ME
M/I
PM; T
ZP
Ba
seline
Y
es
Yes
N
ot a
pplic
able
End
-poi
nt o
f ana
lysi
s Co
loni
zatio
n &
infe
ctio
n ra
tes
Colo
niza
tion
& in
fect
ion
rate
s
Colo
niza
tion
& in
fect
ion
rate
s
Gen
otypin
g of i
solate
s N
ot p
erfo
rmed
N
ot p
erfo
rmed
N
ot p
erfo
rmed
Antib
iotic
use
%
Of p
atie
nts r
ecei
ving
spec
ific
antib
iotic
D
DD
/100
0 pa
tient
-day
s %
Of p
atien
ts re
ceiv
ing
spec
ific
antib
iotic
; D
DD
/100
pat
ient
-day
s C
ontro
l of c
onfo
unde
rs
Intro
ducti
on of
ant
ibioti
c resi
stanc
e A
naly
zed
Not
ana
lyze
d A
naly
zed
Infec
tion
contro
l pra
ctices
N
ot a
nalyz
edd
Not
ana
lyze
d
Not
ana
lyze
de
Con
clus
ion
auth
ors
No
effe
ct o
n re
sista
nce
rate
s Re
duct
ion
in V
RE
acqu
isitio
n
Redu
ctio
n of
FE
P-re
sista
nt P
seudo
mona
s ae
rugin
osa d
urin
g cy
cling
a A
bbre
viat
ions
: GN
B: g
ram
-neg
ativ
e ba
cter
ia; G
PB: g
ram
-pos
itive
bac
teria
; GE
N: g
enta
mici
n; A
MK
: am
ikac
in; N
ET
: net
ilmyc
in; T
OB:
tobr
amyc
in; C
EP:
ce
phalo
spor
ins C
AZ: c
efta
zidi
me;
FEP:
cef
epim
e; C
TX
: cef
otax
ime;
CXM
: cef
urox
ime;
CPI:
cefp
irom
; CR
O: c
eftri
axon
e; T
ZP: p
iper
acill
in/t
azob
acta
m; T
IM:
ticar
cillin
/clav
ulan
ic ac
id; A
MX
: am
oxici
llin;
AM
C: a
mox
icilli
n/cla
vulan
ic a
cid; A
MP:
am
pici
llin;
Pen
: pen
icilli
n; C
LI: c
linda
myc
in; F
Q: f
luor
oqui
nolo
nes;
CIP
: ci
prof
loxa
cin;
LVX
: lev
oflo
xaci
n; S
XT
: trim
etho
prim
/sul
fam
etho
xazo
le; M
EM
: mer
open
em; I
PM: i
mip
enem
. d
Edu
catio
nal c
ampa
ign
to p
reve
nt V
AP
resu
lted
in 7
3% re
duct
ion
of V
AP.
e Inf
ectio
n co
ntro
l cha
nged
bec
ause
of A
cineto
bacte
r out
brea
k..
37
Conclusions
Colonization dynamics of resistant pathogens in small hospital settings, like
intensive care units, are complex with multiple relevant variables, all
interacting. Therefore, there is a high risk of confounding in intervention
studies, which may seriously hamper interpretation of study results. As an
example, nineteen studies on antibiotic interventions were systematically
reviewed to determine to what extent potential confounding had been
controlled for.
In all studies, potential confounders (introduction of resistance, infection
control practices) were either not measured or changed during the study
period. Therefore, it remains uncertain whether observed changes in
antibiotic resistance prevalence after intervention were causally related to the
intervention. Moreover, even absence of efficacy could have resulted from
opposing effects due to confounding, despite an effective intervention.
We, therefore, propose the following four points in order to reduce
confounding in intervention trials addressing antibiotic resistance in ICUs.
First, the optimal study design would be a randomized, controlled trial with
each participating ward as unit of study (i.e., cluster-randomized trial).
Needless to say that such a trial will be expensive. A quasi-experimental
design (such as a before-after study) might be an alternative, as long as results
are interpreted carefully; as such a design inherently increases the likelihood
of confounding, regression to the mean and maturation effects [50]. To
provide more internal validity and potential causation between intervention
and outcome Harris et al. proposed a hierarchy in quasi-experimental designs,
reflecting designs that include or do not include control groups [51]. Second,
colonization rates are preferred over infection rates, as the latter only
represents the tip of the iceberg. Third, determination of the relative
38
importance of acquisition routes is indispensable to optimally target
interventions and for choosing the appropriate statistical methods for analysis.
Genotyping of colonizing isolates in combination with epidemiological
linkage still is the standard to distinguish between exogenous and endogenous
colonization events. As mentioned, these methods are labour-intensive and,
therefore, hardly feasible in daily practice outside dedicated research settings.
Recently, mathematical models have been proposed as alternatives for
determining transmission parameters on the basis of longitudinal prevalence
data [4,27,52]. With these so-called Markov chain models the relative
importance of either cross-transmission or the endogenous route can be
determined upon observed fluctuations in prevalence. Importantly, standard
statistical tests, all assuming independency of observations, are only valid
when, indeed, patient-dependency is not relevant. When cross-transmission is
important, these tests should be interpreted with care as inflated rates of type
I errors are likely to occur [28]. Again, recent developments in biostatistical
analyses may offer better alternatives for the future [27,52]. Finally, potential
confounders should be quantified and included in the final analyses. Again,
for practical reasons, a balance between the optimum and feasibility should
be sought and several, easy to obtain, proxies could be used.
Antibiotic resistance will remain a relevant problem in ICUs in the coming
decades and with no new antibiotic classes on the horizon, minimizing
further emergence of resistance is of utmost importance. Characteristics of
ICU-populations and colonization dynamics enhance the risk of confounding
when analyzing control measures. The measures as proposed in this review
will reduce the risk of false interpretation of study results.
39
References
1. Austin DJ, Bonten MJ, Weinstein RA et al. Vancomycin-resistant
enterococci in intensive-care hospital settings: transmission dynamics,
persistence, and the impact of infection control programs. Proc Natl
Acad Sci USA 1999:6908-13.
2. Bonten MJ, Slaughter S, Ambergen AW et al. The role of "colonization
pressure" in the spread of vancomycin- resistant enterococci: an
important infection control variable. Arch Intern Med 1998:1127-32.
3. Lipsitch M, Samore MH. Antimicrobial use and antimicrobial
resistance: a population perspective. Emerg Infect Dis 2002:347-54.
4. Grundmann H, Hori S, Winter B et al. Risk factors for the
transmission of methicillin-resistant Staphylococcus aureus in an adult
intensive care unit: fitting a model to the data. J Infect Dis 2002:481-8.
5. Bonten MJ, Bergmans DC, Speijer H et al. Characteristics of
polyclonal endemicity of Pseudomonas aeruginosa colonization in intensive
care units. Implications for infection control. Am J Respir Crit Care
Med 1999:1212-9.
6. Garner JS, Jarvis WR, Emori TG et al. CDC definitions for
nosocomial infections, 1988. Am J Infect Control 1988:128-40.
7. Consensus paper on the surveillance of surgical wound infections. The
Society for Hospital Epidemiology of America; The Association for
Practitioners in Infection Control; The Centres for Disease Control;
The Surgical Infection Society. Infect Control Hosp Epidemiol
1992:599-605.
8. ATC index with DDDs. WHO Collaborating Centre for Drug
Statistics Methodology, Oslo. 1999.
40
9. Nijssen S, Bonten MJ, Franklin C et al. Relative risk of physicians and
nurses to transmit pathogens in a medical intensive care unit. Arch
Intern Med 2003:2785-6.
10. Pittet D, Mourouga P, Perneger TV. Compliance with handwashing in
a teaching hospital. Infection Control Program. Ann Intern Med
1999:126-30.
11. Gould D. Systematic observation of hand decontamination. Nurs
Stand 2004:39-44.
12. Pittet D, Dharan S, Touveneau S et al. Bacterial contamination of the
hands of hospital staff during routine patient care. Arch Intern Med
1999:821-826.
13. Harbarth S, Sudre P, Dharan S et al. Outbreak of Enterobacter cloacae
related to understaffing, overcrowding, and poor hygiene practices.
Infect Control Hosp Epidemiol 1999:598-603.
14. Haley RW, Cushion NB, Tenover FC et al. Eradication of endemic
methicillin-resistant Staphylococcus aureus infections from a neonatal
intensive care unit. J Infect Dis 1995:614-24.
15. Vicca AF. Nursing staff workload as a determinant of methicillin-
resistant Staphylococcus aureus spread in an adult intensive therapy unit. J
Hosp Infect 1999:109-13.
16. Larson EL, Hughes CA, Pyrek JD et al. Changes in bacterial flora
associated with skin damage on hands of healthcare personnel. Am J
Infect Control 1998:513-21.
17. Trick WE, Vernon MO, Hayes RA et al. Impact of ring wearing on
hand contamination and comparison of hand hygiene agents in a
hospital. Clin Infect Dis 2003:1383-90.
41
18. Fridkin SK, Pear SM, Williamson TH et al. The role of understaffing in
central venous catheter-associated bloodstream infections. Infect
Control Hosp Epidemiol 1996:150-8.
19. Andersen BM, Lindemann R, Bergh K et al. Spread of methicillin-
resistant Staphylococcus aureus in a neonatal intensive unit associated with
understaffing, overcrowding and mixing of patients. J Hosp Infect
2002:18-24.
20. Reiss F. The Hawthorne Effect in a Pilot Program. Thesis 1979.
21. Reis MD, Moreno R, Iapichino G. Nine equivalents of nursing
manpower use score (NEMS). Intensive Care Med 1997:760-5.
22. Miranda DR, de Rijk A, Schaufeli W. Simplified Therapeutic
Intervention Scoring System: the TISS-28 items--results from a
multicenter study. Crit Care Med 1996:64-73.
23. Urbanowicz JA. An evaluation of an acuity system as it applies to a
cardiac catheterization laboratory. Comput Nurs 1999:129-34.
24. Trundle CM, Farrington M, Anderson L et al. GRASPing infection: a
workload measurement tool for infection control nurses. J Hosp Infect
2001:215-21.
25. Soulier A, Barbut F, Ollivier JM et al. Decreased transmission of
Enterobacteriaceae with extended-spectrum β-lactamases in an
intensive care unit by nursing reorganization. J Hosp Infect 1995:89-97.
26. Nijssen S, Bonten M, Hoepelman IM. Can the therapeutic intervention
severity score_28 (TISS_28) be used as a tool to determine contact
rates between healthcare workers and ICU-patients? 42nd
International Conference on Antimicrobial Agents and Chemotherapy,
San Diego 2002:K-1357.
42
27. Pelupessy, Bonten, Diekmann. How to assess the relative importance
of different colonization routes of pathogens within hospital settings.
Proc Natl Acad Sci USA 2002:5601-5606.
28. Cooper BS, Stone SP, Kibbler CC et al. Isolation measures in the
hospital management of methicillin resistant Staphylococcus aureus
(MRSA): systematic review of the literature. BMJ 2004:533.
29. Merrer J, Santoli F, Appéré-De Vecchi C et al. Colonization pressure
and risk of aquisition of methicillin-resistant Staphylococcus aureus in a
medical intensive care unit. Infect Control Hosp Epidemiol 2000:718-
723.
30. Puzniak LA, Mayfield J, Leet T et al. Acquisition of vancomycin-
resistant enterococci during scheduled antimicrobial rotation in an
intensive care unit. Clin Infect Dis 2001:151-7.
31. Bootsma MCJ, Diekmann O, Nijssen S et al. Estimating transmission
parameters by direct likelihood computation. In preparation.
32. Raz R, Sharir R, Shmilowitz L et al. The elimination of gentamicin-
resistant gram-negative bacteria in a newborn intensive care unit.
Infection 1987:32-4.
33. Hammond JM, Potgieter PD, Forder AA et al. Influence of amikacin
as the primary aminoglycoside on bacterial isolates in the intensive care
unit. Crit Care Med 1990:607-10.
34. Toltzis P, Yamashita T, Vilt L et al. Antibiotic restriction does not alter
endemic colonization with resistant gram-negative rods in a pediatric
intensive care unit. Crit Care Med 1998:1893-9.
35. Toltzis P, Dul MJ, Hoyen C et al. The effect of antibiotic rotation on
colonization with antibiotic-resistant bacilli in a neonatal intensive care
unit. Pediatrics 2002:707-11.
43
36. de Man P, Verhoeven BA, Verbrugh HA et al. An antibiotic policy to
prevent emergence of resistant bacilli. Lancet 2000:973-8.
37. Gruson D, Hilbert G, Vargas F et al. Rotation and restricted use of
antibiotics in a medical intensive care unit. Impact on the incidence of
ventilator-associated pneumonia caused by antibiotic-resistant gram-
negative bacteria. Am J Respir Crit Care Med 2000:837-43.
38. Warren DK, Hill HA, Merz LR et al. Cycling empirical antimicrobial
agents to prevent emergence of antimicrobial-resistant gram-negative
bacteria among intensive care unit patients. Crit Care Med 2004:2450-6.
39. van Loon HJ, Vriens MR, Fluit AC et al. Antibiotic rotation and
development of gram-negative antibiotic resistance. Am J Respir Crit
Care Med 2004:480-7.
40. Toltzis P, Dul M, O'Riordan MA et al. Cefepime use in a pediatric
intensive care unit reduces colonization with resistant bacilli. Pediatr
Infect Dis J 2003:109-14.
41. Du B, Chen D, Liu D et al. Restriction of third-generation
cephalosporin use decreases infection-related mortality. Crit Care Med
2003:1088-93.
42. Kalenic S, Francetic I, Polak J et al. Impact of ampicillin and
cefuroxime on bacterial colonization and infection in patients on a
neonatal intensive care unit. J Hosp Infect 1993:35-41.
43. Lan CK, Hsueh PR, Wong WW et al. Association of antibiotic
utilization measures and reduced incidence of infections with
extended-spectrum β-lactamase-producing organisms. J Microbiol
Immunol Infect 2003:182-6.
44
44. Martinez JA, Nicolas JM, Marco F et al. Comparison of antimicrobial
cycling and mixing strategies in two medical intensive care units. Crit
Care Med 2006:329-36.
45. Winston LG, Charlebois ED, Pang S et al. Impact of a formulary
switch from ticarcillin-clavulanate to piperacillin-tazobactam on
colonization with vancomycin-resistant enterococci. Am J Infect Control
2004:462-9.
46. Allegranzi B, Luzzati R, Luzzani A et al. Impact of antibiotic changes
in empirical therapy on antimicrobial resistance in intensive care unit-
acquired infections. J Hosp Infect 2002:136-40.
47. Raymond DP, Pelletier SJ, Crabtree TD et al. Impact of a rotating
empiric antibiotic schedule on infectious mortality in an intensive care
unit. Crit Care Med 2001:1101-8.
48. Geissler A, Gerbeaux P, Granier I et al. Rational use of antibiotics in
the intensive care unit: impact on microbial resistance and costs.
Intensive Care Med 2003:49-54.
49. Moss WJ, Beers MC, Johnson E et al. Pilot study of antibiotic cycling
in a pediatric intensive care unit. Crit Care Med 2002:1877-82.
50. Harris AD, Lautenbach E, Perencevich E. A systematic review of
quasi-experimental study designs in the fields of infection control and
antibiotic resistance. Clin Infect Dis 2005:77-82.
51. Harris AD, Bradham DD, Baumgarten M et al. The use and
interpretation of quasi-experimental studies in infectious diseases. Clin
Infect Dis 2004:1586-91.
52. Cooper B, Lipsitch M. The analysis of hospital infection data using
hidden Markov models. Biostatistics. 2004:223-37.
45
Chapter
Are active microbiological surveillance and subsequent isolation needed to prevent the spread of methicillin-resistant Staphylococcus aureus?
Nijssen S, Bonten M, Weinstein R
Clinical Infectious Diseases 2005; 40:405-9
Abstract
Infection control strategies usually combine several interventions. The
relative value of individual interventions, however, is rarely determined. We
assessed the effect of daily microbiological surveillance alone (e.g., without
report of culture results or isolating colonized patients) as an infection
control measure on the spread of methicillin-susceptible Staphylococcus aureus
(MSSA) and methicillin-resistant Staphylococcus aureus (MRSA) in a medical
intensive care unit (MICU). Colonization of patients with MSSA and MRSA
was assessed by cultures of nasal swabs obtained daily and, if a patient was
intubated, by cultures of additional endotracheal aspirates. Pulsed-Field Gel
Electrophoresis (PFGE) was used to determine relatedness between MSSA
or MRSA isolates in surveillance cultures (i.e., cultures of nasal swab
specimens obtained daily) and those in clinical cultures (i.e., any other culture
performed for clinical purposes). Adherence to infection control measures by
healthcare workers (HCWs) was determined by observations of HCW-patient
interaction. During a 10-week period, surveillance cultures were performed
for 158 patients. Fifty-five patients (34.8%) were colonized with MSSA, and 9
(5.7%) were colonized with MRSA. Sixty-two patients were colonized before
admission to the hospital (53 had MSSA, and 9 had MRSA). Two patients
appeared to have acquired MSSA in the MICU, but, on the basis of
genotyping analysis, we determined that this was not the result of cross-
acquisition.
Surveillance cultures and genotyping of MRSA and MSSA isolates
demonstrated the absence of cross-transmission among patients in the MICU,
despite ongoing introduction of these pathogens. Reporting culture results
and isolating colonized patients, as suggested by some guidelines, would have
falsely suggested the success of such infection control policies.
48
Introduction
Antibiotic resistance is increasingly problematic in the treatment of critically
ill patients, and colonization with antibiotic-resistant pathogens has become
endemic in many intensive care units (ICUs). Transmission of pathogens
from patient-to-patient via the hands of healthcare workers (HCWs) is a
common route of colonization. Important variables for transmission are the
degree of nursing staff cohorting, rates of contact between HCWs and
patients, adherence to hand hygiene by HCWs, and pathogen colonization
pressure [1].
Many different infection control strategies have been used to reduce
transmission of pathogens in the ICU. In nearly all cases, more than one
infection control strategy has been implemented in addition to pre-existing
(standard) infection control programs, which complicates interpretation of
the value of any specific intervention [2,3]. Active microbiological
surveillance, including reporting culture results to staff and isolating
colonized patients, has been advised as an essential measure to limit the
spread of antibiotic-resistant pathogens [4].
However, the number of interventions reported in studies of active
surveillance is usually 15 (e.g., educate staff, perform surveillance cultures,
increase the number of infection control nurses in unit, report surveillance
culture results, isolate colonized patients on the basis of results of surveillance
cultures, increase environmental cleaning, and increase adherence to hand
hygiene protocols). To assess the value of surveillance cultures (i.e., cultures
of nasal swab specimens obtained daily) alone, without use of additional
measures (such as reporting results to staff and isolating colonized patients),
we studied colonization and transmission of methicillin-susceptible
Staphylococcus aureus (MSSA) and methicillin-resistant Staphylococcus aureus
49
(MRSA) by means of unreported microbiological surveillance. Unobtrusive
observations of patient-HCW interaction were used to determine variables
influencing transmission dynamics.
Methods
Setting
Cook County Hospital is a public teaching hospital in Chicago with beds for
600 patients. The medical ICU (MICU) has 16 beds, 12 in single rooms and 4
in double rooms. The study was commissioned by the infection-control
committee and approved by the institutional review board at Cook County
Hospital. This study was conducted from September 25th through December
12th 2000.
Microbiological surveillance
Microbiological surveillance of colonization with MSSA or MRSA was
performed in the MICU for 10 weeks. All patients admitted to the MICU
were included. Nasal swab specimens (for all patients) and endotracheal
aspirates (for patients who underwent ventilation) were obtained within 12
hours after admission to the hospital and daily throughout the MICU stay. All
specimens were plated directly on mannitol salt agar (BD Diagnostic Systems).
Colonies that grew on mannitol salt agar were plated on trypticase soy agar
with 5% sheep blood (BD Diagnostic Systems) to determine the hemolytic
status of isolates and to test colonies using latex agglutination (Staphaurex;
Abbott Laboratories Diagnostics). Colonies on mannitol salt agar plates were
replicated on Oxacillin Screen Agar (BD Diagnostic Systems) to screen the
50
entire bacterial population for oxacillin-resistance. All isolates were tested for
oxacillin-resistance according to NCCLS guidelines [5]. Data on the number
of clinical isolates of MRSA recovered in the ICU were derived from the
hospitals’ microbiology laboratory. When colonization was identified ≤48 h
after ICU admission, it was considered as having been introduced into the
MICU. When colonization was identified >48 hours after ICU admission,
Pulsed-Field Gel Electrophoresis (PFGE) was used to discriminate between
endogenous colonization (due to selection of preexisting resistant flora by
antibiotic pressure) and cross-transmission, by means of SmaI-generated
restriction fragment–length polymorphism patterns.
All serial surveillance and clinical isolates were typed by PFGE. Criteria
described by Tenover et al. [6] were used to analyze results of PFGE-typing.
Data on infection control and infection rates
For seven weeks of the study period, the degree of nursing staff cohorting,
the rates of contact between patients and nurses, and nursing staff adherence
to hand hygiene were determined by observation of patients and HCWs, as
described elsewhere [7]. Experienced infection control nurses performed
unobtrusive observations daily (during the day or evening, according to a
predetermined schedule), and the MICU staff was unaware of the schedule of
observations. Nurses were observed randomly during 30-minute periods to
assess contact rates and the degree of cohorting. The degree of cohorting
expresses the likelihood that, after a first contact, the second contact will be
with the same patient. Patients were observed randomly during 15-minute
periods to assess contact rates with HCWs, and during the same interval,
these HCWs were monitored for adherence to hand hygiene. As part of the
standard infection-prevention program, infection control nurses had
51
monitored adherence to hand hygiene during an 18-month period that
overlapped that of the current study. These observations were performed
using comparable definitions but were less frequent (twice monthly) than
observations performed for study purposes, and they did not include contact
rates and degrees of cohorting. The long-term observations of adherence to
hand hygiene prescriptions were used to evaluate whether infection-control
compliance had changed over time. The numbers of patients in MICU with
MRSA isolated from clinical cultures (i.e., any other cultures performed for
clinical purposes) were obtained from infection-control records from January
1999 to January 2003. The incidence was expressed as the number of positive
MRSA cultures per three-month period. Other infection control measures
were not performed, and feedback of results was not provided to MICU staff
during the study.
Statistical analysis
Continuous variables were compared using Student’s T-test or the Mann-
Whitney U test, when appropriate. Categorical variables were studied with χ2
analysis. Potential correlations were studied using Pearson’s correlation.
Data are expressed as mean, unless otherwise values ± SD indicated.
Analyses were performed with SPSS software (SPSS Inc., Chicago, Il).
Results
Colonization
During the 10-week study period, 160 patients were admitted to the MICU. A
total of 1216 surveillance cultures were obtained from 158 patients; two
52
patients refused to participate. We were not able to collect specimens from
one patient on one day; however, specimens were obtained from this patient
on the days adjacent to this day. The daily rate of bed occupancy in the
MICU was 81%±11% (range, 56%-100%). The numbers of patients
colonized with MSSA and MRSA during this period were 55 (34.8%) and
nine (5.7%), respectively. Colonization was imported into the MICU by 62
patients (53 were colonized with MSSA, and nine were colonized with
MRSA). Two patients appeared to have acquired MSSA in the MICU, but on
the basis of PFGE results, we determined that colonization was not due to
cross-acquisition and, therefore, that acquisition of MSSA was endogenous.
PFGE of introduced MSSA and MRSA isolates revealed almost as many
different genotypes as patients from whom these strains were recovered. Few
similar genotypes were found among MSSA isolates. Because these strains
were introduced to the MICU at the time of admission and these patients did
not share an overlapping time period in the ICU, cross-transmission is
unlikely to have occurred. The daily endemic prevalence of staphylococcal
colonization was 22.8%±12.5% (range,0%-46.7%) for MSSA and MRSA,
12.2%±10.2% (range, 0%-37.5%) for MSSA only, and 10.5%±6.8% (range,
0%-25%) for MRSA only (Figure 1). Patients colonized with MRSA had a
longer length of stay in the ICU than did patients colonized with MSSA (14.4
±20.9 vs. 3.3±4.6 days; p=.006, by the Mann-Whitney U test).
53
Figure 1. Endemic prevalence of methicillin-susceptible Staphylococcus aureus (MSSA) and methicillin-resistant Staphylococcus aureus (MRSA) in the medical intensive care unit, Cook County Hospital (Chicago, IL).
0
0,05
0,1
0,15
0,2
0,25
0,3
0,35
0,4
0,45
0,5
25-9-20002-10-2000
9-10-2000
16-10-2000
23-10-2000
30-10-20006-11-2000
13-11-2000
20-11-2000
27-11-2000
Prop
ortio
n of
pat
ients
col
onize
d
MSSA MRSA
Infection control variables
Patients and HCWs were observed for a period of 361.5 hours, during which
1133 contacts between HCWs and patients were recorded. Nurses had a
mean of 1.9 patient-contacts/hour, and patients had a mean of 4.2 HCW-
contacts/hour. The mean rate of adherence to glove use was 68%, to hand
hygiene was 53%, and to glove use and/or hand hygiene was 78%. The mean
degree of cohorting of MICU nurses was 77%. Alcoholic hand rub was used
by HCWs in 15% of all hand hygiene opportunities and was highest among
physicians, compared with nurses (32% vs. 8%; p=.01, by χ2 analysis).
Historical infection rates and infection control
The number of patients who had MRSA isolated from clinical cultures per
three-month period varied from three to 10 (Figure 2).
54
Figure 2. Number of clinical isolates of methicillin-resistant Staphylococcus aureus (MRSA) and adherence to hand hygiene among healthcare workers (HCWs), medical intensive care unit, Cook County Hospital (Chicago, IL). Qtr, quarter.
0
2
4
6
8
10
12
1st Q
tr-99
2nd Q
tr-99
3rd Q
tr-99
4th Q
tr-99
1st Q
tr-00
2nd Q
tr-00
3rd Q
tr-00
4th Q
tr-00
1st Q
tr-01
2nd Q
tr-01
3rd Q
tr-01
4th Q
tr-01
1st Q
tr-02
2nd Q
tr-02
3rd Q
tr-02
4th Q
tr-02
Num
ber o
f iso
late
s
0
20
40
60
80
100
Perc
ent a
dher
ance
to
hand
hyg
iene
MRSA HAND HYGIENE
During our study period (which overlapped parts of the third and fourth
quarter of 2000), 11 patients had a total of 14 positive clinical cultures (nine
yielded MSSA, and five yielded MRSA). Five of nine patients for whom
surveillance cultures yielded MRSA had clinical cultures that yielded MRSA.
Six patients for whom surveillance cultures were positive for MSSA had
clinical cultures that were positive for MSSA. Typing of each clinical isolate
revealed a genotype identical to the patient’s surveillance isolate. There was
no discernible trend of changing incidence of colonization with MRSA or
MSSA over the three-year period. In addition, the rate of adherence to hand
hygiene, as determined by infection control nurses, varied from 32% to 48%
and did not change dramatically in the 18-month observation period.
55
Discussion
Staphylococcus aureus is a pathogen that is well adapted for patient-to-patient
spread. Staphylococcal infections are associated with considerable morbidity and
often with attributable mortality (especially when caused by methicillin-
resistant strains) [8,9]. However, whether and how infection prevention
should be performed is a matter of debate, especially when colonization with
MRSA is endemic. Some have argued that active surveillance should be
performed to identify and isolate the iceberg of colonized patients (i.e., the
majority of colonized but usually unidentified patients) [4]. Such a strategy
has been successful for 20 years in The Netherlands, where the proportion of
staphylococcal infections caused by MRSA is <1% [10]. Importantly, in such
a circumstance, patients at high risk for colonization with MRSA (e.g.,
patients who were transferred from foreign hospitals where MRSA is
endemic) can be easily identified, and introduction of MRSA from other
hospitals in The Netherlands or from the community can be neglected.
The dynamics of colonization with Staphylococcus aureus in ICUs in hospitals
where MRSA is endemic are more complicated, and the potential for active
surveillance–based isolation to be a successful infection control strategy in
such a setting is contentious. Using detailed microbiological surveillance—
without reporting results or isolating colonized patients—in a busy, urban
MICU where 6% of all patients were colonized with MRSA at admission and
the mean daily prevalence of MRSA was 10%, we found that cross-
transmission of MRSA did not occur during a ten-week study period; these
results were confirmed by genotyping. Importantly, the period of study
appeared not to be an outlier when considering the number of patients with
MRSA isolated from clinical cultures or the daily practice of HCWs regarding
adherence to hand hygiene. Therefore, the data suggest that cross-
56
transmission did not occur and that if active surveillance cultures for MRSA
would have been combined with reporting of results and isolation of
colonized patients, these would have appeared to be successful interventions,
although, in fact, they were not necessary. These findings were supported
further by the absence of cross-transmission of MSSA. We do not
recommend that all institutions perform active surveillance and bacterial
genotyping as part of their prevention strategies. But we do want to
emphasize that recommendations should be based on sound data. This study
questions the necessity of screening and isolating patients colonized with
MRSA in a high-risk environment. For an evidence-based recommendation,
however, prospective comparative trials with relevant end points should be
performed. In addition, the negative effects of isolating individuals on patient
care should be considered. In an observational study, HCWs were one-half as
likely to enter the rooms of patients in contact isolation [11], and patients
may even suffer psychologically from isolation [12]. The risks of pathogen
transmission depend on several HCW-related variables, such as contact rates,
level of cohorting, and adherence to hand hygiene measures [13]. Data from
our MICU during this study period showed 4.2 HCW-contacts/hour for
patients, a mean degree of nurse cohorting of 77%, and a mean level of
adherence to hand hygiene or gloving of 78%, which apparently were
sufficient in aggregate to prevent cross-transmission [7]. The degree of
cohorting of nurses has been determined in only a few studies. The relevance
of this measure emerged from theoretical models of pathogen transmission,
in which cohorting was expressed as the likelihood that, after a patient
contact, the next contact would be with the same patient [13]. If cohorting
was 100%, there would be no opportunity to transmit pathogens to other
patients. In our MICU, the mean degree of nurse cohorting was 77%, with
57
weekly means of 67%-90% [7]. Because ICU physicians usually care for all
patients in the unit, their level of cohorting is much lower than for nurses,
and as a result, their chance to transmit pathogens is much higher than that
for nurses [7]. Many healthcare-related variables are ward-specific and may
not be constant over time. For example, understaffing can lead to decreased
degrees of cohorting, increased contact rates, and decreased adherence to
infection control measures [14]. The possible effects of these changes were
demonstrated by Grundmann et al. [15], who reported that periods with
lower staffing levels were associated with clustered spread of MRSA in their
ICU. Several other studies have also identified understaffing as a risk factor
for MRSA infection [2], as well as for catheter-related bloodstream infections
[16] and prolonged duration of ICU stay [17,18]. Improved adherence to
infection control measures, fixing staff deficits, or identification and isolation
of carriers, theoretically, could have prevented the spread of MRSA in
periods of understaffing. Because isolation procedures usually increase the
workload for HCWs, it is uncertain whether such a strategy can be
implemented without providing additional staff, especially when a problem
emerges because of understaffing. In fact, the greatest benefit of the multiple
interventions included in programs of active microbiological surveillance and
isolation may derive from the allocation of extra staff required by increased
numbers of patients for whom there are contact precautions and/or from
reduced entry of HCWs into isolation rooms [11]. Our results need to be
interpreted in light of study limitations. Control of MRSA may reflect
improved HCW-adherence, because of the presence of an individual who
obtained cultures, although hand hygiene adherence did not appear to change
markedly (Figure 2). Infection control effects for patients colonized with
MRSA may have been better by chance, but our data do not indicate this.
58
Lack of MRSA and MSSA transmission may also have been chance
phenomena, although this seems unlikely for a period of three months.
Finally, most (12 of 16) of the beds in the MICU were single rooms, which
may not apply to other settings. The emerging picture of these studies is that
general recommendations for targeted infection control measures, such as
performance of active surveillance cultures and subsequent isolation of
colonized patients, require a greater understanding of the epidemiology of
nosocomial pathogens in general and of hospital factors in particular, such as
relative importance of acquisition routes (endogenous or cross-transmission),
colonization pressure, cohorting, adherence to hand hygiene, and staffing
levels. These factors are rarely assessed in studies of infection control
interventions but should greatly influence the choice of infection control
measures.
59
References
1. Bonten MJ, Austin DJ, Lipsitch M. Understanding the spread of antibiotic resistant pathogens in hospitals: mathematical models as tools for control. Clin Infect Dis 2001:1739-46.
2. Haley RW, Cushion NB, Tenover FC et al. Eradication of endemic methicillin-resistant Staphylococcus aureus infections from a neonatal intensive care unit. J Infect Dis 1995:614-24.
3. Montecalvo MA, Jarvis WR, Uman J et al. Infection control measures reduce transmission of vancomycin-resistant enterococci in an endemic setting. Ann Intern Med 1999:269-72.
4. Muto CA, Jernigan JA, Ostrowsky BE et al. SHEA guideline for preventing nosocomial transmission of multidrug-resistant strains of Staphylococcus aureus and enterococcus. Infect Control Hosp Epidemiol 2003:362-86.
5. Nijssen S, Bonten M, Franklin C et al. The Relative Risk of Physicians and Nurses to Transmit Pathogens in a Medical Intensive Care Unit. Arch Intern Med 2003:2785-6.
6. National Comittee for Clinical Laboratory Standards 1999. Methods for dilution antimicrobial susceptibility tests for bacteria that grow aerobically. Approved Standard M7-A4. National Comittee for Clinical Laboratory Standards; Wayne, Pa.
7. Blot SI, Vandewoude KH, Hoste EA et al. Outcome and attributable mortality in critically Ill patients with bacteremia involving methicillin-susceptible and methicillin-resistant Staphylococcus aureus. Arch Intern Med 2002:2229-35.
8. Engemann JJ, Carmeli Y, Cosgrove SE et al. Adverse clinical and economic outcomes attributable to methicillin-resistance among patients with Staphylococcus aureus surgical site infection. Clin Infect Dis 2003:592-8.
9. Vandenbroucke-Grauls CM. Methicillin-resistant Staphylococcus aureus control in hospitals: the Dutch experience. Infect Control Hosp Epidemiol 1996:512-3.
60
10. Kirkland KB, Weinstein JM. Adverse effects of contact isolation. Lancet 1999:1177-8.
11. Peel RK, Stolarek I, Elder AT. Is it time to stop searching for MRSA? Isolating patients with MRSA can have long term implications. BMJ 1997:58.
12. Austin DJ, Bonten MJ, Weinstein RA et al. Vancomycin-resistant enterococci in intensive-care hospital settings: transmission dynamics, persistence, and the impact of infection control programs. Proc Natl Acad Sci USA 1999:6908-13.
13. Pittet D, Mourouga P, Perneger TV. Compliance with handwashing in a teaching hospital. Infection Control Program. Ann Intern Med 1999:126-30.
14. Grundmann H, Hori S, Winter B et al. Risk factors for the transmission of methicillin-resistant Staphylococcus aureus in an adult intensive care unit: fitting a model to the data. J Infect Dis 2002:481-8.
15. Fridkin SK, Pear SM, Williamson TH et al. The role of understaffing in central venous catheter-associated bloodstream infections. Infect Control Hosp Epidemiol 1996:150-8.
16. Needleman J, Buerhaus P, Mattke S et al. Nurse-staffing levels and the quality of care in hospitals. N Engl J Med 2002:1715-22.
17. Thorens JB, Kaelin RM, Jolliet P et al. Influence of the quality of nursing on the duration of weaning from mechanical ventilation in patients with chronic obstructive pulmonary disease. Crit Care Med 1995:1807-15.
18. Tenover FC, Arbeit RD, Goering RV et al. Interpreting chromosomal DNA restriction patterns produced by pulsed-field gel electrophoresis: criteria for bacterial strain typing. J Clin Microbiol 1995:2233-39.
61
Chapter
The relative risk of physicians and nurses to transmit pathogens in a medical intensive care unit
Nijssen S, Bonten M, Franklin C, Verhoef J, Hoepelman A, Weinstein R Archives of Internal Medicine 2003; 163:2785-6
Introduction
Transmission of pathogens from patient-to-patient by the hands of healthcare
workers (HCWs) is the most important source of cross-infections in hospitals,
especially in intensive care units (ICUs). Three variables related to behavior
of HCWs are important in cross-transmission of pathogens: adherence to
hand hygiene, the extent of HCW cohorting, and the number of interactions
between HCWs and patients (interaction rates) [1]. Although levels of
adherence with hand hygiene have been studied extensively [2], little is known
about the extent of cohorting of HCWs, the number of interactions per hour
between HCWs and patients, and how these parameters quantitatively
influence the potential relative risk of transmission of microorganisms for
different groups of HCWs.
Methods
During a seven-week period, the cohorting extent of nursing staff, the
number of interactions between patients and HCWs and HCWs’ adherence
with hand hygiene were assessed by observation of patients and HCWs in a
16-bed medical ICU.
Experienced infection control nurses performed unobtrusive observations
(according to a predetermined schedule during days and evenings), and the
medical ICU (MICU) staff was unaware of the schedule of observations.
Nurses were observed randomly during 30-minute periods to assess contact
rates and the extent of cohorting. The cohorting extent expresses the
likelihood that after a first contact, the second contact will be with the same
patient. If every contact of a specific HCW is with the same patient, the level
of cohorting is 100% (HCW-patient ratio = 1) and the risk for cross-
64
transmission would be zero. Patients were observed randomly during 15-
minute periods to assess contact rates and HCWs’ adherence to hand hygiene.
Nursing workload and staffing levels were measured. The staffing level was
defined as the nurse-patient ratio and was based on actual number of nurses
and patients present in the unit each day. Workload measurement was
expressed by Medicus®Workload Measurement Methodologies (QuadraMed
Corporation, Reston, Va), which is a patient classification tool for the
assessment of nursing care needs based on the physical condition of a patient.
Correlations between nurse staffing levels and interactions rates of nurses and
between nursing workload and nurses’ adherence to hand hygiene were
assessed. Statistical analyses were performed with SPSS statistical software
(SPSS Inc., Chicago, Il). Continuous variables were compared using the
Student’s T-test or Mann-Whitney U test when appropriate. Categorical
variables were studied with the χ2 test. Potential correlations were studied
using Pearson’s correlation.
Results
Patients were observed for 170.5 hours, during which 777 HCW-patient
interactions were recorded; mean number of HCW-patient interactions per
hour (interaction rate) was 4.2. Physicians were responsible for 28% (1.2 per
hour) and nurses for 61% (2.6 per hour) of these interactions. Adherence to
hand hygiene after interaction with a patient was 43% for physicians and 59%
for nurses (p<.001). Nurse staffing levels were inversely associated with nurse
interaction rates (correlation coefficient, -0.30; p=.08) and nursing workload
was inversely associated with nurses’ adherence to hand hygiene (correlation
coefficient, -0.38, p=.02). Nurses were observed for 191 hours during which
65
the cohorting extent was 77%. Since physicians are not cohorted to individual
patients, their behavior can be expressed as random mixing, and their
cohorting extent can be calculated by 1/average number of patients (1/12
≈.08). Based on interaction rates, the total number of nurses and physicians
present, the extent of cohorting of HCWs, and hand hygiene adherence rates,
we estimated the number of interactions by HCWs with different patients
without using hand hygiene per hour and thus the relative risk to transmit
pathogens for nurses and physicians (Table 1). Physicians and nurses (as a
group) had 3.8 and 2.4 interactions per hour, respectively, without using hand
hygiene. Thus, the relative risk to have such a contact was 1.6 for physicians
compared to nurses.
Table 1. Estimate of the number of Sequential Patient Interactions Without Using Hand Hygiene (Spi*) for Physicians and Nurses per Hour
Nurses
Interaction rate (per hour)† 0.61 x 4.2 = 2.6
No. of nurses 9.6
Cohorting‡ 0.77
Adherence to hand hygiene 0.59
No of sequential interactions without hand hygiene per hour 2.4
Physicians
Interaction rate (per hour)† 0.28 x 4.2 = 1.2
No. of nurses 6
Cohorting§ 0.08
Adherence to hand hygiene 0.43
No of sequential interactions without hand hygiene per hour 3.8
*Spi = Interaction x number of healthcare workers x (1-adherence). †Interaction rates were estimated according to the number of times patients were contacted per hour (4.2) and the observed portion of interactions for each type of healthcare worker. The portion of interactions due to physicians was 28%, and the portion of interactions due to nurses was 61%. ‡Average staffing level x average number of patients: 0.8 x 12 patients = 9.6 §Cohorting level of physicians was estimated as if all contacts were randomly mixed. The average number of patients was 12 during this study, and therefore the cohorting level was estimated to be 1/12.
66
Discussion
This study quantitates the relative impact of different groups of HCWs on
pathogen transmission dynamics. Although physicians have less patient
contacts when compared to nurses, physicians’ interactions are less cohorted
(i.e., they take care of more patients) and physicians usually have lower
adherence to hand disinfection. For nurses, greater cohorting reduces mixing
of contacts with different patients and therefore lowers the risk of cross-
transmission. At various times during our observations, understaffing reduced
the nurse-patient ratio and thus the extent of cohorting. Moreover,
understaffing was associated with increased interaction rates of nurses with
patients, and the increased workload was associated with intercurrent
decreases in hand hygiene adherence of nurses, which support the findings of
others [3,4]. Nevertheless, for the overall study period, the chance of having a
potentially contaminated contact was 1.6 times higher for physicians than for
nurses. To compensate for the lack of cohorting, physicians in our ICU
would need to improve their adherence to hand hygiene to 64% or care for
fewer patients to match the nurses’ lower risk of cross-transmission. It should
be noted, however, that we did not evaluate the extent of hand contamination
after patient care, and we assumed that all patient interactions were equal.
Future studies are needed to determine heterogeneity of patient interactions
with regard to hand contamination, which might alter our calculated relative
risks of potentially contaminated contacts. The relative short period might be
considered as a potential limitation of our findings. However, each nurse and
a considerable number of physicians were observed within this period.
Nevertheless, temporary changes in adherence to infection control measures
due to understaffing or increase workload may have been missed. There are
two major implications of our findings. First, infection control policies
67
usually have been focused on improving adherence of nursing staff, but our
data emphasize the need aggressively educate and include physicians in
infection control programs. Second, understaffing will affect almost all
relevant variables of transmission dynamics. Multiple studies have identified
reduced nurse-patient ratios as a risk factor for transmission of nosocomial
pathogens and even patient outcome. Grundmann et al. [4], found that
exposure to relative staff deficit was the only variable significantly associated
with clustered cases of methicillin-resistant Staphylococcus aureus colonization.
Fridkin et al. [5], identified a high patient-nurse ratio as an independent risk
factor for central venous catheter-associated bloodstream infections
occurring in a surgical ICU, and therefore understaffing can be considered as
a potential risk factor for nosocomial infections. Needleman et al. [6], found a
positive association between the proportion of total hours of nursing care by
registered nurses per day and six outcomes in medical patients (i.e., length of
stay, rates of urinary tract infections, upper gastrointestinal tract bleeding,
hospital-acquired pneumonia, and shock or cardiac arrest).
68
References
1. Austin DJ, Bonten MJM, Weinstein RA et al. Vancomycin-resistant enterococci in intensive care hospital settings: Transmission dynamics, persistence, and the impact of infection control programs. Proc Natl Acad Sci USA 1999:6908-13.
2. Pittet D, Boyce JM. Hand hygiene and patient care: pursuing the Semmelweis legacy. Lancet Infect Dis 2001:9-20.
3. Pittet D, Mourouga P, Perneger TV et al. Compliance with hand washing in a teaching hospital. Ann Intern Med 1999:126-30.
4. Grundmann H, Hori S, Winter B et al. Risk factors for the transmission of methicillin-resistant Staphylococcus aureus in an adult intensive care unit: fitting a model to the data. J Infect Dis 2002:481-8.
5. Fridkin SK, Pear SM, Williamson TH et al. The role of understaffing in central venous catheter-associated bloodstream infections. Infect Control Hosp Epidemiol 1996:150-8.
6. Needleman J, Buerhaus P, Mattke S et al. Nurse-staffing levels and the quality of care in hospitals. N Engl J Med 2002:1715-22.
69
Chapter
Unnoticed spread of integron-carrying Enterobacteriacea in intensive care units
Nijssen S, Florijn A, Top J, Willems R, Fluit A, Bonten M Clinical Infectious Diseases 2005; 41:1-9
Abstract
Integrons are strongly associated with multi-drug resistance in Entero-
bacteriaceae. Little is known about the natural history of integron-associated
resistance in hospitals during non-outbreak periods. The prevalence of
integrons and the incidence of cross-transmission and horizontal gene
transfer in Enterobacteriaceae with reduced susceptibility to cephalosporins
(ERSC) were determined for two intensive care units (ICUs).
Microbiological surveillance using rectal swab samples obtained two times per
week and genotyping using Amplified Fragment-Length Polymorphism
(AFLP) were used to determine colonization with and genetic relatedness of
ERSC. IntI1 integrase polymerase chain reaction (PCR), conserved-segment
PCR, Restriction Fragment–Length Polymorphism (RFLP), and DNA
sequencing were used to determine the prevalence and contents of integrons.
Of 457 patients, 121 patients were colonized with ERSC, and 174 isolates
underwent AFLP and PCR. In 34 isolates obtained from 31 patients, 11
different integrons were identified; these integrons encoded resistance to
streptomycin/spectinomycin, gentamicin/tobramycin/kanamycin, chloram-
phenicol and trimethoprim. Integrons could be divided into seven clusters of
≥2 isolates each. Compared with isolates that were negative for integrons,
isolates that were positive for integrons were associated with resistance to
piperacillin, cephalosporins, aminoglycosides, and quinolones. Acquisition
rates of integron-carrying ERSC were 10 cases per 1000 patient-days in the
first ICU and eight cases per 1000 patient-days in the second ICU, with most
cases (26 of 34) being acquired during the ICU stay. Nineteen episodes
resulted from cross-transmission. In addition, two cases of interspecies
transfer and one case of intraspecies transfer of integrons were recorded.
72
Younger age was independently associated with acquisition of integron-
carrying ERSC (Hazard ratio (HR), 0.953; 95% Confidence interval (CI),
0.926-0.987). Surveillance, genotyping, and integron analysis identified
previously unnoticed outbreaks of integron-carrying ERSC. Cross-
transmission appeared to be the dominant route of transmission. Therefore,
barrier precautions are necessary to prevent further spread.
Introduction
Treatment of nosocomial infections is hampered by the worldwide increase
of antibiotic resistance, especially in intensive care units (ICUs) [1].
Numerous studies have described and analyzed nosocomial outbreaks of
infections caused by multi-drug resistant pathogens. Yet, relatively little is
known about the epidemiology of antibiotic-resistant pathogens in non-
outbreak settings, and knowledge of the natural history and spread of
resistance mechanisms will be essential to attempts to reverse the increase of
antibiotic resistance.
Changes in the prevalence of colonization with resistant microorganisms
within a hospital can occur through admission of colonized patients, through
endogenous selection of pre-existent drug-resistant flora during antimicrobial
therapy, through specific mutations in the genome of susceptible bacteria,
through cross-transmission (i.e., through the spread of microorganisms from
one patient to another, usually via the hands of healthcare workers), or
through contamination originating from an environmental source [2]. In
addition, microorganisms can acquire resistance determinants through
horizontal gene transfer. The major agents of horizontal gene transfer in
Enterobacteriaceae include conjugative plasmids and transposons with
73
resistance genes. Multi-drug resistance in Enterobacteriaceae is strongly
associated with integrons [3], which are located either on plasmids or on the
bacterial chromosome and may form part of transposons. Three classes of
integrons are specifically involved in antibiotic resistance. The majority of
integrons found in human clinical isolates belong to class 1. The gene
cassettes confer resistance to antimicrobial agents, antiseptics, and
disinfectants [4,5]. Emergence of integron-associated antibiotic resistance
may occur through cross-transmission of integron-carrying microorganisms
or through horizontal gene transfer of plasmids and transposons [6-9]. As is
the case for most other antibiotic-resistant pathogens, the spread of integron-
carrying, multi-drug resistant Enterobacteriaceae has usually been detected
during outbreaks of nosocomial infection [6,7] and there is little knowledge
about the natural history of integron-associated resistance in hospitals during
non-outbreak periods. Therefore, we determined the prevalence of integrons
as additional resistance determinants in isolates of Enterobacteriaceae with
reduced susceptibility to cephalosporins (ERSC) and the incidences of
horizontal gene transfer and cross-transmission of resistant microorganisms
in two ICUs.
Methods
Microbiological surveillance to determine the prevalence and incidence of
rectal colonization with ERSC was performed during an eight-month period
(September 2001 through May 2002) in two ICUs (ICU-1 and ICU-2) of a
900-bed teaching hospital in The Netherlands (University Medical Center
Utrecht). Rectal swab samples were obtained at admission to the ICU and
twice weekly thereafter, and demographic and clinical data were monitored.
74
The NCCLS recommends using 2 μg/ml of cefpodoxime as a first screening
for drug resistance in Enterobacteriaceae when using pure cultures. We could
not reproduce these results with direct plating of rectal swab samples on
chromogenic agar plates (Chromogenic UTI; Oxoid) with 2 μg/ml of
cefpodoxime, and we performed a study to determine the effective
concentration of cefpodoxime in screening for ERSC. Enterobacteriaceae
isolates with known antimicrobial susceptibility patterns (determined by
microdilution, performed according to NCCLS guidelines) were used to spike
fecal samples and then plated by swab on chromogenic agar plates with
various concentrations of cefpodoxime (either 1, 2, 4, 8, or 16 μg/ml).The
effective concentration of cefpodoxime for screening was determined to be
8 μg/ml. Swab samples were plated directly on chromogenic agar plates
supplemented with 8 μg/ml of cefpodoxime and 6 μg/ml of vancomycin.
Vancomycin was added to the medium to inhibit growth of gram-positive
flora (which we were not interested in).
One colony of each morphological variant was selected and stored. Species
identification was performed using the Vitek IIsystem (bioMérieux).
Susceptibility testing was performed by means of microdilution, according to
NCCLS guidelines [10]. Extended-spectrum β-lactamase (ESBL) production
was detected by means of the disk diffusion test and ESBL E-test [11]. Two
isolates per species per patient (if available) were selected for detection and
characterization of integrons. If 12 isolates were available for a species, the
first and last isolated were used for analyses.
75
Detection and characterization of integrons
Integrons were detected by PCR amplification of the class 1 integrase-specific
IntI gene [12]. The cassette content of the integrons was characterized by
performing conserved-segment PCR (CSPCR) [8], and restriction fragment–
length polymorphism (RFLP) analysis of amplification products of the same
size was performed to demonstrate similarity of integron contents [6]. RFLP
patterns of CS-PCR products were compared with known RFLP patterns
recorded in a database, and CS-PCR products with a unique RFLP pattern
underwent sequencing to identify inserted gene cassettes. Sequencing of gene
cassette contents was performed as described by Peters et al. [13].
Amplified Fragment-Length Polymorphism (AFLP) genotyping. AFLP was
used to demonstrate the genetic relatedness of isolates and was performed as
described by Willems et al. [14], with one modification: the EcoRI-CfoI
adapter was substituted for with the EcoRI-MseI adapter.
Definitions
Cross-transmission was defined as acquired colonization with genotypically
related strains in epidemiologically linked patients. Colonization with ERSC
148 hour after admission to the ICU in a patient with a previous negative
culture result was considered to have been acquired in the ICU. Genetic
relatedness was determined on the basis of both visual and computerized
interpretation of AFLP patterns. Epidemiological linkage was defined as two
patients having an overlap in stay in the ICU. Because of the possibility of
low-level colonization immediately after acquisition, a maximum time
window (between both periods) of seven days was accepted. Intraspecies
transfer of an integron was defined as two genotypically unrelated isolates of
76
the same species carrying the same integron. Interspecies transfer of an
integron was defined as two different species carrying the same integron.
Statistical analysis
Risk factors for acquired colonization with integron-carrying isolates,
compared with acquired colonization with isolates not carrying an integron,
were analyzed by univariate and multivariate Cox’ regression analysis. Cox’
regression analysis controls for the time at risk of individual patients. Patients
colonized with an integron-carrying organism at admission to the ICU were
excluded from analyses.
Results
Patients and colonization
Four hundred fifty-seven patients were admitted to the ICU during an eight-
month period (277 patients to ICU-1 and 180 to ICU-2) (Table 1). In total,
1243 rectal swab samples were obtained, and rectal colonization with ERSC
was found in 121 patients (70 in ICU-1 and 51 in ICU-2). Sixty-one patients
were found to be colonized at admission to the ICU (38 in ICU-1 and 23 in
ICU-2), and 56 patients acquired colonization in the ICU (31 in ICU-1 and 25
in ICU-2). In four patients, it was unclear whether colonization was present
at admission or acquired in the ICU, either because the first culture was
obtained 148 hour after admission or because the patient was admitted to the
ICU before the start of the study. From these 121 patients, 174 isolates were
selected for detection and characterization of integrons, including 67
77
Enterobacter species (38.5%), 40 Citrobacter species (23%), 37 Escherichia species
(21.3%), 28 Klebsiella species (16.1%), and two Serratia species (1.1%).
Table 1. Demographic and laboratory characteristics of patients at two intensive care units (ICUs) with clusters of integron-carrying Enterobacteriaceae.
Variable
Medical ICU
(n = 277)
Neurosurgical ICU
(n = 180)
Both ICUs
(n = 457)
Age, means years ± SD 52 ± 18 54 ± 18 54 ± 18
Male sex 144 (52) 100 (56) 244 (54)
APACHE II score, mean ± SD 21 ± 8 22 ± 18 21 ± 8
Length of ICU stay, means days ± SD 8 ± 11 8 ± 11 8 ± 11
ICU mortality 57 (21) 26 (15) 83 (18)
Colonization with ERSC 70 (25) 51 (28) 121 (26)
Colonization with integron-carrying ERSC 19 (7) 12 (7) 31 (7)
Colonization with integron-carrying ERSC
acquired in the ICU 14 (5) 9 (5) 23 (5)
Time from admission to ICU to acquisition of
integron-carrying ERSC, means days ± SD 10 ± 10 12 ± 10 11 ± 10
Daily endemic prevalence of integron-carrying
ERSC, mean prevalence (range) 8 (0 - 33) 5 (0 - 33) 7 (0 - 33)
Acquisition rate, no. of acquisitions per 1000
patient-days 10 8 9
NOTE. Date are no. (%) of patients, unless otherwise specified. ERSC, Enterobacteriaceae with reduced
susceptibility to cephalosporins.
Presence of integrons
IntI1 integrase PCR identified 54 integron-carrying isolates obtained from 31
patients colonized with ERSC; these included 24 Klebsiella species (43.6%), 12
Enterobacter species (22.2%), 15 Escherichia species (28.3%), and three
Citrobacter species (5.7%). Three patients were colonized with two different
integron-carrying isolates, five patients with 3, two patients with 4, and one
patient with 5. In ICU-1, 19 patients (7%) were colonized with an integron-
carrying ERSC; 14 (74%) of these patients acquired their isolate(s) in the ICU.
78
In ICU-2, 12 patients (7%) were colonized with an integron-carrying ERSC;
nine (75%) of these patients acquired their isolate(s) in the ICU. The mean
daily endemic prevalence (i.e., the percentage of all patients present in the
ICU who were colonized with an integron-carrying isolate) was 8% (range,
0%-33%) in ICU-1 and 5% (range, 0%-33%) in ICU-2, and acquisition rates
were 10 cases and eight cases per 1000 patient-days at risk in ICU-1 and ICU-
2, respectively. ERSC that carried integrons were more frequently resistant to
piperacillin and ticarcillin, aminoglycosides, and quinolones than were ERSC
without integrons (Table 2).
Table 2. Resistance profiles of Enterobacteriaceae isolates with reduced susceptibility to cephalosporins, by integron positivity.
Antimicrobial drug(s), by class
Percentage of integron-
negative isolates with
drug resistance
(n = 120)
Percentage of integron-
Negative isolates with
drug resistance
(n = 54) P
Penicillin
Ampicillin 85 93 .16
Piperacillin 24 94 <.01
Ticarcillin 33 94 <.01
Amoxicillin-clavulanic acid 84 79 .47
Piperacillin-tazobactam 9 13 .36
Cephalosporin
Cefpodoxime 45 78 <.01
Ceftazidime 26 33 .31
Cefotaxime 29 44 .05
Cefoxitin 78 41 <.01
Aminoglycoside
Gentamicin 2 94 <.01
Quinolone
Ciprofloxacin 3 33 <.01
Carbapenem
Meropenem 0 0
NOTE. Date are no. (%) of patients, unless otherwise specified. ERSC, Enterobacteriaceae with reduced susceptibility
to cephalosporins.
79
Risk factors for acquisition of integrons
Time at risk (i.e., time until discharge or death for patients who were not
colonized with integron-carrying ERSC and time until colonization for
patients who were) was shorter for uncolonized patients than for patients
who acquired colonization with integron-carrying ERSC; mean times at risk
(±SD) were 8±5 days and 12±13 days, respectively (p<.01, by Mann-Whitney
U test). In univariate-Cox regression analysis, duration of hospital stay before
admission to the ICU (HR, 1.042; 95% CI, 1.007-1.082) and younger age (HR,
0.956; 95% CI, 0.926-0.987) were associated with a higher risk for acquisition
of an integron-carrying isolate. Exposure to β-lactam antibiotics seemed to
confer protection against acquisition (HR, 0.347; 95% CI, 0.138-0.873). In
multivariate analysis, only a younger age remained independently associated
with the acquisition of integron-carrying isolates (HR, 0.953; 95% CI, 0.926-
0.987).
Typing of integrons
CS-PCR of the 54 isolates carrying the IntI1 gene yielded conserved-segment
amplification products for 39 isolates. Analyses of the amplification products
by RFLP revealed 11 different integrons (Table 3). Four of these integrons
had been identified previously in our hospital (types I, II, VII, and VIII), and
sequencing of two other conserved-segment amplification products revealed
two new integron types (types XV and XVI). Sequencing of the conserved-
segment amplification products revealed the presence of a variant of the dfrV
gene cassette, encoding resistance to trimethoprim, in integron XV and the
presence of the aadB gene cassette, encoding resistance to gentamicin/
tobramycin/kanamycin, in integron XVI. For the remaining five conserved-
segment amplification products (discriminated on the basis of RFLP patterns),
80
gene cassettes could not be sequenced because of insufficient material; they
were designated as integron types A-E.
Table 3. Overview of characteristics of integrons from 2 intensive care units with clustes of integron-carrying Enterobacteriaceae.
Cluster, microorganism
(no. of isolates)
Conserved-segment
products, bp
Integron
type
Gene
cassette(s)
1: Escherichia coli (9) 1000 and 1400 I and II aadA2 and aadB-catB3
1: Enterobacter agglomerans (1) 1400 II aadB-catB3
2: Escherichia coli (4) 1600 VII dfria-aadA1a
3: Enterobacter cloacae (7) 800 XVI aadB
3: Klebsiella pneumoniae (1) 800 XVI
3: Klebsiella oxytoca (1) 800 XVI
4: Klebsiella pneumoniae (2) 2400 A Unknown
5: Klebsiella oxytoca (12) 2000 Ba Unknown
6: Klebsiella oxyoca (2) 2000 Ca Unknown
7: Klebsiella pneumoniae (3) 2400 and 800 D and E Unknown and unknown
8b: Klebsiella pneumoniae (1) 800 XV dfrV
9b: Citrobacter freundii (1) 1800 VIII dfrXII-aadA2 a Difference between integron types B and C, determined on the basis of Restriction Fragment-
Length Polymorphism patterns using Hpall restriction enzyme. b Not a real cluster, sporadic isolates.
Epidemiological clusters
On the basis of the presence of different integron types, 32 representative
isolates (chosen from a total of 39 isolates) could be divided into seven
different clusters of ≥2 isolates each (Figures 3-6 and Table 3). Two integrons
(types VIII and XV) were found in only one isolate each. Prevalence of
different integron types over time is shown in Figure 1 (for ICU-1) and
Figure 2 (for ICU-2).
81
Figure 1. Number of patients in intensive care unit 1 with integron-carrying isolates of Enterobacteriaceae with reduced susceptibility to cephalosporins, by study week.
0
1
2
3
4
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35
Study week
N p
atie
nts w
ith in
tegr
on-c
aryi
ng is
olat
es
Integron types I and II Integron type VII Integron type XV Integron type XVI Integron type A
Figure 2. Number of patients in intensive care unit 2 with integron-carrying isolates of Enterobacteriaceae with reduced susceptibility to cephalosporins, by study week.
0
1
2
3
4
5
6
7
8
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35
Study week
N p
atie
nts w
ith in
tegr
on-c
aryi
ng is
olat
es
Integron type VIII Integron type XVI Integron type B Integron type C Integron type D
82
The first cluster comprised nine patients colonized with Escherichia coli isolates
carrying integrons of types I and II (Figure 3).
Figure 3. Escherichia coli isolates belonging to integron clusters 1 (integron types I and II) and 2 (integron tyepe VII).
Integron type I contained the aadA2 gene cassette, encoding resistance to
streptomycin/spectinomycin. Integron type II contained the aadB and catB3
gene cassettes, encoding resistance to gentamicin/tobramycin/kanamycin and
chloramphenicol, respectively. All Escherichia coli containing integrons of types
I and II were ESBL producers and were isolated exclusively from patients
admitted to ICU-1. Moreover, colonization was acquired during ICU-stay in
all cases, and eight of nine isolates were genotypically related according to
AFLP analysis. Therefore, it is very likely that at least eight cases of acquired
colonization resulted from cross-transmission, although epidemiological
linkage was evident in only four cases. Evidence of intraspecies transfer of
integron types I and II was present in one case. The presumed acceptor strain,
Escherichia coli isolate 1.1, was not genotypically related to the isolates obtained
from other patients; however, this patient was epidemiologically linked to two
of the patients (those from whom isolates 1.2 and 1.3 were isolated).
83
Circumstantial evidence of interspecies transfer was present in a patient
colonized with a strain of Enterobacter agglomerans that carried the same
integron type as that found in a strain of Escherichia coli isolated from an
epidemiologically linked patient.
The second cluster comprised four patients who were colonized with four
genotypically related Escherichia coli strains that carried integron type VII
(isolates 2.1, 2.2, 2.3, and 2.4) (Figure 3). This integron contained the dfrIa and
aadA1a gene cassettes, encoding resistance to trimethoprim and
streptomycin/ spectinomycin, respectively. All four patients were already
colonized with these strains at admission to the ICU and had an extensive
medical history. All had been frequently admitted to either our hospital or
other hospitals in previous years. One of these patients had been hospitalized
for 20 days before admission to the ICU. Integron type VII was previously
found in Escherichia coli and Citrobacter freundii isolates obtained from patients
admitted to various wards of our hospital. The third cluster comprised eight
patients colonized with nine isolates carrying integron type XVI, which
contained the aadB gene cassette (Figure 4).
Figure 4. Enterobacter cloacae isolates belonging to integron cluster 3 (integron type XVI).
Integron XVI was present in three different species: in seven Enterobacter
cloacae isolates (obtained from seven different patients), in one Klebsiella oxytoca
84
isolate, and in one Klebsiella pneumoniae isolate. Six of seven patients who were
colonized with Enterobacter cloacae were admitted to ICU-2, and one was
admitted to ICU-1. Five of these patients acquired colonization during their
stay in the ICU. All five isolates were genetically related according to AFLP
analysis. Therefore, it is likely that at least five cases resulted from cross-
transmission, although epidemiological linkage was evident in only two cases.
The seventh patient (from whom isolate 3.6 was obtained) was admitted to
ICU-1 and was already colonized at admission. On the basis of integron type
and species identification, there was evidence of one case of interspecies
transfer of integron XVI in a patient colonized with Enterobacter cloacae (isolate
3.2) who subsequently acquired colonization with a strain of Klebsiella oxytoca
that also carried integron XVI. The fourth cluster comprised two patients
who were admitted to ICU-1. Both patients acquired colonization with
Klebsiella pneumoniae. The isolates (4.1 and 4.2) were found to have a high
similarity in AFLP patterns, and both carried integron A (Figure 5).
Figure 5. Klebsiella pneumoniae isolates belonging to integron clusters 4 (integron type A) and 7 (integron types D and E).
There was epidemiological linkage between both patients. Clusters 5 and 6
each comprised two patients who were admitted to the same ICU (ICU-2). In
each cluster, patients were epidemiologically linked and acquired colonization
with genetically identical bacteria, which suggests evidence of at least 1 case
of cross-transmission in each cluster (Figure 6).
85
Figure 6. Klebsiella oxytoca isolates belonging to integron clusters 5 (integron type B) and 6 (integron type C).
Moreover, on the basis of AFLP patterns, Klebsiella oxytoca isolates associated
with clusters 5 and 6 were highly related (i.e., had an AFLP similarity of
185%), but restriction with HpaII suggested the presence of different
integron types in genetically related bacteria (Table 3). One patient admitted
to ICU-2 was colonized simultaneously with Enterobacter cloacae carrying
integron type XVI (isolate 3.1) and two genetically related strains of Klebsiella
oxytoca, one carrying integron type B (isolate 5.2) and the other carrying
integron type C (isolate 6.1). On the basis of genotyping and epidemiological
linkage, this patient might have served as a source of colonization for a
patient who acquired colonization with an Enterobacter cloacae strain carrying
integron type XVI (isolate 3.2) and a Klebsiella oxytoca strain carrying integron
type B (isolate 5.1) and for another patient who acquired colonization with a
Klebsiella oxytoca strain carrying integron type C (isolate 6.2).
The seventh cluster comprised three Klebsiella pneumoniae isolates from three
patients admitted to ICU-2, carrying integrons D and E, of which the
contents could not be sequenced (Figure 5). These isolates were all acquired
in the ICU and based on genotyping results and epidemiological linkage, one
case of cross-transmission was identified.
86
In summary, the presence of integrons could be determined in 34
representative isolates obtained from 31 patients (Table 4).
Table 4. Summary of data related to seven integron clusters in 2 intensive care units (ICUs) and the relative importance of different modes of horizontal gene transfer.
No. of isolates
Cluster Overall
Of strains
acquired
during ICU
stay
With cross-
transmission,
determined on
the basis of
genotyping
With cross-
transmission,
determined on
the basis of
epidemiological
linkage and
genotyping
With interspecies
transfer,
determined on the
basis of
epidemiological
linkage and
genotyping
With intraspecies
transfer,
determined on
the basis of
integron type
1 10a 10 8 4 1b 1
2 4 0 0 0 0 0
3 9c 6 5 2 1 0
4 2 2 2 1 0 0
5 2 2 1 1 0 0
6 2 2 1 1 0 0
7 3 3 2 1 0 0
Total 32 25 19 10 2 1 a Includes 9 isolates of Escherichia coli carrying integron types I and II and 1 islolate of Enterobacter agglomerans
carrying integron type II. b Circumstancial evidence of horizontal gene transfer between epidemiologically linked patients of integron type
II only. Evidence of cross-transmission of Escherichia coli is lacking; however, cross-transmission of Enterobacter
cloacae between these patients was confirmed. c Includes 7 isolates of Enterobacter cloacae, 1 isolate of Klebsiella pneumoniae, and 1 isolate of Klebsiella oxytoca.
In the majority of cases (involving 26 [76%] of 34 isolates), colonization with
these bacteria became apparent during stay in the ICU. On the basis of a
comparison of AFLP and integron analysis, we determined that 19 episodes
of acquired colonization resulted from cross-transmission, although
epidemiological linkage of donor and acceptor patient was only evident in 10
cases. In addition, two cases of interspecies and one case of intraspecies
87
transfer of integrons were found. Therefore, of 26 cases of acquired
colonization with integron-carrying gramnegative bacteria, between 10 (38%)
and 19 (73%) of the cases may have resulted from cross-transmission, and
three (12%) may have resulted from horizontal gene transfer (two cases of
interspecies and one case of intraspecies transfer).
Discussion
The natural history of multi-drug resistant, integron-carrying Entero-
bacteriaceae in two ICUs with low levels of resistance to antibiotics was
characterized by its unrecognized spread and the dominance of cross-
transmission, rather than horizontal gene transfer, as the transmission route
responsible for that spread. Most integron-containing Enterobacteriaceae
were acquired after admission to the ICU, with unit-specific clustering that
probably resulted from cross-transmission. Younger age was independently
associated with the acquisition of integron-carrying strains, but this remains
unexplained. The unrecognized, widespread presence of integron-containing,
gram-negative bacteria, both within hospitals and in the community, poses a
serious threat of the spread of antibiotic resistance. Establishment of the
endemicity of antibiotic-resistant pathogens in hospitals occurs through
different epidemiological phases: from sporadic monoclonal outbreaks, to
polyclonal outbreaks, to polyclonal endemicity [15,16]. However, the
monoclonal and polyclonal stages are probably interchangeable, depending
on the relative importance of cross-transmission and horizontal gene transfer.
Polyclonal endemicity can only persist if clonal variation is guaranteed, either
by the introduction of colonized patients, the cross-transmission of bacteria,
or the horizontal transfer of genes within the host. Cross-transmission of a
88
dominant clone will change polyclonal endemicity into a monoclonal
outbreak situation. The epidemiological phase associated with the integron-
containing Enterobacteriaceae reported in our study could be described as
being between polyclonal outbreaks and polyclonal endemicity. Our findings
also illustrate the relevance of surveillance and genotyping of resistance
determinants to understanding the epidemiology of antibiotic resistance.
Without genotyping of integrons or Enterobacteriaceae, a situation of low-
level endemicity with only a few circulating bacterial types would have been
found, instead of nine different clusters, some of which largely resulted from
cross-transmission. States of endemicity that are characterized by different
clones has been described as allodemic by Baquero et al. [17]. In the hospital
described by Baquero et al. [17], an allodemic situation developed as a result
of the horizontal transfer of determinants conferring ESBL production
among Enterobacteriaceae. On the basis of the underlying mechanisms of
resistance spread (i.e., gene transfer), they proposed that, in allodemic
situations, interventions should focus more on the environmental causes of
the problem (i.e., antibiotic selective pressure) than on classical approaches to
limiting patient-to-patient transfer. Our findings demonstrate that allodemic
situations may also result from cross-transmission of multiple genotypes. In
such situations, clinical infection-control measures remain the cornerstone of
infection control. The presence of integrons (types VII and XVI) in culture
samples obtained at admission in our study is compatible with the existence
of a community reservoir of integron-carrying isolates or acquisition of
integrons in regular hospital wards. Recent studies have clearly demonstrated
the widespread presence of class 1 integrons carried by isolates obtained from
human community populations in Europe [12], the United States, Africa [18],
and Asia [19] and from animal populations in Europe [20] and the United
89
States [15, 21]. Moreover, class 1 integrons have been found in isolates from
environmental sources, such as contaminated irrigation canals [22]. Because
of these large community reservoirs among animals and humans, colonization
and infection with integron-containing Enterobacteriaceae will not be limited
to patients in ICUs. A large hospital-wide outbreak (mainly affecting surgical
wards and several ICUs) of different genotypes of Enterobacter cloacae carrying
integron type XVI emerged one year after our study (data not shown). Some
aspects of our study should be commented on. First, we screened for
Enterobacteriaceae with reduced susceptibility to cephalosporins and
determined the presence of integrons within this specific group of isolates.
Because sulfamethoxazole resistance was the strongest predictor of the
presence of class 1 integrons in a previous study [12], systematic screening for
sulfamethoxazole-resistant Enterobacteriaceae might have been more
sensitive than the methods we used. Second, if 12 isolates of a species were
available, the first and last isolates were selected for integron detection. This
may have limited detection of intraspecies transfer, because different
genotypes of a certain species could have been present in the intestinal flora.
Third, environmental sources and personnel were not screened, and their role
in the transmission of resistance remains unknown. Nevertheless, on the
basis of the presence of integron types, epidemiological linkage of patients,
and genotyping, it is clear that the spread of integron-associated resistance
occurred predominantly by cross-transmission and less frequently by inter-
and intraspecies transfer of integrons (plasmids or transposons). Therefore,
appropriate infection control strategies are of key importance in controlling
the spread of these antibiotic-resistant strains.
90
References
1. Kollef MH, Fraser VJ. Antibiotic-resistance in the intensive care unit.
Ann Intern Med 2001:298-314.
2. Lipsitch M, Samore MH. Antimicrobial use and antimicrobial
resistance: a population perspective. Emerg Infect Dis 2002:347-54.
3. Martinez-Freijo P, Fluit AC, Schmitz FJ, et al. Class I integrons in
gram-negative isolates from different European hospitals and
association with decreased susceptibility to multiple antibiotic
compounds. J Antimicrob Chemother. 1998:689-96.
4. Hall RM, Collis CM. Antibiotic-resistance in gram-negative bacteria:
the role of gene cassettes and integrons. Drug Resist Updat 1998:109-
19.
5. Rowe-Magnus DA, Mazel D. Resistance gene capture. Curr Opin
Microbiol 1999:483-8.
6. Leverstein-van Hall MA, Box AT, Blok HE, et al. Evidence of
extensive interspecies transfer of integron-mediated antimicrobial
resistance genes among multi-drug resistant Enterobacteriaceae in a
clinical setting. J Infect Dis 2002:49-56.
7. Gruteke P, GoessensW, Van Gils J, et al. Patterns of resistance
associated with integrons, the extended-spectrum β-lactamase SHV-5
gene, and a multi-drug efflux pump of Klebsiella pneumoniae causing a
nosocomial outbreak. J Clin Microbiol 2003:1161-6.
8. Levesque C, Piche L, Larose C et al. PCR mapping of integrons reveals
several novel combinations of resistance genes. Antimicrob Agents
Chemother 1995:185-91.
91
9. van Belkum A, Goessens W, van der Schee C, et al. Rapid emergence
of ciprofloxacin-resistant Enterobacteriaceae containing multiple
gentamicin resistance-associated integrons in a Dutch hospital. Emerg
Infect Dis 2001:862-71.
10. NCCLS. Methods for dilution antimicrobial susceptibility tests for
bacteria that grow aerobically. Approved Standard M7-A4. Wayne, PA:
NCCLS, 1999.
11. Florijn A, Nijssen S, Schmitz FJ et al. Comparison of E-test and
double disk diffusion test for detection of extended spectrum β-
lactamases. Eur J Clin Microbiol Infect Dis 2002:241-3.
12. Leverstein-van Hall MA, Paauw A, Box AT, et al. Presence of
integron-associated resistance in the community is widespread and
contributes to multi-drug resistance in the hospital. J Clin Microbiol
2002:3038-40.
13. Peters ED, Leverstein-van Hall MA, Box AT et al. Novel gene
cassettes and integrons. Antimicrob Agents Chemother 2001:2961-4.
14. Willems RJ, Top J, van den Braak N, et al. Host specificity of
vancomycin-resistant Enterococcus faecium. J Infect Dis 2000:816-23.
15. Nandi S, Maurer JJ, Hofacre C et al. Gram-positive bacteria are a
major reservoir of class 1 antibiotic resistance integrons in poultry litter.
Proc Natl Acad Sci USA 2004:7118-22.
16. Hayden MK. Insights into the epidemiology and control of infection
with vancomycin-resistant Enterococci. Clin Infect Dis 2000:1058-65.
17. Baquero F, Coque TM, Canton R. Allodemics. Lancet Infect Dis
2002:591-2.
92
18. Gassama A, Aidara-Kane A, Chainier D et al. Integron-associated
antibiotic resistance in enteroaggregative and enteroinvasive Escherichia
coli. Microb Drug Resist 2004:27-30.
19. Mathai E, Grape M, Kronvall G. Integrons and multi-drug resistance
among Escherichia coli causing community-acquired urinary tract
infection in southern India. APMIS 2004:159-64.
20. Guerra B, Junker E, Schroeter A, et al. Phenotypic and genotypic
characterization of antimicrobial-resistance in German Escherichia coli
isolates from cattle, swine and poultry. J Antimicrob Chemother
2003:489-92.
21. Gebreyes WA, Thakur S, Davies PR et al. Trends in antimicrobial-
resistance, phage types and integrons among Salmonella serotypes from
pigs, 1997-2000. J Antimicrob Chemother 2004:997-1003.
22. Roe MT, Vega E, Pillai SD. Antimicrobial-resistance markers of class 1
and class 2 integron-bearing Escherichia coli from irrigation water and
sediments. Emerg Infect Dis 2003:822-6.
93
Chapter
Determining the relative importance of bacterial transmission routes in hospital settings
In preparation
Abstract
Nosocomial infections are usually preceded by colonization. Colonization
may either be present already on admission or be acquired during hospital
stay. In the latter case, we need to distinguish between the exogenous route
(i.e, cross-transmission) and the endogenous route (e.g., mutation with
subsequent selection). The likelihood of cross-transmission depends on
colonization pressure, which may change daily, creating so-called non-linear
dynamics. Combining information obtained by genotyping of bacteria with
epidemiological linkage data of patients yields the gold standard to discern
acquisition routes, but these labour-intensive techniques preclude on-going
monitoring of transmission dynamics. Yet, knowledge about the relative
importance of acquisition routes is crucial for infection control.
Based on fundamental differences between dynamics dependent or
independent of colonization pressure, a Markov model was developed [1] and
its ability to determine the relative importance of exogenous and endogenous
acquisition is prospectively evaluated here.
Daily colonization rates of third-generation cephalosporin resistant
Enterobacteriaceae (CRE) were determined by way of microbiological
surveillance in two intensive care units (ICUs). Endogenous and exogenous
transmission rates based upon either model predictions or based upon
genotyping and epidemiological linkage as reference standard were compared.
Daily prevalence of CRE in both ICUs was 26.1±15.4% and 15.1±13.4%,
respectively. According to the reference standard, five out of 23 (21.7%) and
six out of 21 (28.6%) cases of acquired colonization were exogenous,
respectively. The model concludes that the endogenous route is the most
important transmission route in both ICUs (p=.01 for both ICUs).
96
The algorithm introduced in [1] correctly quantified the endogenous routes as
the most important acquisition route in both ICUs and the algorithm can be
used to monitor colonization trends and to analyze intervention effects.
Introduction
The epidemiology of antibiotic resistance in ICUs is complex and quantitative
understanding of the dynamics is essential for designing optimal infection
control strategies. The true volume of antibiotic resistance is best represented
by asymptomatic carriage (i.e., colonization) as only a fraction of colonized
patients will develop infections [2].
Changes in the prevalence of colonization with antibiotic-resistant
microorganisms within hospital settings may occur through different
processes: admission and discharge of colonized and non-colonized patients;
mutations, changing susceptible bacteria into resistant ones, followed by
selection due to antibiotic pressure; and cross-transmission, usually via
temporarily contaminated hands of healthcare workers [3,4]. A key
characteristic of cross-transmission is dependence among patients. The risk
of acquisition (also called ‘colonization pressure’) is influenced by the
colonization status of other patients [5]. This has also been demonstrated for
methicillin-resistant Staphylococcus aureus (MRSA) [6], vancomycin-resistant
Enterococci (VRE) [7] and Enterobacteriaceae [8].
Because of the typically small patient populations in ICUs (usually <20) and
the rapid patient turnover, large fluctuations in proportions of colonized
patients occur naturally [3]. In relatively short time series, with large
fluctuations in numbers of colonized patients, the dependence created by
97
cross-transmission leads to overdispersion and autocorrelation [9]. So, the
distribution of the number of patients colonized at a given day will be skewed
and the variance to mean ratio of the number of patients colonized per day
will exceed 1. Processes in which patients interact are usually called ‘non-
linear’. In contrast, mutations, selection of resistant flora and admission of
colonized patients occur independently of the colonization status of other
patients and these processes are called linear. For these processes there is still
autocorrelation in the number of colonized patients per day (as patients stay
in the unit for some time), but, when data cover a long time period, the
number of patients colonized each day will be binomially distributed.
The distinction between linear and non-linear processes is relevant for the
design of infection control strategies, as well as for the interpretation of the
observed effects of interventions [9,10]. Barrier precautions, for instance, can
only prevent cross-transmission. And routinely used statistical tests assume
independence of events, which is clearly violated when patient-to-patient
transmission is involved (Figure 1, chapter 1) for a worked example how
wrong conclusions can be when dependence is simply neglected [1].
98
In this study, we apply an extension of the Markov model proposed by
Pelupessy et al. [11] for longitudinal colonization data to allow direct
likelihood computation. The adaptations, described in detail in [1], are
• that admission rates are explicitly distinguished from endogenous
selection rates
• that actual changes in bed occupancy are used (as in [12])
• that there is no need to assume that length of stay is exponentially
distributed
• that the moments cultures are performed and the results of these
cultures are the bookkeeping cornerstone of the model while a
stochastic model estimates the status of patients in-between culture
sampling moments
So the model formulation is data driven from the very beginning and
incorporates all the information that is available. It yields maximum
likelihood estimates (as well as confidence regions) for transmission
parameters thus enabling the determination of the predominant acquisition
route.
In this study, we have performed a ‘proof of principle’ by a prospective
comparison of model predictions on the relative importance of endogenous
and exogenous acquisition of third-generation cephalosporin-resistant
Enterobacteriaceae (CRE) in two ICUs with the gold standard derived from
extensive surveillance and genotyping data.
99
Methods
Setting
Colonization with CRE was studied between September 2001 and May 2002
in a medical (ICU-1) and neurosurgical ICU (ICU-2) of the University
Medical Center Utrecht, The Netherlands. ICU-1 has 10 beds, four of which
are in separate rooms and ICU-2 has 8 beds, one in a separate room. Nursing
and medical staff is not shared between these ICUs. Standard infection
control measures were used in both units.
Microbiological surveillance and genotyping
During an eight-month period, rectal colonization with CRE was determined
in all patients admitted to two ICUs. Rectal swabs were obtained on
admission and twice weekly thereafter. Swabs were plated on Chromogenic
UTI Agar (Oxoid Limited, Basingstoke, UK) supplemented with 8 μg/ml
cefpodoxime (Aventis Pharma, Paris, France) and 6 μg/ml vancomycin.
Species identification was performed using VITEK II (bioMérieux Benelux
B.V., ‘s Hertogenbosch, Netherlands). Additional susceptibility testing was
performed by microdilution according to NCCLS guidelines [13] and,
subsequently, all isolates not resistant to either cefpodoxime or ceftazidime
were excluded from analysis. Two isolates per species per patient (if available)
were genotyped using Amplified Fragment-Length Polymorphism (AFLP)
[14]. If more than two isolates of one species were available, first and last
isolates were selected. Genetic relatedness was determined on the basis of
both visual and computerized interpretation of AFLP patterns of isolates of
‘donor’ and ‘acceptor’ patients. A similarity of more than 80% was used as
100
cut-off point and was based on similarities in AFLP-patterns among multiple
isolates obtained from individual patients.
Definitions for microbiological/longitudinal surveillance
Colonization with CRE was classified as ‘present on admission’ when CRE
was demonstrated in cultures obtained <48 hours after admission and as
‘acquired’ when demonstrated in cultures obtained >48 after admission with a
previous negative culture. Two patients in the same ICU were considered to
be epidemiological linked when these patients had either an overlapping
period of stay, or, to allow for survival of pathogens in unidentified reservoirs
[15], when the time between discharge from the ICU of one of the patients
and admission to the ICU of the other patient was at most seven days.
Possible unidentified reservoirs are healthcare workers, environmental
contamination and other patients, which are not sampled at the site of
colonization.
Cross-transmission was defined as acquired colonization with a genetically
similar CRE previously found in an epidemiologically linked patient.
Acquired colonization without epidemiological linkage or genetic relatedness
was considered to be endogenous.
We evaluated the effect of the length of the time window in the definition of
epidemiological linkage.
101
The Markov model
Here we concisely describe the mathematical methodology. For an extensive
description we refer to [1]. The force of colonization consists of a linear term
(parameter a) for endogenous processes and a non-linear term for cross-
transmission, which is given by a constant parameter (β) multiplied by the
prevalence (I/N; with I the number of colonized patients and N the number
of patients in the ICU). The probability per day for an uncolonized patient to
acquire colonization is 1-e(-a-β.I/N), which is approximately equal to the force of
colonization (a+βI/N) when the force of colonization is small. Culture results
are considered 100% reliable and it is assumed that, once established,
colonization persists throughout ICU-stay. As surveillance cultures are not
obtained daily, uncertainty exists about the colonization status of patients in
days when no cultures are obtained and a stochastic model is used to calculate
the likelihood of colonization.
We calculate [1] the values of the parameters a and β (by maximum likelihood
estimation (MLE)) that optimally fit the observations (moments and results
of the culturing and the period of stay per patient). Based on these MLEs the
fraction of patient-days with colonization and the prevalence of colonization
over a period of time can be estimated. The proportion of cross-transmission
is (β I/N)/(a+βI/N) and the proportion of endogenous processes is
a/(a+(βI/N)).
A central concept in infectious disease dynamics is the basic reproduction
number, R0, which corresponds to the average number of secondary infected
cases in a wholly susceptible population [15,16]. Within ICUs, R0 represents
the number of secondary cases generated through cross-transmission by a
primary case in a pathogen-free ward. Infection prevention aims to reduce R0
to an effective R (RN) value below unity. However, for small units (like ICUs)
102
no sharp distinction exists between R0 being above and below unity, and
further reduction of RN may be useful even when RN is already below unity.
Although R0 cannot be determined with this model, RN can be estimated
from the formula: RN = β L, with L being the average duration of stay
(assuming an exponential distributed duration of stay).
Results
Colonization characteristics
In all, 457 patients were studied: 277 admitted to ICU-1 and 180 to ICU-2
and 1243 rectal swabs were obtained (753 in ICU-1 and 490 in ICU-2) (Table
1). Forty-eight patients in ICU-1 and 35 patients in ICU-2 were colonized
during their stay. In ICU-1 23 patients were colonized on admission and 23
patients acquired colonization. In ICU-2 ten patients were colonized on
admission and 21 patients acquired colonization. Routes of acquisition could
not be determined for six patients (two in ICU-1 and four in ICU-2), because
first cultures were taken >48 hours after admission or because patients had
been admitted to the ICU before the start of the study.
The mean daily prevalence of colonization with CRE was 26.1±15.4% in
ICU-1 and 15.1±13.4 in ICU-2. Acquisition rates were 13/1000 and 16/1000
patient-days at risk in ICU-1 and ICU-2, respectively. Mean time to acquire
colonization was 6±8 days in ICU-1 and 8±11 days in ICU-2 (Table 1).
103
Table 1. Colonization characteristics of patients admitted to either ICU.
ICU-1 ICU-2
Admitted patients 277 180
Rectal swabs 753 490
Patients colonized (%) 48 (17.3) 35 (19.4)
Patients with CRE colonization of unknown origin 2 (0.7) 4 (2.2)
Patients colonized on admission (%) 23 (8.3) 10 (5.6)
Patients with acquired colonization (%) 23 (8.3) 10 (5.6)
Endemic prevalence mean ± SD 26.1 ± 15.4 15.1 ± 13.4
Range (%) 0 - 60 0 - 50
Acquisition rate (N acquisitions / 1000 pt-days at risk) 13 16
Mean time to acquisition ± SD 6 ± 8 8 ± 11
Length of stay, mean (SD) 8 ± 11 9 ± 11
In total, 174 isolates (107 patients from ICU-1 and 67 patients from ICU-2)
were genotyped. Based on AFLP results and epidemiological linkage, five
patients in ICU-1 and six patients in ICU-2 acquired colonization via cross-
transmission. Therefore, five out of 23 (21.7%) and six out of 21 (28.6%)
acquired colonizations resulted from cross-transmission in ICU-1 and ICU-2,
respectively, representing cross-transmission rates of 2.9 and 4.5 per 1000
patient-days at risk in ICU-1 and ICU-2, respectively. The ratios between
endogenous and exogenous acquisition were 3.6:1 for ICU-1 and 2.5:1 for
ICU-2.
The time interval in the definition of epidemiological linkage had no major
impact on the number of acquisitions of colonization defined as cross-
transmission. In both ICU-1 and ICU-2 only one case classified as cross-
transmission would be classified as endogenous if the length of the time
window would be zero (time window of 3 and 4 days for ICU-1 and ICU-2,
respectively) and the time window should exceed 21 days to classify more
cases as cross-transmission.
104
Results from modeling while ignoring the information obtained by genotyping
The MLEs for the parameters a, describing endogenous processes, and β,
describing cross-transmission, with their 95% confidence areas and lines of
equal importance of both acquisition routes (β <I/N>= a, with <I/N> being
the mean prevalence) are depicted in Figures 1 and 2. In ICU-1, MLEs for a
and β were 0.022 (95% confidence interval: 0.013-0.032) and 0 (95% CI: 0.0-
0.035) respectively (Figure 1). In ICU-2, MLEs for a and β were 0.025 (0.016-
0.036) and 0 (0.0-0.055), respectively (Figure 2).
Figure 1 and 2. Contour plots of the likelihood of variables a (endogenous acquisition) and b (exogenous acquisition) for third-generation cephalosporin-resistant Enterobacteriaceae in ICU-1 and ICU-2. The shaded area represents the 95% confidence interval. The line indicates equality between endogenous and exogenous acquisition.
ICU-1 ICU-2 Proportions of exogenous colonization was estimated to be 0% for both
ICUs, 95% CI 0-38% and 0-32% for ICU-1 and ICU-2, respectively (using
the profile likelihood method [1]). Both confidence intervals include the
calculated proportions based on epidemiological linkage and genotyping,
which were 21.7 and 28.6% for ICU-1 and ICU-2, respectively (Table 2).
105
Table 2. Epidemiological variables of cephalosporin-resistant Enterobacteriaceae according to genotyping in combination with epidemiological linkage (=observation) and according to model predictions (=model).
ICU-1 ICU-2
Admitted patients Observation Model Observation Model
Endemic prevalence 26.1 ± 15.41 26.0 15.1 ± 13.41 15.5
Proportion cross-transmission (%) 21.7 0 (0-38)2 28.6 0 (0-32)2 1 mean ± SD, 2 95% confidence intervals
Using the profile likelihood method [1], the probability that endogenous
colonization was dominant over cross-transmission was 98.6% and 99.5% for
ICU-1 and ICU-2, respectively. Note that the confidence intervals are
conservative when one of the colonization routes is of no importance [1].
This makes the observation that the endogenous route is dominant even
stronger.
The estimated endemic prevalences based on the MLEs for a and β were
26.0% and 15.5% for ICU-1 and ICU-2, respectively. Both values match the
observed endemic prevalence (26.1±15.4% for ICU-1 and 15.1±13.4%for
ICU-2) very well. Calculated RN values were 0 (95% CI: 0.0-0.025) and 0 (0-
0.44) for ICU-1 and ICU-2, respectively. According to a goodness of fit χ2
test (with 2 parameters) there was no reason to reject the model (p=.29 and
p=.28 for ICU-1 and ICU-2, respectively).
106
Discussion
According to the gold standard provided by a combination of genotyping and
epidemiological linkage data, the Markov chain model accurately quantified
acquisition routes of colonization with third-generation cephalosporin-
resistant Enterobacteriaceae in two ICUs and correctly established
predominance of endogenous over exogenous acquisition. This method,
therefore, seems a promising tool to provide essential information on the
dynamics of microorganisms in hospital settings, without requiring any
labour-intensive and costly genotyping procedures.
Antibiotic resistance is emerging in hospital settings worldwide and with a
diminishing number of antibiotics remaining available for treatment,
prevention of spread will become more important. Up till now, genotyping of
multiple isolates in combination with interpretation of epidemiological data
has been the method of choice to reliably determine dynamics of antibiotic
resistance, especially in endemic settings. However, extensive genotyping is
costly and time-consuming and, therefore, hardly feasible on a daily basis.
The Markov model, as proposed in [1] and in this study, fulfills the need for
an easy and reliable tool to evaluate the dynamics of antibiotic resistance and
is able to disentangle the relevance of patient-dependent and independent
acquisition routes on the basis of longitudinal colonization data only.
Although based on previous theoretical work, the current model differs
significantly in that actual bed occupancy data are used (instead of assuming
constant full occupation), admission of colonized patients is distinguished
from endogenous selection and a stochastical methodology is used to
calculate how likely a patient is colonized during days in-between sampling
moments. When reanalyzing the data with the ‘old’ model [11] higher MLEs
107
and considerably wider confidence intervals are obtained for endogenous
processes (data not shown), which is logical as patients colonized on
admission are also considered as acquisition through endogenous selection.
Consequently, estimated proportions of endogenous selection would increase.
Although this would not influence the interpretation in our setting (where
endogenous colonization is by far more important than exogenous
transmission anyhow) the ‘old’ model would be less reliable for settings with
opposite dynamics.
Our model, as it is now, offers three important advantages for clinical
practice. First, it allows quantifying of the relative importance of exogenous
and endogenous acquisition routes, which is relevant for designing infection
control strategies. Exogenous transmission, usually occurring via temporarily
contaminated hands of staff, depends on healthcare worker related variables,
such as contact rates, level of cohorting and compliance with hand hygiene,
as well as on patient (body site and bacterial load) and microbial
characteristics (such as survival time of microorganisms on hands) [3].
Interventions to reduce cross-transmissions are a good strategy when
exogenous transmission is an important acquisition route. Endogenous
selection is driven by selective antibiotic pressure and does not depend on
colonization pressure in the unit. Transformation from susceptible to
resistant microorganisms can occur through mutations, upregulation of
resistance genes or horizontal gene transfer. In fact, the term ‘acquired’ may
not always be correct, as selection of pre-existing, but undetectable on
admission, flora may only become apparent after some time in ICU.
Reduction of selective antibiotic pressure is needed when endogenous
selection is an important acquisition route.
108
Second, the Markov methodology may improve the reliability of the
interpretation of intervention studies, as it takes patient-dependency into
account. Many infection control interventions (such as improving hand
hygiene, use of gloves and gowns and antibiotic cycling) have been analyzed
in quasi-experimental designs, such as before-after studies [7,10,18-21].
Results were evaluated by standard statistical tests, such as χ2 test, Student’s
T-test and regression analysis that neglect dependence among patients.
Therefore, if cross-transmission is relevant, differences between baseline and
intervention period, considered to be statistically significant according to
these statistical tests, do not necessarily prove causality between intervention
and outcome [1]. The Markov model provides estimates of endogenous and
exogenous transmission rates, in itself correcting for autocorrelation when
cross-transmission is relevant.
Moreover, chance processes such as a temporary lower admission rate of
colonized patients will not influence results. The latter may decrease endemic
prevalence, thus falsely suggesting that an intervention is effective.
Third, this method allows quantification of infection control practices. A
central concept in infectious disease dynamics is the basic reproduction
number R0, which corresponds to the average number of secondary infected
cases in a wholly susceptible population [15,16]. Within hospital settings, R0
represents the number of secondary cases through cross-transmission
generated by a primary case in a pathogen-free ward. Infection prevention
aims to reduce R0 to an effective R (RN) value below unity. In our study the
RN values for third-generation cephalosporin-resistant Enterobacteriaceae
were close to zero in both wards. These findings suggest that an intervention
aimed at reducing cross-transmission can hardly reduce resistance prevalence
109
any further. In other settings or for other pathogens, where RN is >1,
calculation of RN after an intervention allows quantification of its effects.
Our model has some limitations. First, the role of environmental
contamination is not explicitly incorporated. In theory, the colonization status
of a patient might determine the likelihood of contamination of the inanimate
environment and with discharge of the colonized patient, environmental
contamination might disappear as well. In that case, the inanimate
environment could be considered as an extension of the patient and the
Markov model still would apply. Second, the role of persistently colonized
healthcare workers has not been incorporated. Such healthcare workers might
act as a continuous source for transmission. However, despite the fact that
multiple examples of outbreaks caused by healthcare workers exist,
persistently colonized healthcare workers are, in general, not considered
relevant sources for most nosocomial pathogens. Permanently colonized
healthcare workers would impose a colonization pressure that would not
depend on the prevalence of colonized patients and would, therefore, be part
of the so-called endogenous process. Third, colonization is assumed to
remain until discharge, which holds true for many but not all antibiotic-
resistant nosocomial pathogens. Yet, the possibility of intermittent
colonization, or eradication, can easily be included.
These limitation are the reason that a time window was introduced in the
definition of epidemiological linkage. However, omitting the time window in
the definition of epidemiological linkage would only slightly reduce the
fraction of acquisitions classified as cross-transmission.
The observed prevalence and the prevalence calculated with the model
coincides very well. However, there are two differences between both
methods. First, the observed prevalence only looks at patients proven to be
110
colonized while the model also takes into account possible colonization of
patients for which the colonization status is unknown. Second, the model
prediction weights the prevalence each day according to the number of
patients in the ICU. Hence, days with extreme prevalences because only few
patients are present, are less important.
Finally, we emphasize that generalization to other settings with different
patient populations, infrastructure, ecology, antibiotic use, infection control
adherence, patient-staff ratio and colonization pressure requires care. Yet, the
underlying concepts of acquisition and transmission of our model apply to all
nosocomial settings.
111
References
1. Bootsma MCJ, Bonten MJM, Nijssen S et al. Estimating bacterial transmission parameters in hospital settings by direct likelihood computation. In preparation.
2. Bonten MJM, Weinstein RA. The role of colonization in the pathogenesis of nosocomial infections. Infect Control Hosp Epidemiol. 1996:193-200.
3. Bonten MJ, Austin DJ, Lipsitch M. Understanding the spread of antibiotic resistant pathogens in hospitals: mathematical models as a tool for control. Clin Infect Dis. 2001:39-46.
4. Lipsitch M, Samore MH. Antimicrobial use and antimicrobial resistance: a population perspective. Emerg Infect Dis 2002:347-54.
5. Bonten MJM, Slaughter S, Ambergen AW et al. The role of "colonization pressure" in the spread of vancomycin-resistant enterococci. An important infection control variable. Arch Intern Med 1998:1127-32.
6. Merrer J, Santoli F, Appéré-De Vecchi C et al. Colonization pressure and risk of acquisition of methicillin-resistant Staphylococcus aureus in a medical intensive care unit. Infect Control Hosp Epidemiol 2000:718-23.
7. Puzniak LA, Leet T, Mayfield J et al. To gown or not to gown: The effect on acquisition of vancomycin-resistant enterococci. Clin Infect Dis 2002:18-25.
8. Man P de, Veeke E van der , Leemreijze M et al. Enterobacter species in a pediatric hospital: horizontal transfer or selection in individual patients? J Infect Dis 2001:211-4.
9. Cooper B, Lipsitch M. The analysis of hospital infection data using hidden Markov models. Biostatistics 2004:223-37.
10. Harris AD, Bradham DD, Baumgarten M et al. The use and interpretation of quasi-experimental studies in infectious diseases. Clin Infect Dis 2004:1586-91.
112
11. Pelupessy I, Bonten MJ, Diekmann O. How to assess the relative importance of different colonization routes of pathogens within hospital settings. Proc Natl Acad Sci USA 2002:5601-5.
12. Forrester M, Pettitt AN. Use of stochastic epidemic modeling to quantify rates of colonization with methicillin-resistant Staphylococcus aureus in an intensive care unit. Infect. Control Hosp Epidemiol 2005: 598-606.
13. National Comittee for Clinical Laboratory Standards 2001. Performance standards for antimicrobial susceptibility testing. Fourteenth informational supplement. NCCLS document M100-S14. NCCLS, Wayne PA. 2001.
14. Willems RJL, Top J, van den Braak N et al. Host specificity of vancomycin-resistant Enterococcus faecium. J Infect Dis 2000:816-23.
15. Anderson RM, May RM. Infectious diseases of humans. Dynamics and control. Oxford: Oxford University Press; 1991.
16. Diekmann O, Heesterbeek H. Mathematical Epidemiology of Infectious Diseases: Model Building, Analysis and Interpretation. Chichester, U.K.: Wiley; 2000.
17. Grundmann H, Bärwolff S, Tami A et al. How many infections are caused by patient-to-patient transmission in intensive care units? Crit Care Med 2005:946-51.
18. Pittet D, Mourouga P, Perneger TV. Compliance with handwashing in a teaching hospital. Ann Intern Med 1999:126-30.
19. Slaughter S, Hayden MK, Nathan C et al. A comparison of the effect of universal use of gloves and gowns with that of glove use alone on acquisition of vancomycin-resistant enterococci in a medical intensive care unit. Ann Intern Med 1996:448-56.
20. Dominguez EA, Smith TL, Reed E et al. A pilot study of antibiotic cycling in a hematology-oncology unit. Infect Control Hosp Epidemiol 2000:S4-S8.
21. Raymond DP, Pelletier SJ, Crabtree TD et al. Impact of a rotating empiric antibiotic schedule on infectious mortality in an intensive care unit. Crit Care Med 2001:1101-8.
113
Chapter
Comparison of E-tests and double disk diffusion tests for the detection of Extended Spectrum Beta-Lactamases (ESBLs)
Florijn A, Nijssen S, Smitz F, Verhoef J, Fluit A European Journal of Clinical Microbiology and Infectious Diseases 2002; 21:241-43
Introduction
Extended-Spectrum Beta-Lactamases (ESBL) are becoming an increasing
problem in hospitals world-wide. ESBL are plasmid-encoded
cephalosporinases that are inhibited in vitro by clavulanic acid, which
generally belong to the TEM and SHV family of β-lactamases [1].
The detection of ESBL is not straightforward, especially since ESBL may
show MIC values for cefotaxime or ceftriaxone, which are below the
breakpoint for susceptibility defined by the National Committee for Clinical
Laboratory Standards (NCCLS) [2]. In addition, some cephalosporins are also
substrates for chromosomal and chromosomally derived β-lactamases [3].
Demonstration of inhibition by clavulanic acid is commonly used for the
phenotypic detection of ESBL. However, this test can be compromised by
porin changes, TEM-1 overproduction, or inhibitor-resistant TEM (IRT) [1,
4].
Automated diagnostic systems are not always reliable for the detection of
ESBL [5]. Alternatively, a screening method with either, ceftriaxone,
ceftazidime, or aztreonam should be used. According to NCCLS criteria any
bacterium belonging to the Enterobacteriaceae isolated with a MIC greater
than 1 μg/ml for any of these antibiotics should be considered as a potential
ESBL requiring further testing [2]. Potential ESBL should be confirmed by
determining the inhibition by clavalanic acid. Most commonly this done by
E-test or double disk diffusion. The former test employs strips coated with
the relevant antibiotics, which form a gradient after placing them on agar
plates. One end of the strips contains the cephalosporin and the other end
the cephalosporin/clavulanic acid combination. After overnight incubation of
the isolate with the potential ESBL the MIC-values are read from the strips
116
and the inhibition ratio (MIC-value for cephalosporin divided by the MIC-
value for the cephalosporin/clavulanic acid combination) is calculated. In
some cases an inhibition zone can be observed. In the double disk diffusion
test (DDT) a disk with the cephalosporin is placed on an agar plate
inoculated with the test isolate 20-30 mm from a disk containing clavulanic
acid [6,7]. After overnight incubation, the presence or absence of an
inhibition zone is determined. The value of Etest and double disk diffusion is
still a matter of discussion [5,8,9]. Here we report on the performance of
both tests on 404 isolates suspected of carrying ESBL from 25 European
university hospitals, who participated in the SENTRY Program.
Methods
A total of 404 isolates, including 168 Escherichia coli, 33 Klebsiella oxytoca, 161
Klebsiella pneumoniae, and 42 Proteus mirabilis showing a MIC of at least 2 μg/ml
for either ceftriaxone, ceftazidime, or aztreonam were tested with ceftriaxone-
ceftriaxone/clavulanic acid and ceftazidime-ceftazidime/ clavulanic acid E-
test strips (AB Biodisk, Solna, Sweden) and Neo-Sensitabs with 30 μg
ceftriaxone, 30 μg ceftazidime, and 30 μg aztreonam centered around a
30/15 μg amoxicillin/clavulanic acid disk (AB Biodisk) with 30 mm distance
between the centers of the latter disk and the centers of the other disks. Both
strips and disk were placed on 9 cm Mueller-Hinton agar plates inoculated
with a 0.1-0.15 Mc Farland suspension of the isolate suspected of carrying an
ESBL. MIC’s and inhibition zones were read after overnight incubation at
35°C. Discrepant results between E-test and double-disk diffusion were
repeated.
117
Results
Results showed that double-disk diffusion identified 227 ESBL-positive
isolates using all three antibiotics compared to 205 ESBL-positive isolates for
E test using both test strips (Table 1). Concordance between E-test and DDT
was observed in 69.1% of the analyses. This low percentage is mainly due to
the large number of E-tests that could not be interpreted, because the reading Table 1. Comparison between the E-test using both test strips and the DDT using three antibiotics.
DDT result (% of total)
E-test result Positive Negative Total
Positive 199 (49.3) 6 (1.5) 205 (50.7)
Negative 6 (1.5) 80 (19.8) 86 (21.3)
Could not be determined 22 (5.4) 91 (22.5) 113 (28.0)
Total 227 (56.2) 177 (43.8) 404 (100)
of one of the antibiotics was out of range on the test strip and no ratio could
be determined. In 1.5% of the comparisons the DDT was positive, whereas
the E-test was negative. Also in 1.5% of the analyses the E-test was positive,
while the DDT was negative. In 5.4% the DDT test indicated ESBL carriage,
while the E-test yielded a result that could not be interpreted (E-test ND). E-
test ND/DDT-positive isolates were 7 Escherichia coli, 11 Klebsiella pneumoniae,
1 Klebsiella oxytoca, and 3 Proteus mirabilis isolates, whereas the 6 DDT-
negative/E-test-positive isolates belonged to Klebsiella pneumoniae (n=4) and
Escherichia coli (n=2). When ceftazidime alone was used in the DTT 92% of all
DDT-positive isolates were detected, whereas 91% was detected when
ceftriaxone was used. No additional isolates were obtained when aztreonam
was used. These results indicate that the use of aztreonam has no additional
118
value and ceftazidime and ceftriaxone are sufficient for the detection of
ESBL in the DDT.
Discussion
Our results concerning the comparison of the two tests disagree somewhat
with the data from Cormican et al., who compared the E-test using
ceftazidime-ceftazidime/clavulanic acid strips with a disk diffusion test for
225 clinical strains of Escherichia coli and Klebsiella spp., which probably carried
an ESBL and concluded that the E-test was more sensitive than the disk
diffusion test [8]. But these authors took an at least four-fold inhibition by
clavulanic acid as criterion for the presence of an ESBL and only ceftazidime
was tested.
Our results are more in line with the data of Vercauteren et al., who
compared the ceftazidime-ceftazidime/clavulanic acid Etest strip with a
double disk method with ceftriaxone, ceftazidime, aztreonam and cefepime
using 33 well-defined ESBL containing isolates [9]. A ratio for ceftazidime to
ceftazidime/clavalunaic acid greater or equal to 8 was considered indicative
for an ESBL. These investigators noted that all four disks in the disk
diffusion test scored equally by recognizing 31 of the ESBL isolates, although
two false-positive results were obtained with cefepime. The E-test detected
26 of 33 ESBL containing isolates. Most isolates not identified (5 out of 16)
belonged the SHV- family of enzymes. Clavulanic acid interfered in 10 cases
with the reading of the ceftazidime alone on the same strip and an additional
strip with ceftazidime alone had to be used to obtain a test result. In some
cases the range of the strips was insufficient and an inconclusive result was
119
obtained. The same authors used ceftriaxone for double disk diffusion
testing of 86 blood isolates suspected of carrying ESBL. Only six isolates
were positive in this assay and also with the other three antibiotics, except
one Klebsiella oxytoca isolate, which was negative with ceftazidime. This isolate
proved also to be negative in the E-test. In addition, these authors also used a
more complex three dimensional test for the 33 known ESBL-positive strains,
but only ceftriaxone performed well in this assay. Based on the results the
authors proposed to use a combination of double-disk and three-dimensional
tests with ceftriaxone Neo-Sensitabs for ESBL-screening.
M’Zali reported a variant to the DDT called MAST double disk test (MDD)
in which the zones for a cephalosporin combined with clavulanic acid and the
cephalosporin alone are determined [10]. A zone ratio of greater or equal
than 1.5 is considered positive for an ESBL. Results for ceftazidime as
cephalosporin yielded a sensitivity of 86%, whereas this was 65.5% when
cefotaxime was used against a defined set of ESBL carrying isolates. A score
of 93% was obtained when both antibiotics were used. The authors
recommended the MDD as an inexpensive alternative to other methods for
the detection of ESBL-production.
Conclusions
Based on our data we conclude, that the DTT using both ceftazidime and
ceftriaxone and amoxicillin/clavulanic acid Neo-Sensitabs is a cheap and
reliable method to detect Escherichia coli, Klebsiella spp., and Proteus mirabilis
suspected for carrying ESBL in a routine setting when compared to the E-
test, which often yields a result that can not be interpreted.
120
References
1. Livermore DM. β-Lactamases in laboratory and clinical resistance. Clin Microbiol Rev 1995:557-584.
2. National Committee for Clinical Laboratory Standards. Performance standards for antimicrobial susceptibility testing. NCCLS, Wayne, PA, (1998) Supplement Tables M100-S8 to Methods for dilution antimicrobial susceptibility test for bacteria that grew aerobically-fourth edition: approved standard.
3. Pornuil KJ, Rodrigo G, Dornbush K. Production of a plasmid mediated extended-spectrum β-lactamase by a Klebsiella pneumoniae septicaemia isolate. J Antimicrob Chemother 1994:943-954.
4. Vedel G, Belaaouaj A, Gilly L et al. Clinical isolates of Escherichia coli producing TRI β-lactamases: novel TEM-enzymes conferring resistance to β-lactamase inhibitors. J Antimicrob Chemother 1992:449-462.
5. Tenover FC, Mohammed MJ, Gorton TS et al. Detection and reporting of organisms producing extended-spectrum β-lactamases: survey of laboratories in Connecticut. J Clin Microbiol 1999:4065-70.
6. Jarlier V, Nicolas M-H, Fournier G et al. Extended broad-spectrum β-lactamases conferring transferable resistance to newer β-lactam agents in Enterobacteriaceae: hospital prevalence and susceptibility patterns. Rev Infect Dis 1988:867-878.
7. Emery CL, Weymouth LA. Detection and clinical significance of extended-spectrum β-lactamases in a tertiary-care medical center. J Clin Microbiol 1997:2061-67.
8. Cormican MG, Marshall SA, Jones RN. Detection of extended-spectrum β-lactamase (ESBL)-producing strains by E-test ESBL screen. J Clin Mircrobiol 1996:1880-84.
9. Vercauteren E, Descheemaeker P, Ieven M et al. Comparison of screening methods for detection of extended-spectrum β-lactamases and their prevalence among blood isolates of Escherichia coli and Klebsiella spp. in a Belgian teaching hospital. J Clin Microbiol 1997:2191-97.
121
10. M'Zali FH, Chanawong A, Kerr KG et al. Detection of extended-spectrum β-lactamases in members of the family Enterobacteriaceae: comparison of the MAST DD test, the double disc and the E-test ESBL. J Antimicrob Chemother 2000:881-5.
122
Chapter
β- Lactam susceptibilities and prevalence of ESBL-producing isolates among more than 5000 European Enterobacteriaceae isolates
Nijssen S, Florijn A, Bonten M, Smitz F, Verhoef J, Fluit A International Journal of Antimicrobial Agents 2004; 24:585-91
Abstract
In vitro susceptibility to 15 β-lactam antibiotics was evaluated in
Enterobacteriaceae isolated during the SENTRY Antimicrobial Surveillance
Program. Piperacillin/tazobactam was the most active penicillin against
Escherichia coli, Proteus mirabilis, Klebsiella oxytoca and Klebsiella pneumoniae (94.9%,
98.3%, 87.4% and 82.9% of isolates susceptible). Cefepime was the most
effective cephalosporin against Escherichia coli, Proteus mirabilis and Enterobacter
cloacae (99.2%, 96.3% and 95.2% of isolates susceptible, respectively) and
cefoxitin was the most effective cephalosporin against Klebsiella oxytoca and
Klebsiella pneumoniae (98.6% and 95.6% of isolates susceptible). Carbapenems
demonstrated excellent activity (≥99.5%). ESBL-production was confirmed
with ESBL-E-test and disk diffusion test in 1.3% of Escherichia coli isolates,
18.4% of Klebsiella pneumoniae isolates, 12.6% of Klebsiella oxytoca isolates and
5.3% of Proteus mirabilis isolates.
Introduction
Resistance to β-lactam antibiotics has increased significantly in the last two
decades and has been documented in both community and hospital settings
[1,2,3,4]. Extended-Spectrum β-Lactamases (ESBLs) are plasmid-encoded
enzymes, which can arise from point mutations in the TEM-1, SHV-1 and
OXA β-lactamase genes and can hydrolyze β-lactams, including third
generation cephalosporins such as ceftriaxone, ceftazidime, cefotaxime and
the monobactam aztreonam [5,6,7]. These plasmids are easily transmissible in
and between bacterial species. ESBLs occur predominantly in Klebsiella
species and Escherichia coli, but can also be present in other members of
Enterobacteriaceae [8]. Extended-Spectrum β-Lactamases were first
126
recognized in Europe in the early eighties, but are widely distributed all over
the world nowadays [9,10]. Nosocomial outbreaks of infections caused by
ESBL-producing bacteria have been reported repeatedly [11,12,13], but the
prevalence of ESBL-mediated resistance remains unknown for most hospitals
[14].
The first aim of the presented study is to describe the β-lactam susceptibility
patterns of European isolates of the five most prevalent members of the
Enterobacteriaceae, i.e. Escherichia coli, Klebsiella pneumoniae, Enterobacter cloacae,
Proteus mirabilis and Klebsiella oxytoca, collected as part of the SENTRY
Antimicrobial Surveillance Program. In addition, the prevalence of isolates
with an ESBL-phenotype was assesses by using MIC-values for aztreonam or
ceftazidime or ceftriaxone (potential ESBL-phenotype), and confirmed by an
ESBL E-test and a disk diffusion test (DDT) [7,15]. Based on the
susceptibility rates, options for empiric therapy are discussed.
Methods and Materials
The 25 participating European hospitals were requested to send the first 20
blood stream infection isolates of each month, and up to a maximum of 100
isolates from nosocomial pneumonia, 50 isolates from skin and soft tissue
infections, and 50 isolates from urinary tract infections during both 1997 and
1998. Isolates were consecutive, but only one isolate per patient was allowed.
For nosocomial pneumonia only specimen of bronchoalveolar lavage,
tracheal aspirate, and high-quality sputum samples were allowed. The isolates
were speciated at the source and when deemed clinically significant by local
criteria, and were sent to Eijkman-Winkler Institute (the European reference
center for the SENTRY Antimicrobial Surveillance Program) using Amies
127
Charcoal Medium transport swabs (Difco, USA), together with relevant
information for the isolate. Isolates were cultured on blood agar and stored at
-70°C using Microbank (Oxoid, UK) until further testing.
MICs to a range of antibiotics were determined using a broth microdilution
(Sensititre, USA) method and using standard methods defined by the
National Committee for Clinical Laboratory Standards [16].
All 404 SENTRY isolates with MICs ≥2 µg/ml for aztreonam or ceftriaxone
or ceftazidime (potential ESBL-phenotype), were tested with ESBL-E-test
(AB BIODISK, Solna Sweden) as well as disk diffusion test (Rosco,Taastrup
Denmark) with ceftazidime, ceftriaxone, aztreonam and amoxycillin/
clavulanate to confirm ESBL-production. ESBL-E-test and disk diffusion
tests were performed and interpreted as recommended by manufacturer and
NCCLS [16]. ESBL-production was considered to be confirmed if either the
double disk diffusion test or the ESBL E-test was positive [17].
128
Results
A total of 17934 isolates were collected and tested as part of the European
Arm of SENTRY Antimicrobial Surveillance Program from 1997 to 1999,
including 3325 (18.5%) Escherichia coli, 767 (4.3%) Klebsiella pneumoniae, 505
(2.9%) Enterobacter cloacae, 400 (2.2%) Proteus mirabilis and 215 (1.2%) Klebsiella
oxytoca. In Tables 1 to 5, range of antimicrobial activity and susceptibility rates
of 15 β-lactam agents against these members of the Enterobacteriaceae are
shown.
Table 1. Antimicrobial activity spectrum (range, MIC50, MIC90) and susceptibility rates of antimicrobial agents tested against Escherichia coli (n=3325)
Antimicrobial agent
Range
(mg/l)
MIC50
(mg/l)
MIC90
(mg/l)
Susceptible
(%)
Ampicillin ≤ 0.12 to > 16 4 > 16 51.4
Ticarcillin ≤ 1 to > 128 4 > 128 53.0
Piperacillin ≤ 1 to > 128 2 > 128 55.2
Ticarcillin/clavulanate ≤ 1 to > 128 4 32 80.1
Piperacillin/clavulanate ≤ 0.50 to > 128 1 8 94.9
Amoxicillin/clavulanate ≤ 0.12 to > 16 4 16 77.7
Cefazolin ≤ 2 to > 16 ≤ 2 > 16 85.5
Cefoxitin 0.25 to > 32 2 4 96.0
Cefuroxime ≤ 0.12 to > 16 2 8 95.0
Ceftazidime ≤ 0.12 to > 16 ≤ 0.12 0.25 98.1
Ceftriaxone ≤ 0.25 to > 32 ≤ 0.25 ≤ 0.25 98.3
Cefepime ≤ 0.12 to > 16 ≤ 0.12 ≤ 0.12 99.2
Aztreonam ≤ 0.12 to > 16 ≤ 0.12 ≤ 0.12 97.9
Imipenem ≤ 0.06 to 8 0.25 0.50 100
Meropenem ≤ 0.06 to 4 ≤ 0.06 ≤ 0.06 100
129
Table 2. Antimicrobial activity spectrum (range, MIC50, MIC90) and susceptibility rates of antimicrobial agents tested against Proteus mirabilis (n=400)
Antimicrobial agent
Range
(mg/l)
MIC50
(mg/l)
MIC90
(mg/l)
Susceptible
(%)
Ampicillin ≤ 0.25 to > 16 1 > 16 58.0
Ticarcillin ≤ 1 to > 128 ≤ 1 > 128 66.8
Piperacillin ≤ 1 to > 128 ≤ 1 > 128 72.8
Ticarcillin/clavulanate ≤ 1 to > 128 ≤ 1 8 95.0
Piperacillin/clavulanate ≤ 0.50 to > 64 ≤ 0.50 4 98.3
Amoxicillin/clavulanate ≤ 0.12 to > 16 1 16 86.0
Cefazolin ≤ 2 to > 16 4 > 16 71.3
Cefoxitin 0.25 to > 32 2 4 94.8
Cefuroxime ≤ 0.12 to > 16 0.50 > 16 88.3
Ceftazidime ≤ 0.12 to > 16 ≤ 0.12 1 95.3
Ceftriaxone ≤ 0.25 to > 32 ≤ 0.25 0.50 93.8
Cefepime ≤ 0.12 to > 16 ≤ 0.12 0.50 96.3
Aztreonam ≤ 0.12 to > 16 ≤ 0.12 ≤ 0.12 98.0
Imipenem ≤ 0.06 to 8 1 2 99.5
Meropenem ≤ 0.06 to 8 ≤ 0.06 0.12 99.8
Table 3. Antimicrobial activity spectrum (range, MIC50, MIC90) and susceptibility rates of antimicrobial agents tested against Klebsiella oxytoca (n=215)
Antimicrobial agent
Range
(mg/l)
MIC50
(mg/l)
MIC90
(mg/l)
Susceptible
(%)
Ampicillin 2 to > 8 > 16 > 16 3.7
Ticarcillin ≤ 1 to > 128 64 > 128 19.1
Piperacillin ≤ 1 to > 128 16 > 128 63.7
Ticarcillin/clavulanate ≤ 1 to > 128 2 64 85.1
Piperacillin/clavulanate ≤ 0.50 to > 64 1 64 87.4
Amoxicillin/clavulanate ≤ 0.50 to > 16 2 > 16 82.8
Cefazolin ≤ 2 to > 16 8 > 16 56.7
Cefoxitin ≤ 0.25 to > 32 1 2 98.6
Cefuroxime ≤ 0.12 to > 16 1 > 16 84.2
Ceftazidime ≤ 0.12 to > 16 ≤ 0.12 1 94.0
Ceftriaxone ≤ 0.25 to > 32 ≤ 0.25 4 91.2
Cefepime ≤ 0.12 to > 16 ≤ 0.12 2 97.7
Aztreonam ≤ 0.12 to > 16 ≤ 0.12 > 16 87.0
Imipenem ≤ 0.06 to 8 0.50 1 99.6
Meropenem ≤ 0.06 to > 8 0.06 0.12 99.5
130
Table 4. Antimicrobial activity spectrum (range, MIC50, MIC90) and susceptibility rates of antimicrobial agents tested against Klebsiella pneumoniae (n=767)
Antimicrobial agent
Range
(mg/l)
MIC50
(mg/l)
MIC90
(mg/l)
Susceptible
(%)
Ampicillin ≤ 1 to > 16 > 16 > 16 3.4
Ticarcillin ≤ 1 to > 128 128 > 128 6.0
Piperacillin ≤ 1 to > 128 2 > 128 51.8
Ticarcillin/clavulanate ≤ 1 to > 128 2 128 75.0
Piperacillin/clavulanate ≤ 0.50 to > 64 2 64 82.9
Amoxicillin/clavulanate ≤ 0.50 to > 16 4 16 78.6
Cefazolin ≤ 2 to > 8 ≤ 2 > 16 72.0
Cefoxitin ≤ 0.25 to > 32 2 4 95.6
Cefuroxime ≤ 0.12 to > 16 0.50 > 16 80.6
Ceftazidime ≤ 0.12 to > 16 ≤ 0.12 > 16 81.3
Ceftriaxone ≤ 0.25 to > 32 ≤ 0.25 > 32 83.8
Cefepime ≤ 0.12 to > 16 ≤ 0.12 8 91.8
Aztreonam ≤ 0.12 to > 16 ≤ 0.12 > 16 82.1
Imipenem ≤ 0.06 to > 8 0.50 1 99.9
Meropenem ≤ 0.06 to > 8 ≤ 0.06 0.12 99.7
Table 5. Antimicrobial activity spectrum (range, MIC50, MIC90) and susceptibility rates of antimicrobial agents tested against Enterobacter cloacae (n=505)
Antimicrobial agent
Range
(mg/l)
MIC50
(mg/l)
MIC90
(mg/l)
Susceptible
(%)
Ampicillin ≤ 1 to > 16 > 16 > 16 28.3
Ticarcillin ≤ 1 to > 128 2 > 128 64.2
Piperacillin ≤ 1 to > 128 2 > 128 68.7
Ticarcillin/clavulanate ≤ 1 to > 128 2 > 128 65.3
Piperacillin/clavulanate ≤ 0.50 to > 64 1 64 78.4
Amoxicillin/clavulanate ≤ 0.50 to > 16 > 16 > 16 14.5
Cefazolin ≤ 2 to > 16 > 16 > 16 5.5
Cefoxitin ≤ 0.25 to > 32 > 32 > 32 8.1
Cefuroxime ≤ 0.12 to > 16 8 > 16 57.4
Ceftazidime ≤ 0.12 to > 16 0.25 > 16 78.4
Ceftriaxone ≤ 0.25 to > 32 ≤ 0.25 > 32 76.4
Cefepime ≤ 0.12 to > 16 ≤ 0.12 4 95.2
Aztreonam ≤ 0.12 to > 16 ≤ 0.12 > 16 78.4
Imipenem ≤ 0.06 to 8 0.50 2 99.6
Meropenem ≤ 0.06 to 4 ≤ 0.06 0.25 100
131
Of the group of penicillins without a β-lactamase inhibitor, piperacillin
showed the highest activity (51.8-72.8% of the isolates were susceptible) and
ampicillin the lowest (3.4-58% of the isolates were susceptible) activity against
all species. The combinations of penicillins with β-lactamase inhibitors all
showed comparable activity against all species: 75.0-98.3%, with piperacillin/
tazobactam as the most effective combination. Enterobacter cloacae was an
exception to this: susceptibility rates of 14.5 and 65.3% respectively for
amoxycillin/clavulanate and ticarcillin/clavulanate). The first-generation
cephalosporin cefazolin, was the least effective cephalosporin, considering all
five species, with the following susceptibility rates: Escherichia coli: 85.5%,
Proteus mirabilis:71.3%, Klebsiella pneumoniae: 72%, Klebsiella oxytoca: 56.7% and
Enterobacter cloacae: 5.5%. The fourth-generation cephalosporin cefepime was
the most effective cephalosporin against Escherichia coli, Proteus mirabilis and
Enterobacter cloacae: 99.2%, 96.3% and 95.2% of the isolates were susceptible,
respectively. The second-generation cephalosporin cefoxitin was the most
effective cephalosporin against Klebsiella oxytoca and Klebsiella pneumoniae:
98.6% and 95.6% of the isolates were susceptible, respectively. The third-
generation cephalosporins ceftazidime and ceftriaxone rendered good activity
against Escherichia coli (susceptibility rates of 98.1% and 98.3%, respectively),
Proteus mirabilis (susceptibility rates of 95.3% and 93.8, respectively) and
Klebsiella oxytoca (susceptibility rates of 94% and 91.2%, respectively).
Ceftazidime and ceftriaxone were less effective against Klebsiella pneumoniae
(susceptibility rates of 81.3% and 83.8, respectively) and Enterobacter cloacae
(susceptibility rates of 78.4% and 76.4, respectively).
The monobactam aztreonam showed moderate activity against Enterobacter
cloacae (78.4%), Klebsiella pneumoniae (82.1%) and Klebsiella oxytoca (87%), but
was highly effective against Escherichia coli and Proteus mirabilis of which 98%
132
of the isolates were susceptible. The carbapenems demonstrated excellent
activity with more than 99.5% of all isolates susceptible.
All members of the Enterobacteriaceae are known to be potential ESBL-
producers, especially Escherichia coli and Klebsiella pneumoniae. A total of 404
isolates out of 4707 (8.6%), with MICs of ≥2 µg/ml for ceftazidime or
ceftriaxone or aztreonam were tested for confirmation of ESBL-production,
including 168 (5.1%) Escherichia coli, 161 (21.0%) Klebsiella pneumoniae, 33
(15.3%) Klebsiella oxytoca and 42 (10.5%) Proteus mirabilis isolates (Table 6).
Table 6. Number of isolates expressing ESBL-phenotypes and confirmed ESBLs
Species Total Potential ESBL-phenotype Confirmed ESBL-phenotype
n
Of potential ESBL-
phenotype (%)
Of total
(%)
Escherichia coli 3325 168 (5.1%) 44 26.2 1.3
Klebsiella pneumoniae 767 161 (21%) 141 87.6 18.1
Klebsiella oxytoca 215 33 (15.3%) 27 81.8 12.6
Proteus mirabilis 400 42 (10.5%) 21 50 5.3
Total 4707 404 (8.6%) 233 57.7 4.9
Enterobacter cloacae isolates were not tested for confirmation of ESBL-
production. ESBL-production was confirmed in 44 of 3325 (1.3%) isolates of
Escherichia coli, 141 of 767 (18.4%) isolates of Klebsiella pneumoniae, 27 of 215
(12.6%) isolates of Klebsiella oxytoca and 21 of 400 (5.3%) isolates of Proteus
mirabilis (Table 6). ESBL-production is not equally distributed throughout
Europe (Table 7).
Greece, Italy, Portugal, Turkey and Israel show ESBL higher prevalence rates
amongst Enterobacteriaceae than other European countries. No ESBLs were
133
Tab
le 7
. Pot
entia
l and
con
firm
ed E
SBL-
phen
otyp
es fo
r eac
h sp
ecies
and
cou
ntry
Co
untry
To
tal
Esch
erich
ia col
i Kl
ebsie
lla pn
eumo
niae
Kl
ebsie
lla ox
ytoca
Pr
oteus
mira
bilis
n
Pote
ntial
ESB
L-
phen
otyp
e
(%)
Conf
irmed
EBS
L-
phen
otyp
e
(%)
n
Pote
ntial
ESB
L-
phen
otyp
e
(%)
Conf
irmed
EBS
L-
phen
otyp
e
(%)
n
Pote
ntial
ESB
L-
phen
otyp
e
(%)
Conf
irmed
EBS
L-
phen
otyp
e
(%)
n
Pote
ntial
ESB
L-
phen
otyp
e
(%)
Conf
irmed
EBS
L-
phen
otyp
e
(%)
n
Conf
irmed
EBS
L-
phen
otyp
e
(%)
Isra
el 36
19
.4
13.9
18
5.6
0 11
36.4
36
.4
30
0 4
50.0
25.0
Turk
ey
294
28.9
23
.8
172
14.0
8
103
53.3
48
.5
1442
.9
42.9
5
0 0
Alb
ania
28
14.2
0
1910
.5
0 0
0 0
00
0 9
22.2
0
Eng
land
104
4.8
0.96
85
4.7
0 3
33.3
33
.3
00
0 16
0 0
Switz
erlan
d 30
3 2.
0 0.
7 22
88.
3 0.
9 47
0 0
170
0 11
0 0
Spain
79
9 5.
5 1.
6 58
15.
2 0.
3 10
46.
7 5.
8 38
10.5
10
.5
763.
9 1.
3
Portu
gal
230
12.6
10
14
510
.3
5 69
26.1
23
.2
00
0 16
6.3
0
Polan
21
7 11
.5
4.6
159
7.5
2 33
18.2
15
.2
714
.3
14.3
18
33.3
5.6
The
Net
herla
nds
279
8.6
2.5
173
6.9
0 48
10.4
8.
3 24
25.0
12
.5
342.
9 0
Italy
32
0 15
.6
11.9
21
99.
5 7
4035
.0
30.0
16
12.5
12
.5
4528
.928
.9
Gre
ece
357
19.3
12
.6
220
8.2
2 94
42.6
38
9
33.3
22
.2
3423
.55.
9
Ger
man
y 46
6 3.
2 1.
1 34
12.
3 0
704.
3 2.
9 31
9.7
9.7
244.
2 0
Fran
ce
1020
3.
1 1.
1 77
62.
1 0
117
6.8
4.3
368.
3 8.
3 91
5.5
3.3
Belg
ium
11
7 5.
1 1.
7 84
2.4
0 12
0 0
1233
.3
16.7
9
0 0
Aus
tria
137
2.2
0.7
105
1.9
0 16
0 0
812
.5
12.5
8
0 0
47
07
3325
767
215
400
134
found in Albania, but this is probably a consequence of the low number of
isolates collected. ESBL-producing Escherichia coli, were mainly found in
Southern Europe: 39 out of 44 isolates (89%). No ESBL-producing
Escherichia coli were found in North-Western Europe and Austria. In contrast,
Klebsiella pneumoniae isolates producing ESBLs were more equally distributed
throughout Europe although highest prevalences were seen in Southern
Europe (Spain excluded), Poland and Israel. Spain and Western Europe,
except for England, showed low prevalences. The latter was probably due to
low number of isolates collected. Throughout Europe the prevalence of
ESBLs in Klebsiella oxytoca seems relatively high which is partly due to the low
number of isolates collected (Table 7). ESBL-production in Proteus mirabilis
was found in Israel, Greece, Italy, Spain, France and Poland. Of 21 ESBLs
found in Proteus mirabilis, 13 (62%) came from Italy.
135
Discussion
The analysis of the antimicrobial activities of 15 β-lactam antibiotics against
five members of the Enterobacteriaceae family illustrated that most of these
antibiotics still have moderate to high in vitro activities. Resistance rates
against ampicillin and ticarcillin were among highest of all antibiotics. Poor
susceptibility of Enterobacteriaceae to ampicillin has been well known [8].
Combinations of piperacillin and ticarcillin with a β-lactamase inhibitor are
more successful than piperacillin and ticarcillin alone. Escherichia coli showed
susceptibility rates comparable to North America, for the tested
cephalosporins. However, European Klebsiella pneumoniae isolates displayed
considerably higher resistance rates to cephalosporins in general, but
especially to cefazolin and the third-generation cephalosporins, than those
isolated in North America. Enterobacter spp. showed slightly higher resistance
rates against cephalosporins in Europe compared to the U.S [18]. Cefepime
and cefoxitin were the most consistently active cephalosporins. Others have
reported the high susceptibility rates to cefepime as well [2,18,19,20].
Cefoxitin, which was the most successful cephalosporin against European
Klebsiella pneumoniae (95.6% susceptible), was the least successful one in North
America (87% susceptible). Resistance against third-generation
cephalosporins in this SENTRY analysis was 1.7% for ceftriaxone and 1.9%
for ceftazidime among Escherichia coli, and 16.2% for ceftriaxone and 18.7%
for ceftazidime among Klebsiella pneumoniae. Enterobacter cloacae showed
resistance rates of 23.6% to ceftriaxone and 21.6% to ceftazidime. The
SENTRY analysis of resistance to cephalosporins among members of the
Enterobacteriaceae in North America reported resistance rates for ceftriaxone
and ceftazidime among Escherichia coli of 0.2% and 0.8%, respectively. Among
136
Klebsiella pneumoniae, these rates were 0.7% for ceftriaxone and 3.6% for
ceftazidime. Similar resistance rates (20% to ceftriaxone and 23%to
ceftazidime) were reported for Enterobacter spp. [18]. Resistance rates against
imipenem and meropenem are still very low among members of
Enterobacteriaceae in Europe (≤0.5%), United States & Canada, Latin
America and the Western Pacific [2, 8, 19, 20].
As has been described for other parts of the world [17,18], regional variations
in ESBL prevalence are present in Europe. Potential ESBL-phenotypes (MIC
≥2 µg/ml for ceftriaxone, ceftazidime or aztreonam) among Klebsiella
pneumoniae are much more prevalent in Latin America (45.5%) and Western
Pacific (24.6%) than in the U.S. and Canada (7.6% and 4.9%, respectively)
[19]. In this study, 21% of the Klebsiella pneumoniae isolates had MICs ≥ 2
µg/ml and 87.6% of the isolates with a potential ESBL-phenotype were
confirmed to be ESBL-producers. Winokur et al. [19] reported that 42.9%
(North America), 82.4% (Western Pacific region) and 83.8% (Latin America)
of Klebsiella pneumoniae isolates expressing the potential ESBL-phenotype, were
confirmed to be ESBL-producers. The prevalence of confirmed ESBL
phenotypes among Klebsiella pneumoniae in North America was 6.6% [18],
compared to 18.1% in European isolates.
In this analysis Escherichia coli isolates, 26.2% of the isolates expressing the
potential ESBL-phenotype were confirmed to be ESBL-producers. Winokur
et al. [19] reported the following: 27.8% (North America), 61.9% (Latin
America) and 75% (Western Pacific region) of all isolates with a potential
ESBL-phenotype were confirmed to be ESBL-producers. Jones et al. [18]
reported a prevalence of ESBLs among Escherichia coli in North America of
2.7% compared to 1.3% in European isolates. The overall prevalence of
ESBL-producing isolates among Escherichia coli and Klebsiella pneumoniae in
137
North America was 3.8% [18], compared to 4.5% among Escherichia coli and
Klebsiella pneumoniae in this analysis. Proteus mirabilis isolates showing a potential
ESBL-phenotype were also more prevalent in Latin America (22.4%) and
Europe (10.5% in our study) than in the U.S. and Canada (4.9% and 3.1%,
respectively) [19]. The Western Pacific region showed a very low prevalence
(1.8%) [19].
As has been described for the U.S. and Canada [18], the prevalence of ESBLs
is not uniformly distributed throughout Europe. The Southern region of
Europe has higher prevalence of ESBLs than the Northern and Central parts.
This is probably a consequence of different antibiotic policies in these regions
[8]. It should be noted that some prevalences might be higher than they truly
are, because less isolates were sent from those countries.
Enterobacter cloacae isolates show higher resistance rates compared to the other
four members, due to chromosomally encoded AmpC hyperproduction.
Mechanisms of resistance to β-lactams other than ESBL-production are
probably present in the isolates in which ESBL-production could not be
determined. Isolates of Klebsiella pneumoniae resistant to cefoxitin and
cefotaxime, in addition to penicillins, narrow-spectrum cephalosporins,
ceftazidime, aztreonam, and in which the ESBL-phenotype could not be
determined, the likely mechanism of resistance would be AmpC production.
In Klebsiella oxytoca resistance can be caused by K1 β-lactamase
hyperproduction when ceftazidime is not and ceftriaxone and aztreonam
both are hydrolyzed. In this study resistance to aztreonam was 23% and to
ceftazidime 6%. Insignificant levels of uninducible AmpC enzymes in
Escherichia coli are seen in about 2% of most surveys. Chromosomal β-
lactamase expression is negligible in Proteus mirabilis [8].
138
In summary, the presented study illustrated that combinations of a β-lactam
with a β-lactamase inhibitor, e.g. piperacillin/tazobactam, are still good
choices for therapy in the treatment of infections with these five pathogens.
When the combination of a β-lactam with a β-lactamase inhibitor fails, third-
generation cephalosporins are still good options for therapy. However, local
surveillance should be taken into account. For treatment of patients with an
isolate showing lower susceptibility rates to these agents or an ESBL-
producing isolate, cefepime as a fourth-generation cephalosporin and
ultimately the carbapenems could be used [21]. Great differences in regional
prevalences of ESBLs exist throughout the world. Prevalences of ESBLs are
still low with the exception of Klebsiella spp. Screening for ESBL-phenotypes
by using MIC-values is a poor indicator, therefore confirmation by using
ESBL E-test or disk diffusion test is required to get a more reliable estimation
of true ESBL-phenotypes.
139
Acknowledgements
The authors wish to thank H. Mittermayer (Austria), M. Struelens (Belgium),
F. Goldstein (France), V. Jarlier (France), J. Etienne (France), P. R. Courcol
(France), F. Daschner (Germany), U. Hadding (Germany), N. Legakis
(Greece), G.-C. Schito (Italy), G. Raponi (Italy), P. Heczko (Poland), W.
Hyrniewicz (Poland), D. Costa (Portugal), E. Perea (Spain), F. Baquero
(Spain), R. Martin Alvarez (Spain), J. Bille (Switzerland), G. French (UK), R.
Andoni (Albania), V. Korten (Turkey), S. Unal (Turkey), D. Gür (Turkey),
and N. Keller (Israel) for the shipment of isolates. The authors wish to thank
M. Klootwijk, K. Kusters, and S. de Vaal for expert technical assistance. The
SENTRY Antimicrobial Surveillance Program is funded by an educational
grant from Bristol-Myers Squibb Pharmaceutical Company and the European
Network for Antimicrobial Resistance and Epidemiology (ENARE) by a
grant (ERBCHRCT940554) from the European Union.
140
References
1. Medeiros AA. Evolution and dissemination of β-lactamases accelerated by generations of β-lactam antibiotics. Clin Infect Dis 1997: S19-45.
2. Jones RN, Croco MTA, Kugler KC et al., the SENTRY Participants Group (North America). Respiratory tract pathogens isolated from patients hospitalized with suspected pneumonia: frequency of occurrence and antimicrobial susceptibility patterns from SENTRY Antimicrobial Surveillance Program (United States and Canada, 1997). Diagn Microbiol Infect Dis 2000:115-25.
3. Jones RN. Impact of changing pathogens and antimicrobial susceptibility patterns in the treatment of serious infections in hospitalized patients. Am J Med 1996 100:3S-12S.
4. Jones RN. Can antimicrobial activity be sustained? An appraisal of orally administered drugs used for respiratory tract infections. Diagn Microbiol Infect Dis 1997:21-8.
5. Kenneth S, Thomson KS, Sanders CC. A simple and reliable method to screen isolates of Escherichia coli and Klebsiella pneumoniae for the production of TEM- and SHV-derived extended spectrum β-lactamases. Clin Microbiol Infect Dis 1997:549-53.
6. Coudron PE, Moland ES, Sanders CC. Occurrence and detection of extended-spectrum β-lactamases in members of the family Enterobacteriaceae at veterans medical center: seek and you may find. J Clin Microbiol 1997:2593-97.
7. Vercauteren E, Descheemaeker P, Ieven M et al. Comparison of screening methods of extended-spectrum β-lactamases and their prevalence among blood isolates of Escherichia coli and Klebsiella spp. in a Belgian hospital. J Clin Microbiol 1997:2191-97.
8. Livermore DM. β-Lactamases in Laboratory and Clinical Resistance. J Clin Microbiol 1995:557-84.
9. Knothe H, Shah P, Kremery V et al. Transferable resistance to cefotaxime, cefoxitin, cefamandole and cefuroxime in clinical isolates of Klebsiella pneumoniae and Serratia marcescens. Infection 1983:315-317.
141
10. Jacoby GA, Medeiros AA. More Extended-spectrum β-Lactamases. Antimicrob Agents Chemother 1991:1697-1704.
11. Bradford PA, Cherubin CE, Idemyor V et al. Multiply resistant Klebsiella pneumoniae isolates from two Chicago hospitals: identification of the extended- spectrum TEM-12 and TEM-10 ceftazidime-hydrolyzing β-lactamases in a single isolate. Antimicrob Agents Chemother 1994:1211-33.
12. Meyer KS, Urban C, Eagan JA et al. Nosocomial outbreak of Klebsiella infection resistant to late generation cephalosporins. Ann Intern Med 1993:353-58.
13. Rice LB, Willey SH, Papanicolaou GA et al. Outbreak of ceftazidime resistance caused by extended-spectrum β-lactamases at a Massachusetts chronic-care facility. Antimicrob Agents Chemother 1990:2193-99.
14. Emery CL, Weymouth LA. Detection and clinical significance of extended-spectrum β-lactamases in tertiary-care medical centre. J Clin Microbiol 1997:2061-67.
15. Jarlier V, Nicolas G, Fournier G et al. Extended broad-spectrum β-lactamases conferring transferable resistance to newer β-lactam agents in Enterobacteriaceae: hospital prevalence and susceptibility patterns. Rev Infect Dis 1988:867-78.
16. National Committee for Clinical Laboratory Standards. Methods for dilution antimicrobial susceptibility tests for bacteria that grow aerobically. Approved Standard M7-A4. National Committee for Clinical Laboratory Standards, Wayne, Pa 1999.
17. Florijn A, Nijssen S, Schmitz FJ et al. Comparison of E test and double disk diffusion test for detection of extended spectrum β-lactamases. Eur J Clin Microbiol Infect Dis 2002:241-3.
18. Jones RN, Jenkins SG, Hoban DJ et al. In vitro efficacy of six cephalosporins tested against Enterobacteriaceae isolated in 38 North American medical centers participating in the SENTRY Antimicrobial Surveillance Program 1997-1998. Int J Antimicrob Agents 2000: 111-118.
142
19. Winokur PL, Canton R, Casellas JM et al. Variations in prevalence of strains expressing an extended-spectrum β-lactamase phenotype and characterization of isolates from Europe, the Americas and the Western Pacific Region. Clin Infect Dis 2001:S94-103.
20. Mendes C, Hsiung A, Kiffer C et al. Mystic Study Group: Evaluation of the in vitro activity of 9 antimicrobials against bacterial strains isolated from patients in intensive care units in Brazil: MYSTIC Antimicrobial Surveillance Program. Braz J Infect Dis 2000:236-44.
21. Sirot J, Nicolas-Chanoine MH, Chardon H et al. Susceptibility of Enterobacteriaceae to β-lactam agents and fluoroquinolones: a 3-year survey in France. Clin Microbiol Infect 2002:207-213.
143
Chapter
A step-wise reduction of β-lactam exposure, with control of all relevant confounders, failed to reduce acquisition of third-generation cephalosporin-resistant Enterobacteriaceae in two intensive care units
In preparation
Abstract
Understanding the colonization dynamics of resistant pathogens is important
for the design of strategies to control resistance.
The aim of our study was to determine colonization dynamics of third-
generation cephalosporin-resistant Enterobacteriaceae (CRE) in two ICUs
(during a non-outbreak, baseline period) and to evaluate the effects of a single
intervention on acquisition of colonization with CRE. Microbiological
surveillance (including bacterial genotyping), monitoring of infection control
and clinical variables were used to determine all variables relevant in
colonization dynamics. During the eight-month baseline period, the
acquisition rate for CRE was 14/1000 patient-days at risk and CRE-
colonization was predominantly acquired endogenously, with the use of β-
lactams (amoxicillin-clavulanic acid in particular) as a potential and modifiable
risk factor. A heterogeneous regimen (weekly cycling of ceftriaxone,
amoxicillin-clavulanic acid and quinolones) and a homogeneous regimen
(quinolones) to reduce β-lactam exposure were implemented in a randomized
crossover study design to evaluate the effects on acquisition rates of CRE,
with control of all relevant variables.
A step-wise reduction in β-lactam use failed to reduce CRE-acquisition (HR
heterogeneous regimen: 1.0, 95%CR: 0.5-2.2; p=.95 and HR homogeneous
regimen: 1.1, 95% CI: 0.5-2.5; p=.69), but facilitated a dramatic increase in
ciprofloxacin-resistant CRE ( HR heterogeneous regimen: 1.6, 95% CI:0.4-6.5;
p=.50 and HR homogeneous regimen: 4.1 (95% CI: 1.4-11.9; p<.01).
Infection control variables remained comparable during all periods.
The findings of this study demonstrate that a straightforward reduction of β-
lactam exposure, with control of all relevant confounders, does not reduce
acquisition of CRE. The beneficial effects of cycling or rotation of antibiotics
146
on colonization rates of β-lactam-resistant pathogens have not been
unequivocally demonstrated.
Introduction
Antimicrobial resistance is a serious - and continuously increasing - threat to
patient treatment in intensive care units (ICU), worldwide. It is obvious that
antimicrobial use contributes to the emergence and spread of resistant
pathogens, both in hospitals and the community at large. Within ICUs
patients are frequently colonized with antibiotic-resistant bacteria, and the
number (or proportion) of patients colonized is considered a measure of the
resistance problem in a unit. Asymptomatic carriage of resistant pathogens
(i.e. colonization) in a patient may become evident through development of
resistance de novo in previously susceptible bacteria and/or antibiotic-
induced selection. Furthermore, bacteria may be transferred from patient to
patient (i.e. cross-transmission). The risk of cross-transmission is associated
with the colonization pressure (i.e. proportion of patients colonized) [1].
Therefore, this risk of cross-transmission may also increase through
admission of colonized patients. Transmission of resistant pathogens in
hospitals, therefore, is a multi-factorial process in which patient
characteristics, contact rates, staffing and cohorting levels, adherence to hand
hygiene and antimicrobial use are important determinants. Within ICUs, with
typically small patient populations (10-20 patients) and a rapid turnover,
prevalence of resistant pathogens continuously fluctuates because of
stochastic events. Understanding the underlying epidemiological dynamics,
though, is important for the design of strategies to control resistance. And
when analyzing the effects of such an intervention, it is of utmost importance
147
to control for confounding, which may be created by all variables that were
not modulated.
The aim of our study was twofold: first, to determine colonization dynamics
of third-generation cephalosporin-resistant Enterobacteriaceae (CRE) in two
ICUs (during a non-outbreak, baseline period) and second to evaluate the
effects of a single intervention on acquisition of colonization with CRE.
During baseline, acquired CRE colonization appeared to be predominantly
from endogenous origin, with the use of β-lactams (amoxicillin-clavulanic
acid in particular) as a risk factor. Therefore, a stepwise reduction of β-lactam
use was evaluated using a heterogeneous and homogeneous regimen in a
randomized crossover study design.
Patients and Methods
Setting and study design
This study was conducted in two intensive care units, a medical (MICU) and
neurosurgical ICU (NSICU), of the University Medical Centre Utrecht, the
Netherlands. The MICU has ten beds, of which four are in separate rooms,
and is situated in the hospital’s basement. The NSICU has eight beds, one in
a separate room, and is situated on the 4th floor. No medical or nursing staff
is shared between both ICUs.
During a baseline period of eight months (September 9th, 2001 through May
13th, 2002), colonization dynamics of third-generation cephalosporin-resistant
Enterobacteriaceae (CRE) were analyzed by means of microbiological
surveillance and genotyping, collection of demographical and clinical data and
monitoring of infection control practices.
148
Based on the epidemiological findings obtained during baseline (i.e., relative
importance of acquisition routes and risk factors for acquisition) an
intervention was designed to reduce the acquisition rate of CRE.
In a crossover design, the effects of homogeneous (3 months) and
heterogeneous (3 months) antibiotic regimens to reduce patient exposure to
the most frequently used β-lactam antibiotics (amoxicillin-clavulanic acid and
ceftriaxone) were evaluated. During the heterogeneous regimen, first choice
for empirical therapy changed weekly from ceftriaxone in the first week, to
amoxicillin-clavulanic acid in the second week, to levofloxacin or
ciprofloxacin in the third week, and back to ceftriaxone in the fourth week,
etc. Within individual patients, antibiotics started, according to the weekly
schedule, were not adjusted when the weekly schedule changed. During the
homogenous regimen, levofloxacin or ciprofloxacin were the first choice of
empirical therapy. The MICU was randomized to start with the
heterogeneous regimen and the NSICU with the homogeneous regimen.
This study was approved by the institutional review board, which waved the
need of informed consent.
Data collection and microbiological surveillance
All patients admitted were included and age, gender, APACHE II score,
admission indication were recorded on admission. Antibiotic use was
monitored throughout the patient’s ICU-stay. Collection of all variables
(demographics, antibiotic use, infection control measures and microbiological
data) was similar in all three study periods.
Colonization with CRE was determined by means of rectal swabs taken on
admission and twice weekly thereafter. As a first screening, swabs were
incubated overnight at 37°C on Chromogenic UTI agar plates (Oxoid
149
Limited, Basingstoke, UK) supplemented with 8 μg/ml cefpodoxime and
6 μg/ml vancomycin. This growth medium allows identification of subgroups
of Enterobacteriaceae members by distinct coloration, based on chemical
characteristics of these organisms. Species identification of every
morphological distinct colony was performed using VITEK II (bioMérieux,
Lyon, France). Resistance to third-generation cephalosporins was then
confirmed by determination of the MIC values for cefpodoxime and
ceftazidime using Micronaut-S β-lactamase III (Merlin Diagnostika GMBH,
Bornheim-Hersel, Germany). Isolates not resistant to either cefpodoxime or
ceftazidime, according to NCCLS guidelines [2], were considered susceptible
in further analyses. In addition, susceptibility of CRE-isolates to ciprofloxacin
was determined using microdilution susceptibility testing according to
NCCLS guidelines [2].
Colonization on admission was defined as colonization within the first 48
hours after ICU admission. Acquired colonization was defined as
colonization after 48 hours of ICU admission after a previous negative
culture. The primary end-point of analysis was number of acquisitions per
1000 patient days at risk (i.e. CRE acquisition rate).
Cross-transmission was defined as acquired colonization with genotypically
related strains in epidemiologically linked patients. CRE-acquisitions not
fulfilling this definition were considered endogenous acquisition, i.e. selection
of pre-existing flora or de novo resistance development. Epidemiological
linkage was defined as two patients having an overlap in ICU-stay. Because of
the possibility of low-level colonization directly after acquisition, a maximum
time window (between discharge and admission of “ donor” and “acceptor”
patients) of 7 days was accepted [3]. CRE-isolates were genotyped by means
of Amplified Fragment-Length Polymorphism (AFLP) [4]. Genetic
150
relatedness of isolates was based on both visual and computerized
interpretation of AFLP patterns. A similarity of more than 80% was used as
cut-off point and was based on similarities in AFLP-patterns among multiple
isolates obtained from individual patients. Results of surveillance cultures and
genotyping were not available for the hospital’s infection control department
or ICU physicians.
Infection control practices
Observations of patients and nurses were used to determine contact rates (for
both patients and nurses), cohorting levels of nursing staff and adherence to
hand hygiene. Cohorting is defined as the likelihood that after a previous
contact, the next contact of a healthcare worker (HCW) is with the same
patient. Observations were performed by trained infection control nurses,
according to predetermined schedules (unknown to the ICU staff), and were
evenly distributed between 7 am and 11 pm. Two types of observations were
performed - nurse oriented and patient oriented observations - to calculate
contact rates, cohorting of nursing staff and adherence to hand hygiene.
Nurses were observed for 20 minutes, during which the number of contacts
and number of contacted patients (nurse-patient contacts) were recorded, in
order to calculate contact rates and level of cohorting. Patients were also
observed for 20 minutes, during which number of contacts (HCW-patient
contacts), type of healthcare worker (physician, nurse, physical therapist,
radiology assistant, etc.), type of contact, use and removal of gloves, use and
type of hand hygiene were recorded, in order to determine contact rates and
adherence to hand hygiene.
A patient contact was defined as any contact with a patient’s skin or gown.
Contacts with other inanimate objects were considered environmental.
151
Gloves needed to be removed and hand hygiene to be used before returning
to the communal environment of the ICU. Appropriate hand hygiene was
considered to be either washing hands with soap and water or use of
alcoholic hand rub. Both ICUs are provided with two sinks with both soap
and alcohol hand rub dispensers, with each bedside also having its own
alcohol hand rub dispenser.
Statistical and risk factor analysis
Continuous variables were analysed by Student’s T-test or Mann-Whitney U
test and categorical variables by χ2 statistics. Acquisition rates were compared
using a multivariate Cox’ proportional hazard model, with subsequent
addition of potential confounders (all variables with p<.10). The model
calculates hazard ratios (HR) and 95% confidence intervals. All analyses were
performed with SPSS software (SPSS Inc. Chicago, IL).
Results
Baseline
Patient characteristics, CRE-colonization and infection control
All analyses were performed separately for both ICUs but did not reveal
relevant differences (data not shown). Therefore, we present the combined
data of both wards. During the 8-month baseline period, 457 patients were
admitted (Table 1).
Thirty-three patients (7.2%) were colonized with CRE on admission and 44
patients (9.6%) acquired colonization during their stay in ICU (Table 2). Of
six patients origin of colonization could not be determined because either
152
cultures were taken more than 48 hours after admission or these patients
were already admitted before the start of the study. The CRE-acquisition rate
was 14/1000 patient-days at risk with a mean time to acquisition of seven
days. Based on epidemiological linkage and genotyping CRE-colonization was
predominantly acquired endogenously: 11 of 44 (25%) cases of acquired
colonization (five in the MICU (21.7%) and six in the NSICU (28.6%))
resulted from cross-transmission. Therefore, the endogenous route was
considered the dominant route for acquired colonization. Of all patients
colonized with CRE, 16 were colonized with a ciprofloxacin-resistant isolate.
Of these, nine were colonized on admission and six acquired such a CRE in
the ICU with a mean time to colonization of 7±10 days. The acquisition rate
of ciprofloxacin-resistant CRE was 2.1/1000 patient days.
153
Tab
le 1.
Pop
ulat
ion
char
acte
ristic
s. Va
riabl
e Pe
riod
Patie
nts d
emog
raph
ics
Bas
elin
e H
eter
ogen
eous
H
omog
eneo
us
p-va
lue
Patie
nt-d
ays
38
18
12
81
11
76
Adm
itted
pat
ient
s
457
17
6
135
Age
, yea
rs
53
± 1
9
57 ±
18
56
± 1
5 0.
01/0
.08
APA
CHE
II sc
ore
21
± 8
23 ±
8
21
± 7
0.
07/0
.98
MIC
U st
ay, d
ays
8
± 1
1
7 ±
9
9
± 1
0 0.
61/0
.31
Male
sex,
no.
(%) o
f pat
ient
s
244
(53.
4)
10
4 (5
9.1)
82 (6
0.7)
0.
20/0
.13
Mor
talit
y, no
. (%
)
83 (1
8.2)
23 (1
3.6)
29 (2
1.5)
0.
35/0
.39
Adm
issi
on in
dica
tion
Card
iova
scul
ar
30
(6.6
)
9 (5
.1)
9
(6.7
) 0.
50/0
.97
Pulm
onar
y
94 (2
0.6)
36 (2
0.5)
34 (2
5.2)
0.
98/0
.25
Gas
tro-in
test
inal
10
(2.2
)
4 (2
.3)
0
0.95
/0.0
8
Neu
rolo
gica
l
72 (1
5.8)
28 (1
5.9)
21 (1
5.6)
0.
96/0
.96
Seps
is
26 (5
.7)
6
(3.4
)
7 (5
.2)
0.24
/0.8
2
Trau
ma
45
(9.8
)
14 (8
.0)
5
(3.7
) 0.
46/0
.03
Surg
ery
13
9 (3
0.4)
72 (4
0.9)
49 (3
6.3)
0.
01/0
.20
Oth
er
41
(9.0
)
7 (4
.0)
10
(7.4
) 0.
03/0
.57
Antib
iotic
use
(DD
D/1
000-
pt d
ays a
nd n
(%))
Patie
nts r
eceiv
ing
antib
iotic
ther
apy
30
6 (6
7%)
10
7 (6
1%)
97
(72%
) 0.
12/0
.33
Am
oxici
llin-
clavu
lanic
acid
32
6 16
8 (3
7%)
131
(-59.
8%)
37 (2
1%)
31 (-
90.5
%)
21 (1
6%)
<0.
01/<
0.01
Ceftr
iaxon
e 13
4 78
(17%
) 13
0 (-3
.0%
) 33
(19%
) 55
(-59
.0%
) 13
(10%
) 0.
68/0
.04
Oth
er β
-lact
ams1
39
4 14
3 (3
1%)
265
(-32.
7%)
52 (3
0%)
469
(19.
0%)
55 (4
1%)
0.91
/0.2
0
Am
inig
lycos
ides
15
9 10
4 (2
3%)
142
(-10.
7%)
37 (2
1%)
91 (-
42.8
%)
19 (1
4%)
0.64
/0.0
3
Qui
nolo
nes
150
38 (8
%)
129
(-14.
0%)
23 (1
3%)
514
(243
%)
69 (5
1%)
0.07
/<0.
01
154
Table 2. Colonization characteristics.
Variable Period
Cephalosporin-resistant Enterobacteriaceae Baseline Heterogeneous Homogeneous p-value
Patients with CRE colonization (%) 83 (18.2) 26 (14.8) 29 (21.5) 0.31/0.39
Patients with colonization on admission (%) 33 (7.2) 11 (6.3) 12 (8.9) 0.65/0.53
Patients with acquired colonization (%) 44 (9.6) 14 (8.0) 16 (11.9) 0.50/0.45
Acquisition rate/ 1000 patient-days at risk 14 14 18 0.95/0.69
Mean days to acquisition 7 ± 9 6 ± 7 7 ± 8 0.71/0.73
Ciprofloxacin-resistant CRE
Patients ciprofloxacin-resistant CRE 16 (3.5) 4 (2.3) 11 (8.1) 0.43/0.02
Patients ciprofloxacin-resistant CRE-isolate on
admission 9 (2.0) 1 (0.6) 1 (0.8) 0.30/0.47
Patients acquired ciprofloxacin-resistant CRE-isolate 6 (1.3) 3 (1.7) 8 (6.0) 0.71/<0.01
Mean days to acquisition 7 ± 10 7 ± 8 7 ± 8 0.98/0.47
Acquisition rate ciprofloxacin-resistant isolate/1000
patient-days 2.1 2.5 8.3 0.50/0.01
In total, 352 nurse-patient contacts (nurse observations) and 435 HCW-
patient contacts (patient observations) were observed during 197 hours
(Table 3). Nurses had 3.2±1.3 patient contacts/hour and their level of
cohorting was 71%±22%. Patients received 4.0±1.8 contacts/hour from
healthcare workers (nurses, physicians, radiology technicians, physical
therapists, etc.). Adherence to hand hygiene after patient contact was 55%
overall, 59% for physicians and 53% for nurses (p=.399) (Table 3).
Risk factors for acquisition with CRE
In univariate analysis, CRE-acquisition was associated with a pulmonary
admission indication, trauma, admission after surgery and admission for
‘other’ indications (Table 4). Furthermore, all patients acquiring CRE had
received antibiotics, as compared to 63% of non-affected patients (p<.01).
Among antibiotics, amoxicillin-clavulanic acid and aminoglycosides were
155
Table 3. Contact rates, compliance and cohorting.
Variable Period
Baseline Heterogeneous Homogeneous p-value
Observation of patients
No. of HCW-patient contacts 435 132 186
Contact rates patients (contacts/hour) 4.0 ± 1.8 2.9 ± 1.3 4.3 ± 2.5 0.01/0.70
Compliance (%) 55 57 53 0.77/0.73
Physicians (%) 59 63 43 0.75/0.28
Nurses (%) 53 55 58 0.73/0.42
Observation of nurses
No. of HCW-patient contacts 352 119 148
Contact rates nurses (contacts/hour) 3.2 ± 1.3 2.5 ± 1.0 2.8 ± 1.4 0.02/0.28
Cohorted contacts (%) 71 ± 22 74 ± 23 74 ± 25 0.66/0.65
associated with CRE-acquisition. ICU-ward, APACHE II score and patient-
specific contact rates and hand hygiene were not associated with CRE-
acquisition. For this analysis, hand hygiene and contact rates were calculated
on patient-level instead of using the means of each period. As observations
were not performed daily, data were not available for all patients and these
were excluded from analysis. In multivariate analysis, only admission because
of trauma remained an independent risk factor (Hazard ratio (HR):2.7,
Confidence interval (CI): 1.1-6.6) (Table 4). As aminoglycosides were most of
the time prescribed in combination with β-lactam antibiotics, their association
with acquisition disappeared in multivariate analysis. For amoxicillin-
clavulanic acid only a non-significant trend remained in multivariate analysis.
156
Table 4. Risk factors for CRE acquisition.
Variable Without CRE
(n = 374) Acquired CRE
(n = 44) p-value HR 95% CI p-value
Demographics
Age 54± 18 55 ± 18 0.81
Male sex 195 (52%) 26 (59%) 0.38
APACHE II score 21 ± 8 22 ± 6 0.58
MICU 229 (61%) 23 (52%) 0.25
Contact rate 4.6 ± 3.1 5.5 ± 3.6 0.22
Hand hygiene 47% ± 41% 53% ± 39% 0.59
Admission diagnoses
Pulmonairy disease 66 (18%) 15 (34%) 0.01 2.10 0.88 – 5.03 0.09
Cardiovasculair disease 24 (6%) 1 (2%) 0.27
Neurological disease 62 (17%) 5 (11%) 0.37
Trauma 30 (8%) 14 (32%) <0.01 2.71 1.11 – 6.61 0.03
Surgery 124 (32%) 6 (14%) 0.01 1.50 0.52 – 4.35 0.45
Other 68 (18%) 3(7%) 0.06 1.45 0.18 – 11.91 0.73
Antibiotic therapy 237 (63%) 44 (100%) <0.01
Amoxicillin-clavulanic acid 127 (34%) 32 (73%) <0.01 1.49 0.70 – 3.14 0.30
Ceftriaxone 57 (15%) 8 (18%) 0.61
Aminoglycosides 71 (19%) 22 (50%) <0.01 0.97 0.50 – 1.90 0.93
Quinolones 28 (8%) 5 (11%) 0.37
In summary, acquired colonization with CRE predominantly occurred
endogenously (75% of acquisitions) with the use of amoxicillin-clavulanic
acid and aminoglycosides identified as potential and modifiable risk factors in
univariate analysis. After controlling for admission categories, only a trend
towards an increased risk of CRE-acquisition after or during use of
amoxicillin-clavulanic acid remained. Based on the association of β-lactam use
and acquisition of gram-negatives resistant to these antibiotics, reported by
others [5,6], we hypothesized that reducing the use of β-lactam antibiotics in
general, and amoxicillin-clavulanic acid more particularly, could decrease
endogenous acquisition.
157
Intervention period
During intervention periods, patient characteristics remained comparable to
baseline (Table 1). Again, all analyses were performed first for both ICUs
separately (data not shown). As there were no relevant demographic
differences between the wards (apart from indication of admission) and
intervention effects were comparable, data are presented simultaneously.
Percentages of patients colonized on admission with CRE or ciprofloxacin-
resistant CRE were comparable in all three study periods.
Antibiotic use
As compared to baseline, overall usage of antibiotics did not change. During
baseline 67% of all patients received antibiotics, as compared to 61% and
72% during heterogeneous and homogeneous study periods, respectively
(p=.12 and p=.33). Yet, amoxicillin-clavulanic acid use decreased from 37%
of all patients in baseline to 21% (p<.01) and 16% (p<.01) during
heterogeneous and homogeneous periods, respectively. Expressed in
DDD/1000 patient days, use of amoxicillin-clavulanic acid was reduced from
326 in baseline, to 131 in the heterogeneous and 31 in the homogeneous
period, respectively. Usage of ceftriaxone during the heterogeneous regimen
(19%, 130 DDD/1000 patient-days) was comparable to baseline (17%, 134
DDD/1000 patient days), but decreased to 10% (55 DDD/1000 patient days)
during the homogeneous regimen (p<.01). Quinolone use slightly decreased
in the heterogeneous period when analyzed as DDD/1000 patient days (150
versus 129 in baseline and heterogeneous period, respectively), although
proportions of patients being exposed slightly increased (from 8% during
baseline to 13% in the heterogeneous period, (p=.07)). Yet, quinolone use
was multiplied in the homogenous period, with 51% of all patients receiving
158
quinolones and a total exposure of 514 DDD/1000 patient days. From this
we can conclude that both antibiotic regimens were successfully implemented.
Infection control
Contact rates received by patients were lowest during the heterogeneous
period; 2.9±1.3, as compared to 4.0±1.8 (p=.01) during baseline and 4.3±2.5
(p=.02) during the homogeneous period (Table 3). The contacts rates of
nurses were also lowest during the heterogeneous period: 2.5±1.0, as
compared to 3.2±1.3 (p=.02) during baseline and 2.8±1.4 (p=.28) during the
homogeneous period. Adherence to hand hygiene and cohorting levels were
comparable during all three study periods.
Introduction of resistance
Percentages of patients colonized on admission with CRE or ciprofloxacin-
resistant CRE were comparable in all baseline and intervention periods.
Acquisition rates
Acquisition rates of CRE did not differ significantly during the three study
periods. Actually, a non-significant trend towards a higher CRE-acquisition
rate (18/1000 patient-days at risk) was found during the homogeneous period
(in which use of β-lactam antibiotics was lowest).
In a Cox’ proportional hazards model (using baseline as reference), with
adjustment for ICU, age, APACHE II score, admission indication and
contact rates as potential confounders, no differences in acquisition rates
were observed (HR: 1.0, 95% CI:0.5-2.2; p=.95 for the heterogeneous
159
regimen and HR: 1.1, 95% CI: 0.5-2.5; p=.69 for the homogeneous regimens)
(Figure 1a).
Figure 1a. Hazard functions CRE for all periods.
Genotyping of CRE was not routinely performed during the intervention
periods, although some isolates were genotyped for other reasons. With that
information and using species determination and epidemiological linkage, we
could estimate the number of potential cases of cross-transmission. Although
probably still underestimating the real number of cases, we were able to
confirm at least some of these potential cases to be the result of cross-
transmission. Based on species determination and epidemiological linkage 12
of 14 patients (86%) could have acquired colonization through cross-
transmission during the heterogeneous period and 13 of 16 patients (81%)
during the homogeneous period, of which three (of 12, 21%) during the
heterogeneous and six (of 13, 38%) during the homogeneous period were
160
confirmed by genotyping. If all uncertain cases had resulted from cross-
transmission, exogenous acquisition would have been dominant during the
intervention. Thus, the reduction in endogenous acquisition (as targeted by
the intervention) was, at least to some extent, counterbalanced by an increase
of exogenous transmission.
Acquisition rates of ciprofloxacin-resistant CRE were highest (8.3 per 1000
patient-days at risk) during the homogeneous regimen in which quinolone-
use was also highest (Table 2) with a HR of 4.1 (95% CI: 1.4-11.9; p<.01) in
Cox’ regression (using baseline as reference). No significant difference in
acquisition could be demonstrated between baseline and the heterogeneous
period (HR: 1.6, 95% CI: 0.4-6.5; p=.50) (Figure 1b). Figure 1b. Hazard functions ciprofloxacin-resistant CRE for all periods.
161
In all, 17 patients acquired colonization with ciprofloxacin-resistant CRE.
Five patients had acquired isolates that were genotypically related and
epidemiologically linked to those of other patients (no isolates during baseline,
one (33%) during the heterogeneous period and four (50%) during the
homogeneous period). The remaining 12 patients did either not have
epidemiological linkage or acquired colonization with other genotypes, and
were, thus, considered to have acquired colonization via the endogenous
route.
Discussion
In this study, we evaluated the effects of two antibiotic regimens on
acquisition of third-generation cephalosporin-resistant Enterobacteriaceae
(CRE) in two ICUs, with monitoring of all variables relevant for colonization
dynamics. In a baseline period acquisition of CRE was predominantly
endogenous, with only few cases resulting from cross-transmission, and with
the use of amoxicillin-clavulanic acid as the most important modifiable risk
factor for acquisition. A subsequent step-wise reduction of amoxicillin-
clavulanic acid use by 60% and 91% and ceftriaxone use by 3% and 59%, at
the costs of increased usage of quinolones, failed to reduce CRE-acquisition,
but facilitated a dramatic increase in ciprofloxacin-resistant CRE.
The findings of this study demonstrate that a straightforward reduction of β-
lactam exposure, with control of all relevant confounders, does not reduce
acquisition of β-lactam-resistant gram-negatives.
As antibiotic use is the driving force in emergence and spread of
antimicrobial resistance, it seems logical that reducing antibiotic pressure will
also reduce the prevalence of antimicrobial resistance [7]. Apart from
162
restriction of particular antimicrobial classes, cycling and rotation of classes
also have been proposed as alternative strategies to control emergence of
resistance. The rationale is that heterogeneous antimicrobial use, compared to
homogeneous use, creates fluctuating selection pressure, reduces adaptation
ability by resistance acquisition and spread of present resistance. The shorter
the cycles are, the larger heterogeneity is. This effect is enforced by high
patient turnover (usually the case in ICUs), which dilutes resistance
prevalence in the absence of continuous selection, assuming that patients
admitted are not colonized with resistant pathogens [8]. Based on a
mathematical model, Bergstrom et al. predicted that cycling on patient-level is
more effective in preventing resistance development than temporal cycling on
a ward-level [7]. Moreover, long cycling duration might even increase, rather
than decrease, resistance.
Several studies have evaluated the effects of cycling or rotational
antimicrobial use on colonization with different pathogens in ICUs [9-13].
Three of these studies compared rotation/cycling of different antimicrobial
classes to unrestricted antibiotic use, but failed to demonstrate significant
decreases in colonization rates [9,10,12]. In another study quinolones
(levofloxacin) were cycled with β-lactams (cefpirome and piperacillin-
tazobactam) during 4-month periods [13]. In each period, though, resistance
rates to the antibiotic of choice during that period increased, especially for
levofloxacin and piperacillin-tazobactam. Moreover, there was a relative
increase of cross-transmission during the first cycle of quinolone use. These
findings, therefore, do not support the presumed beneficial effects of
antibiotic cycling. In a recent study, the effects of monthly cycling of different
antimicrobial classes on acquisition of resistant gram-negatives was compared
to cycling on patient-level (mixing), using a crossover design in two intensive
163
care units [11]. Acquisition of cefepime-resistant Pseudomonas aeruginosa was
significantly higher during mixing and, in addition, trends towards more
acquisition of resistance to ceftazidime, imipenem and meropenem were
observed. Two studies investigated the effects of cycling or rotation on
infection rates (compared to colonization rates) in ICUs and reduced
resistance and infection rates were indeed found [14,15]. Unfortunately,
potential confounders were not carefully determined or analyzed in any of
these studies. Moreover, many variables, apart from antibiotic use, (infection
control practices, diagnostic strategies) changed in the two studies reporting
reduced incidence rates [14,15].
From this it can be concluded that, so far, the beneficial effects of cycling or
rotation on colonization rates of resistant pathogens have not been
unequivocally demonstrated. Importantly, all studies suffer, at least to some
extent, from methodological flaws, such as a lack of baseline measurement,
not controlling of all relevant variables during intervention, implementation
of more than one intervention and a quasi-experimental design rather than
randomized controlled trials.
To our knowledge, this the first study, using prospective surveillance to
evaluate the effects of a single intervention in antimicrobial prescription in
two ICUs in a crossover design, with baseline measurement and subsequent
control of all variables relevant to colonization dynamics. Determining
introduction of resistance, route of acquisition, contact rates, cohorting of
nurses and adherence of hand hygiene during a baseline period is important
in order to design appropriate control strategies. In addition, these variables
are potential confounders in intervention studies and should therefore be
determined during intervention periods as well. In addition, this study is
unique in that it compares a heterogeneous antimicrobial regimen of weekly
164
cycling of three antimicrobial classes to a homogeneous regimen of quinolone
use in two intensive care units, using a crossover design.
Antibiotic use, contact rates, cohorting levels and adherence to hand hygiene
were all comparable to those found in other studies.
Data for reliable comparison of antimicrobial use between different wards,
hospitals or countries are limited, as uniform reporting is lacking.
Antimicrobial use during our baseline period appeared to be comparable to
antibiotic use in German ICUs participating in SARI (Surveillance of
Antimicrobial Use and Antimicrobial Resistance in ICUs), which also used
the World Health Organization definition (DDD/Anatomical Therapeutical
Classification). [16]. Total antibiotic use (1380 DDD/1000 patient-days in our
setting) was 1332 DDD/1000 patient-days in German ICUs. Amoxillin-
clavulanic acid was the most frequently used drug in our and German ICUs
(326 and 208 DDD/1000 patient-days, respectively). Median usage of third-
generation cephalosporins ranged from 90 to 125 DDD/1000 patient-days
depending on ICU type (surgical, interdisciplinary, or medical ICU) in SARI
ICUs compared to 155 in our ICUs. Median fluoroquinolone use ranged
from 130-150 DDD/1000 patient-days in SARI ICUs compared to 150
DDD/1000 patient-days in our study. Median aminoglycoside use was 50 in
SARI ICUs vs. 159 DDD/1000 patient-days in our ICUs. However, careful
interpretation is needed as severity of illness of patient with dosage
adjustment as a consequence, is not taken into account. There was a positive
correlation between length of stay and total antibiotic use in SARI ICUs. In
ICUs in which patients had a mean length of stay of eight days (comparable
to our ICUs during baseline) mean total antibiotic use was 2000 DDD/1000
patient-days, which is 45% higher than in our setting. Considering the average
total antibiotic use (1332 DDD/1000 patient-days) in SARI ICUs, containing
165
only one ICU with a mean length of stay of eight days, implies that antibiotic
use in our ICUs was comparable to that of German ICUs with a shorter
length of stay (five to six days according to data) and thus relatively lower.
In a report by Kern et al., antimicrobial use in 53 ICUs in Southern Germany
according to WHO-definitions ranged form 14.4-182.9 DDD/100 patient-
days (comparable to 144-1829 DDD/1000 patient-days) with β-lactams and
fluoroquinolones used most frequently: 59% and 11% of total use (61% and
8% in our ICUs) [17]. Total antibiotic use in our ICUs was within the same
range as reported in this study as was β-lactam use. Aminoglycosides were
used more frequently than fluoroquinolones in our ICUs compared to SARI
ICUs.
Contact rates between healthcare workers and ICU-patients have been
measured in only a few studies [18-21]. Grundmann et al. found an average
contact rate of 3.0 contacts/patient/hour in a British mixed-ICU [20]. In
another study, contact rates, expressed as the average number of
opportunities to use hand hygiene after patient care per hour in an ICU (all
healthcare workers included), ranged from 0 to more than 60 contacts per
hour (complete ICU) depending on time of day and ICU [19]. We expressed
patient contact rates as the number of contacts received by an individual
patient per hour (instead of complete ICU), which ranged from 0 to 9
contacts/hour. In an eight-bed ICU with 100% occupancy our contact rates
are comparable to those found by Pittet et al. [19]. A study by Kim et al.
observed 589 opportunities for hand disinfection (i.e. patient contacts) during
40 hours of observation in two ICUs [22]. Although this study was not
designed to calculate contact rates, the average contact rate can be estimated
at 15 contacts/hour (for the complete ICU), which is in range with the
contact rate found by Pittet et al. and our data. McArdle et al. reports a
166
patient contact rate of 350/patient/day, distinguishing between direct (45%)
and indirect patient contacts (55%) [21]. The estimated number of
contacts/patient/hour (according to our definition) is approximately 6.6. So,
compared to literature, contact rates in these ICUs were comparable to those
found by others.
Cohorting of nurses has not been determined in many studies either.
Grundmann et al. reported an average nurse cohorting level, using definitions
comparable to the ones used in this study, of 70% (range 46-84%) [20].
McArdle et al. observed that 74% of all direct patient contacts was by nurses
who cared for a single patient, which is comparable to what we found in our
study [21]. In another study, the cohorting level of nurses in a 16-bed ICU
was 77% [18]. The nurse-patient ratio can serve as a surrogate marker for
cohorting levels as the latter is a derivative of the former and several studies
used the nurse-patient ratio to describe its role in ICU pathogen transmission
[23,24]. Although data on cohorting levels are not abundant, the former
shows cohorting levels measured so far, all lie between 70 and 80%.
Adherence to hand hygiene has been measured in numerous studies although
not in a universal manner. Adherence rates in ICUs have ranged from 12% to
81%, but usually do not exceed 50%. Moreover, adherence rates depend on
type of healthcare worker, time of day and type and intensity of patient care
[19,20]. Although still not optimal, adherence rates, as observed in our study,
appeared to be above-average. As acquisition in our ICUs was predominantly
endogenous, it is questionable whether changes in hand hygiene would have
influenced acquisition of CRE, but it could have changed the number of
exogenous acquisitions and, therefore, the relative importance of both routes.
Our study has several limitations. First, duration of our intervention period
was relatively short and with low acquisition rates, it is imaginable that
167
interpretation would have become more reliable with a prolonged duration of
intervention. Yet, with regard to acquisition of CRE, differences were very
low. Assuming these hazard ratios to be true, very long periods of
observation would have been needed to demonstrate significant, though still
small, differences. Moreover, intervention periods lasted long enough to
demonstrate a statistically significant difference in acquisition of quinolone-
resistance among CRE. Second, ciprofloxacin-resistance was only measured
in CRE-isolates collected during surveillance and not in Enterobacteriaceae
susceptible to third-generation cephalosporins. In addition, quinolone-
resistance in Enterobacteriaceae from clinical isolates was monitored to
identify potential rises in resistance, but not further analysed in this study.
Therefore, prevalence and acquisition rates of ciprofloxacin-resistant gram-
negatives are probably underestimated in this study. Third, bacterial
genotyping was only performed on CRE isolates from patients admitted
during baseline to determine the predominant acquisition route in order to
design an appropriate intervention. Genotyping of isolates collected during
the intervention period was not performed for all isolates, making it
impossible to determine the exact number of cross-transmission and thus
potential changes in predominant acquisition route.
Fourth, in the second phase of the intervention (after the crossover) an
outbreak caused by Enterobacter cloacae, displaying variable phenotypes, was
detected predominantly within the surgical department of our hospital.
Because of patient transfer from this department to the ICU, this strain was
also identified in nine patients (4 in ICU 1 and 5 in ICU 2). In eight of them
colonization appeared to be acquired in ICU, with epidemiological linkage
with another patient present in six (three in ICU 1 and three in ICU 2). Yet,
even when excluding the last weeks of the intervention periods (in which the
168
hospital outbreak was detected), colonization and infection control data
remained unchanged.
Despite these potential shortcomings, we conclude that the beneficial effects
of cycling or rotation of antibiotics on colonization rates of β-lactam resistant
pathogens have not been unequivocally demonstrated. Furthermore,
considering the disappointing results reported by others [9-13], even though
these studies suffered to some extent from methodological flaws, there is no
evidence to recommend antibiotic cycling or rotation as a strategy to control
antibiotic resistant pathogens.
169
References
1. Bonten MJ, Slaughter S, Ambergen AW et al. The role of "colonization pressure" in the spread of vancomycin- resistant enterococci: an important infection control variable. Arch Intern Med 1998:1127-32.
2. National Comittee for Clinical Laboratory Standards 2001. Performance standards for antimicrobial susceptibility testing. Fourteenth informational supplement. NCCLS document M100-S14. NCCLS, Wayne, PA. 2001.
3. Grundmann H, Barwolff S, Tami A et al. How many infections are caused by patient-to-patient transmission in intensive care units? Crit Care Med 2005:946-51.
4. Willems RJ, Top J, van den Braak N et al. Host specificity of vancomycin-resistant Enterococcus faecium. J Infect Dis 2000:816-23.
5. Sandoval C, Walter SD, McGeer A et al. Nursing home residents and Enterobacteriaceae resistant to third-generation cephalosporins. Emerg Infect Dis 2004:1050-5.
6. Kim PW, Harris AD, Roghmann MC et al. Epidemiological risk factors for isolation of ceftriaxone-resistant versus -susceptible Citrobacter freundii in hospitalized patients. Antimicrob Agents Chemother 2003:2882-7.
7. Bergstrom CT, Lo M, Lipsitch M. Ecological theory suggests that antimicrobial cycling will not reduce antimicrobial resistance in hospitals. Proc Natl Acad Sci USA 2004:13285-90.
8. Lipsitch M, Bergstrom CT, Levin BR. The epidemiology of antibiotic resistance in hospitals: paradoxes and prescriptions. Proc Natl Acad Sci USA 2000:1938-43.
9. Moss WJ, Beers MC, Johnson E et al. A pilot study of antibiotic cycling in a pediatric intensive care unit. Crit Care Med 2002:1877-82.
10. Warren DK, Hill HA, Merz LR et al. Cycling empirical antimicrobial agents to prevent emergence of antimicrobial-resistant Gram-negative bacteria among intensive care unit patients. Crit Care Med 2004:2450-6.
170
11. Martinez JA, Nicolas JM, Marco F et al. Comparison of antimicrobial cycling and mixing strategies in two medical intensive care units. Crit Care Med 2006:329-36.
12. Toltzis P, Dul MJ, Hoyen C et al. The effect of antibiotic rotation on colonization with antibiotic-resistant bacilli in a neonatal intensive care unit. Pediatrics 2002:707-11.
13. Van Loon HJ, Vriens MR, Fluit AC et al. Antibiotic rotation and development of gram-negative antibiotic resistance. Am J Respir Crit Care Med 2004:130-34.
14. Gruson D, Hilbert G, Vargas F et al. Rotation and restricted use of antibiotics in a medical intensive care unit. Impact on the incidence of ventilator-associated pneumonia caused by antibiotic-resistant gram-negative bacteria. Am J Respir Crit Care Med 2000:837-43.
15. Raymond DP, Pelletier SJ, Crabtree TD et al. Impact of a rotating empiric antibiotic schedule on infectious mortality in an intensive care unit. Crit Care Med 2001:1101-8.
16. Meyer E, Schwab F, Jonas D et al. Surveillance of antimicrobial use and antimicrobial resistance in intensive care units (SARI): 1. Antimicrobial use in German intensive care units. Intensive Care Med 2004:1089-96.
17. Kern WV, de With K, Steib-Bauert M et al. Antibiotic use in non-university regional acute care general hospitals in southwestern Germany, 2001-2002. Infection 2005:333-9.
18. Nijssen S, Bonten MJ, Franklin C et al. Relative risk of physicians and nurses to transmit pathogens in a medical intensive care unit. Arch Intern Med 2003:2785-6.
19. Pittet D, Mourouga P, Perneger TV. Compliance with handwashing in a teaching hospital. Infection Control Program. Ann Intern Med 1999:126-30.
20. Grundmann H, Hori S, Winter B et al. Risk factors for the transmission of methicillin-resistant Staphylococcus aureus in an adult intensive care unit: fitting a model to the data. J Infect Dis 2002:481-8.
171
21. McArdle FI, Lee RJ, Gibb AP et al. How much time is needed for hand hygiene in intensive care? A prospective trained observer study of rates of contact between healthcare workers and intensive care patients. J Hosp Infect 2006:304-10.
22. Kim PW, Roghmann MC, Perencevich EN et al. Rates of hand disinfection associated with glove use, patient isolation, and changes between exposure to various body sites. Am J Infect Control 2003:97-103.
23. Fridkin SK, Pear SM, Williamson TH et al. The role of understaffing in central venous catheter-associated bloodstream infections. Infect Control Hosp Epidemiol. 1996:150-8.
24. Vicca AF. Nursing staff workload as a determinant of methicillin-resistant Staphylococcus aureus spread in an adult intensive therapy unit. J Hosp Infect. 1999:109-13.
172
Chapter
General Discussion
Nederlandse Samenvatting
Dankwoord
Curriculum Vitae
List of Publications
General Discussion The time span needed for antibiotic resistance to emerge seems like seconds
compared to the time span needed to find suitable solutions to control
resistance. Despite the efforts of many to gather knowledge on microbial
genetics, microbial population dynamics, interactions between pathogen, host,
environment, and clinical epidemiology, there is yet not a ‘magic-bullet-
strategy’ to control antibiotic resistance. This thesis focused on the
determinants of colonization dynamics in intensive care units (ICUs) partly
by exploring existing knowledge on these determinants from scientific
literature and to identify knowledge gaps and partly by investigating
colonization dynamics of two key nosocomial pathogens to close, at least,
some of these gaps.
From the literature analysis, described in chapter 1, we conclude that
colonization dynamics in ICUs are a function of pathogen-related and
healthcare worker-related determinants, all interacting with each other, and
thereby, increasing complexity. When designing (and executing) strategies to
control emergence and spread of antibiotic resistance in such a setting, routes
of acquisition (endogenous or exogenous), antibiotic use and different
infection control parameters should be measured, and these determinants
should, preferably, not change when they are not the subject of intervention.
Ideally, a single intervention is implemented at a time, while maintaining all
other relevant variables unchanged or being carefully determined, so the
effects of the intervention can be evaluated using methods identical to
baseline. When relevant variables do not remain unchanged, careful
determination will allow subsequent adjustment in statistical analysis. As an
illustration, we reviewed the studies that used modification of antibiotic
176
strategies in the ICU as a measure to reduce antibiotic resistance (e.g.,
antibiotic cycling or rotation). After reviewing how these studies fulfilled to
these important methodological aspects (chapter 1), we had to conclude that
the majority of studies in fact, to variable extents, do not. Consequently, the
validity of conclusions drawn from such studies seems rather limited.
To further explore the determinants of colonization and their interactions, we
first studied the effects of microbiological surveillance (without feedback of
results to the ICU or subsequent isolation of colonized patients) as a single
intervention on the spread of methicillin-sensitive and methicillin-resistant
Staphylococcus aureus (MSSA and MRSA) in an ICU where both were
endemic (chapter 2). During a ten-week period, after obtaining 1216
surveillance cultures and genotyping 142 isolates, two out of 158 patients
admitted to this ICU acquired colonization with MSSA and none acquired
MRSA. Based on bacterial genotyping results, both cases of acquired
colonization appeared not to result from cross-transmission, despite
continuous presence and admission of patients colonized with these
pathogens. This at first glance rather awkward intervention – performing
surveillance without providing feedback of results – had never been evaluated.
Yet, if this study had been executed with feedback of results and isolation of
colonized patients, this would certainly have been interpreted as a successful
intervention for bacterial transmission. These findings underscore the need of
more carefully designed studies to evaluate the true contribution of individual
measures of infection control, before generally accepting that a package of
measures is necessary.
The risk of transmission is, probably, influenced by the bacterial load of a
patient, contact rates of healthcare workers, bacterial contamination of hands
and devices after contact, the extent to which different patients are contacted
177
by healthcare workers and adherence to and efficacy of hand disinfection.
Based upon multiple observational studies it is generally believed that
physicians adhere less to hand disinfection rules than nurses. Yet, hand
disinfection, though probably important, is just one of the variables in a
cascade of events leading to (or better stated preventing) cross-transmission.
One might simply argue that nurses spent most of their time on patient care,
thus probably have more patient contacts and thus are more at risk to
transmit pathogens from one patient to another, as compared to physicians.
On the other hand, nurses better adhere to hand disinfection rules and might
contact fewer patients, than physicians. The question ‘who is the main vector
in pathogen transmission’ is interesting for obvious reasons, but may also
guide to whom education on infection control practices should be focused. In
chapter 3, the relative risk for nurses and physicians to transmit pathogens in
a medical ICU was estimated using contact rates, cohorting levels and hand
hygiene adherence. A contact between two different patients, defined as a
potentially contaminated contact, was used as proxy for actual pathogen
transmission. Despite lower contact rates, physicians, as a group, had a 1.6
times higher risk to transmit pathogens in this ICU than nurses, mainly
because of their lower cohort level and lower adherence to hand disinfection
(43% as compared to 59% for nurses). In fact, an adherence rate to hand
disinfection of 64% would be needed to counterbalance the lower cohorting,
which is inevitable for physicians as they need to see all patients in the unit.
This observational study, once again, shows the complexity of interactions
between variables determining transmission risk, and that this risk is not
reflected by the amount of time spent on direct patient care.
In the design (and analysis) of effective infection prevention strategies, the
route of colonization is an important feature. So-called ‘endogenous’
178
colonization (i.e. within-host mutation, within-host horizontal gene transfer
and/or subsequent selection of bacteria previously present below detection
limits) is mainly driven by antimicrobial use, but does not depend on the
colonization status of other patients (colonization pressure). In contrast, so-
called ‘exogenous’ colonization (i.e. cross-transmission) results from failures
in infection control and highly depends on colonization pressure: cross-
transmission is not possible when there are no other patients in the unit with
a certain pathogen. The underlying dynamics of these processes are
fundamentally different (see chapters 1, 5). As a result of these dynamical
differences, strategies needed to prevent either of these routes, are also
different. Yet, antimicrobial use and adherence to infection control are
important issues for both routes, but their relative importance may differ. For
instance, improved adherence to hand hygiene is unlikely to have much of an
effect on pathogen acquisition, when acquisition occurs predominantly
through endogenous selection. These dynamical differences in acquisition
and spread of bacterial resistance within hospital settings have become a kind
of ‘red line’ through this thesis, and are addressed in chapters 1, 4, 5, and 8.
Antibiotic resistance can spread by various modes: through cross-
transmission of complete pathogens or through transfer of genetic resistance
determinants (within and between species). In chapter 4, the contribution of
these various transfer modes was studied in Enterobacteriaceae with reduced
susceptibility to third-generation cephalosporins (ERSC) (selected on
screening media supplemented with 8 μg/ml cefpodoxime). Genotyping was
used to determine strain-relatedness and presence of mobile genetic elements
conferring resistance to antimicrobial agents (i.e., integrons). In an ICU
setting characterized by low levels of antibiotic resistance – as compared to
international standards - 121 (27%) of 457 admitted patients were colonized
179
with ERSC and integrons were detected in 34 isolates of 31 patients. Nine
integron clusters could be distinguished, and integrons were associated with
resistance to multiple antibiotics. Twenty-four patients acquired integron-
carrying isolates (26 isolates) during ICU-stay and based on genotyping of
integrons, cross-transmission of integron-carrying species had occurred 19
times (73%), of which 10 between epidemiological linked patients (38%). In
two instances inter-species and in one case intra-species transfer of integrons
was demonstrated. The relative contribution of the different transmission
routes for integrons, therefore, was between 38% and 73% for cross-
transmission, 8 % for inter-species transfer and 4% for intra-species transfer.
The remaining cases were acquired endogenously.
These findings underscore the relevance of surveillance, genotyping and
analysis of resistance determinants - like integrons - for better understanding
of antibiotic resistance epidemiology. Without this combined approach, the
situation in our unit would have been described as low-level endemicity of
ERSC with few circulating bacterial types. In fact, multiple, though obviously
unrecognized outbreaks, occurred, mainly through cross-transmission of
complete organisms rather than by horizontal gene transfer.
Bacterial genotyping is the gold standard to determine clonality of isolates
and is, therefore, needed to demonstrate the occurrence of cross-transmission.
However, typing methods are time-consuming, labour-intensive and costly.
Based on the differences between linear (such as endogenous selection) and
non-linear processes (such as cross-transmission), Pelupessy et al. has
proposed to use a Markov model to predict the relative importance of both
acquisition routes [1]. With this approach, the absolute rates of both
acquisition routes are estimated from longitudinal surveillance data, without
180
the need of bacterial genotyping. The initial model, as described by Pelupessy
et al., had several limitations: endogenous selection and admission of
colonized patients could not be distinguished, full occupancy of the ward was
assumed and periods of uncertain colonization status (between a negative and
the first positive culture) were, artificially, recoded to either non-colonized or
colonized. In chapter 5, the accuracy of a modified Markov model to
estimate the predominant acquisition route of third-generation
cephalosporin-resistant Enterobacteriaceae (CRE) in two intensive care units
ICUs was evaluated. The adaptations, to be described in detail by Bootsma et
al. (in preparation), are
• that admission rates are explicitly distinguished from endogenous
selection rates
• that actual changes in bed occupancy are used
• that there is no need to assume that length of stay is exponentially
distributed
• that the moments cultures are performed and the results of these
cultures are the bookkeeping cornerstone of the model while a
stochastic model estimates the status of patients in-between culture
sampling moments
So the model formulation is data driven from the very beginning and
incorporates all the information that is available.
Model predictions of the mean daily endemic prevalence and proportions of
endogenous and exogenous colonization were estimated upon admission and
discharge dates and culture results on subsequent dates during ICU-stay.
Using bacterial genotyping by Amplified Fragment-Length Polymorphism
181
(AFLP) and epidemiological linkage as reference, the Markov chain model
accurately quantified acquisition routes of colonization with CRE in two
ICUs and correctly established predominance of endogenous over exogenous
acquisition. This method, therefore, seems a promising tool to provide
essential information on the dynamics of microorganisms in hospital settings,
without requiring any labour-intensive and costly genotyping procedures.
As resistance to β-lactam antibiotics in Enterobacteriaceae is rapidly
increasing worldwide, we investigated the prevalence of β-lactam resistance in
Enterobacteriaceae in Europe (chapter 6, 7).
Extended-Spectrum Beta-Lactamases (ESBLs) are plasmid-enocoded
enzymes that can hydrolyze β-lactams, including third-generation
cephalosporins. ESBLs have been described mainly in Escherichia coli and
Klebsiella pneumoniae, but are found in other members of Enterobacteriaceae as
well. The detection of ESBLs in bacterial isolates is not straightforward, as
the Minimal Inhibitory Concentration (MIC) for cefotaxime and ceftriaxone
in such isolates may be below breakpoints for susceptibility defined by the
National Committee for Clinical Laboratory Standards (NCCLS) [2]. In
addition, resistance to cephalosporins may be mediated by other resistance
mechanisms than ESBL-production. Growth inhibition in the presence of a
β-lactamase inhibitor, like clavulanic acid, is indicative for ESBL-production.
Enterobacteriaceae with a MIC for aztreonam, ceftriaxone or ceftazidime
≥1 μg/ml are potential ESBL-producers (NCCLS) and such isolates require
further testing. Automatic diagnostic systems are not always reliable for the
detection of ESBLs. As prevalence levels of ESBL-mediated resistance are
unknown in most hospitals, we determined ESBL-production in European
Enterobacteriaceae as well (chapters 6, 7).
182
E-test ESBL and DDT are most commonly used, however, their accuracy
remains debated. The DDT is inexpensive and easy to use. The E-test is
more expensive and requires more time for interpretation of results. In
chapter 6, we compared and evaluated the performance of both tests in 404
isolates displaying the potential ESBL-phenotype. DDT identified 227
isolates (56%) as ESBL-producers, as compared to 205 isolates (51%) by E-
test, with a 69% concordance between both tests. In 1.5% either DDT or E-
test yielded a positive test result whereas the other yielded a negative test
result. All together, ESBL-production was identified in 233 of 404 isolates
(57%) displaying a potential ESBL-phenotype by either of these tests. In
DDT, ESBL-detection was comparable for ceftriaxone (91%) and
ceftazidime (92%) and the addition of aztreonam did not improve diagnostic
accuracy. Molecular identification of ESBLs was not performed, thus, the
actual proportion of ESBLs remains unknown. A practical problem with E-
test is that results often cannot be interpreted when inhibition zones grow
beyond the scale range of the E-test strip. In 5.4%, DDT identified ESBL-
production where E-test results could not be interpreted. Based on these
results, and taking user friendliness and costs of both tests into account as
well, double disk diffusion test is preferred above E-test for ESBL-detection
(chapter 6).
Susceptibilities to 15 different β-lactam antibiotics were determined for 5000
Enterobacteriaceae collected in 25 European hospitals between 1997-1998 as
part of the SENTRY Antimicrobial Surveillance Program (chapter 7). The
main findings were that, in general, the majority of Enterobacteriaceae in
Europe were susceptible to piperacillin-tazobactam (>78%) and third-
generation cephalosoprins (>76%). These antibiotics are frequently used as
first choice for empirical therapy of nosocomial infections caused by
183
Enterobacteriaceae. ESBL-production was found in 233 of 4707 isolates
(4.9%), with large regional differences in prevalence of ESBLs; prevalence
was highest in Southern European countries (>13%), as compared to <9% in
most other European countries. As susceptibility rates for cefepime and
carbapenems in Enterobacteriaceae (including ESBL-producing isolates) were
above 95% and 99%, these antimicrobials remain useful alternatives to treat
nosocomial infections caused by pathogens with lower susceptibility rates or
producing ESBLs. Importantly, though, results from local surveillance are
most indicative for the choice of appropriate therapy.
From chapters 6 and 7 we can conclude that the detection of ESBLs is
problematic and that, with all diagnostic problems, the prevalence in
European isolates was low (5.4%). We, therefore, decided to focus on
resistance to third-generation cephalosporins in Enterobacteriaceae (CRE),
without specific determination of the underlying resistance mechanism,
instead of resistance resulting from ESBL-production alone (chapter 8). The
aim of the study, described in chapter 8, was twofold: first, to determine
colonization dynamics of third-generation cephalosporin-resistant
Enterobacteriaceae (CRE) in two ICUs (during a non-outbreak, baseline
period) and second to evaluate the effects of a single intervention on
acquisition of colonization with CRE, with monitoring of all variables
relevant for colonization dynamics. In the baseline-period, acquisition of
CRE was predominantly endogenous, with only few cases resulting from
cross-transmission (AFLP), and with the use of amoxicillin-clavulanic acid as
the most important modifiable risk factor for acquisition. Therefore, a
heterogeneous and homogeneous regimen, to reduce exposure of patients to
the most commonly used β-lactams (amoxicllin-clavulanate and ceftriaxone),
were implemented in a randomized crossover study design.
184
The heterogeneous regimen (3 months) consisted of weekly changes in
empirical therapy from ceftriaxone, to amoxicillin-clavulanic acid, to
quinolones (levofloxacin or ciprofloxacin) and back to ceftriaxone in the
fourth week. During the homogeneous (3 months) antibiotic regimen,
levofloxacin (or ciprofloxacin) was the first choice of empirical therapy. A
subsequent step-wise reduction of amoxicillin-clavulanic acid use and
ceftriaxone use at costs of increased usage of quinolones (from baseline, to
heterogeneous to homogeneous), failed to reduce CRE-acquisition (HR
heterogeneous regimen: 1.0, 95% CI: 0.5-2.2; p=.95 and HR homogeneous
regimen: 1.1, 95% CI: 0.5-2.5; p=.69), but facilitated a dramatic increase in
ciprofloxacin-resistant CRE ( HR heterogeneous regimen: 1.6, 95% CI:0.4-6.5;
p=.50 and HR homogeneous regimen: 4.1 (95% CI: 1.4-11.9; p<.01). Only
contact rates apparently changed during the study periods (with lowest rates
during heterogeneous), which was accounted for in Cox’ regression analysis.
When comparing antibiotic use, contact rates, cohorting levels and adherence
to hand hygiene (all potential confounders) to reported data from the
international scientific literature, our findings all appeared to be comparable,
suggesting that our observations can be generalized to at least some extent.
The findings of this study demonstrate that a straightforward reduction of β-
lactam exposure, with control of all relevant confounders, does not reduce
acquisition of β-lactam-resistant gram-negatives.
There are some potential limitations of this study that should be discussed
briefly. Prevalence and incidence of CRE colonization were low in these
ICUs, which reduces the statistical power to demonstrate effects of an
intervention on acquisition rates. However, considering the low hazard ratios
for CRE in the study periods, even with prolonged duration these hazard
ratios would not reach statistical significance. Moreover, the effects of
185
quinolones on ciprofloxacin-resistant CRE (with even lower acquisition rates
during baseline) were highly significant in this short study period. Secondly,
genotyping was not performed on isolates collected during the intervention
period and a shift in dominance of acquisition routes could only be estimated
by potential cases of cross-transmission based on species determination and
epidemiological linkage between patients, leaving insecurity about the exact
number of cases.
Despite these potential shortcomings, we conclude that the beneficial effects
of cycling or rotation of antibiotics on colonization rates of β-lactam-resistant
pathogens have not been unequivocally demonstrated. Furthermore,
considering the disappointing results reported by others [3-7], even though
these studies suffered to some extent from methodological flaws, such as a
lack of baseline measurement, insufficient (or absent) control of confounding
and simultaneous implementation of more than one intervention, there is no
evidence to recommend antibiotic cycling or rotation as a strategy to control
antibiotic resistant pathogens.
Future perspectives
So, at the end of this thesis: where should we go from here? One thing that
must be obvious now, is that the question on how to control antimicrobial
resistance is difficult, but not necessarily impossible, to be answered.
The multi-disciplinary approach
Well-designed studies combining molecular biology to investigate genetic
resistance determinants, population biology and clinical epidemiology are
needed to obtain reliable data on resistance epidemiology and the effects of
186
interventions. The ideal design for such studies would be a cluster-
randomized controlled trial, as this would offer the assurance that
confounding is avoided. Quasi-experimental studies can be a good - and
more feasible - alternative, as long as confounding is adequately controlled.
The latter can only be achieved with extensive microbiological surveillance,
bacterial genotyping, monitoring antibiotic usage and observation of infection
control practices.
Microbiological surveillance
In many studies infection rates rather than colonization rates have been used
as end-points of analyses. As infection rates only represent the tip of the
‘resistance-iceberg’, only a fraction of dynamical changes will be detected by
infection rates. Especially studies reporting no differences in infection
outcome, therefore, may have been underpowered to detect that difference,
even though significant changes may have occurred on colonization level.
Therefore, colonization rates instead of infection rates should be determined
in intervention studies in order to evaluate the true effects of an intervention
on resistance dynamics.
Quantification of different colonization routes, by bacterial genotyping and
epidemiological linkage is not performed regularly in intervention studies. As
optimal efficacy of controlling antibiotic resistance can only be achieved by
targeting the predominant colonization routes, such knowledge would be
helpful in designing future intervention studies. Reality, though, is that
bacterial genotyping is currently not available on a real-time basis. Yet,
technology now allows such determination, but the associated costs of these
new techniques will probably be an important barrier in the coming years.
187
The approach of mathematical modeling, as introduced in this thesis, might
be of clinical use for this purpose.
Mathematical modeling
Because of the complexity of bacterial dynamics, especially in hospital
settings, mathematical modeling might offer major advantages above more
classical epidemiological and statistical methods for data analysis and
determination of bacterial transmission routes. The concept of patient-
dependency has, up till now, been almost completely neglected in hospital
epidemiology. The Markov model, which explicitly takes patient-dependency
into account, is a first example of this approach. Initially proposed by
Pelupessy et al. [1] and further adapted by Bootsma et al. (in preparation),
chapter 5 of this thesis is its first prospective validation. Currently, more
validation studies are being conducted. Moreover, mathematicians are now
working on further improvements of the model and techniques to use this
model for statistical analysis. Ultimately, this model might be used as a real-
time continuous monitoring system of antibiotic resistance in hospitals. With
continuous addition of culture results, estimates of transmission dynamics
will be provided and changes herein, such as increases in cross-transmission,
might become visible, without genotyping.
Infection control strategies
For designing highly effective control strategies, we propose that future
studies should include determination of the relative importance of different
acquisition routes, before introducing control measures. For instance,
acquisition of MRSA does not occur by de novo resistance development in
188
an individual patient, but usually results from patient-to-patient transmission.
Moreover, high prevalence rates may also result from high admission rates of
MRSA-positive patients. In contrast, ciprofloxacin-resistance may well
develop during antimicrobial therapy within individual patients. When
endogenous colonization is dominant, with low rates of cross-transmission,
changes in antimicrobial use are probably more effective than increasing
adherence to hand hygiene or introducing other barrier precautions. However,
the previous sentence should not be interpreted as downplaying the relevance
of appropriate hand hygiene.
Infection control variables like contact rates, cohorting levels and adherence
to hand hygiene - extremely important determinants of colonization dynamics
- have not been measured extensively. Yet, quantitative information of these
variables may provide useful information about the local status of infection
control and, thus, to what extent these measures need to or can be improved
(chapter 3). Adherence to hand hygiene in ICUs seldom exceeds 50% and
changing behavior of healthcare workers, in this regard, has been proven
difficult, especially for prolonged periods of time. Interestingly, cohorting
levels of nurses were between 70% and 80% in all studies in which this
variable has been determined. This seems a rather high value, especially
considering staffing problems in many ICUs. Therefore, more data -
especially from ICUs considered to be understaffed - are needed. With
cohorting levels as high as 70-80%, it is questionable whether further
improvement of this cohorting level will have much of an effect on
colonization dynamics, especially in non-outbreak situations. During
outbreaks, though, temporary cohorting levels close to 100% - resembling
one-to-one nursing - could be beneficial (or be necessary) as transmission is
almost impossible to occur when no other patients are contacted.
189
From the above we conclude that determination of infection control variables
is useful:
• to determine the variable with the highest potential for successful
modulation of colonization dynamics
• to control for confounding during intervention studies
• to target infection control strategies on proportional importance of
colonization routes
Antibiotic usage
‘Antibiotic prescription should be appropriate, prudent and restrictive’, has
been frequently stated. How can we be against prudent use of anything?
The statement means that antimicrobial therapy should be based on a clear
indication, is appropriate for the suspected pathogen, is adequately dosed and
is adjusted according to culture results (taken before start of therapy), if
possible. Local microbiological surveillance and susceptibility testing provides
fundamental information for appropriate antimicrobial prescription.
Considering antimicrobial regimens proposed for the control of antibiotic
resistance, this thesis does not support the use of antimicrobial cycling or
rotation as a solution to control antimicrobial resistance. Restriction of a
specific antimicrobial class - during long periods or during cycling - usually
induces more extensive use of overuse (an)other class(es) with subsequent
resistance to these agents, as the total antibiotic prescription does not change.
This has been called ‘squeezing the balloon of resistance’. In one of our
studies, reductions of 91% and 59% in amoxicillin-clavulanic acid and
ceftriaxone did not have an effect on acquisition rates of third-generation
190
cephalosporin-resistant Enterobacteriaceae, but a 243% increased use of
quinolones was associated with a significant increase in the prevalence of
resistance to this class, within a short period of time. This occurred, while
controlling all potential confounders. Quinolone-resistance has been known
to rise from point-mutations and higher levels of resistance are reached under
selective pressure. Taking all this into account, homogenous regimens with
quinolones as a first choice of empirical therapy, is not recommended. In fact,
antibiotic mixing, thereby minimizing homogeneous antibiotic exposure,
might be the preferred strategy of antibiotic use in ICUs.
191
Nederlandse Samenvatting Dit proefschrift beschrijft de epidemiologie van antibioticaresistente bacteriën
in ziekenhuizen en in het bijzonder op intensive care-afdelingen (IC’s). Deze
bacteriën reageren niet of onvoldoende op antibiotische therapie en vormen
dus een bedreiging voor de behandeling van patiënten met infecties
veroorzaakt door deze bacteriën. Patiënten die in een ziekenhuis worden
opgenomen lopen het risico om drager te worden van antibioticaresistente
bacteriën, een fenomeen wat kolonisatie wordt genoemd. Niet alle patiënten
die gekoloniseerd raken met antibioticaresistente bacteriën krijgen ook
daadwerkelijk een infectie, dit is slechts het topje van de ijsberg. Gezien dit
feit is de prevalentie van kolonisatie van patiënten met deze bacteriën en de
verspreiding ervan op ziekenhuisafdelingen niet direct zichtbaar en niet
eenvoudig te controleren.
Vele factoren dragen bij aan de ontwikkeling en verspreiding van
antibioticaresistentie en de onderlinge interacties tussen deze factoren maken
de kolonisatiedynamiek zeer complex. Teneinde de ontwikkeling en
verspreiding van resistentie te beperken met de juiste maatregelen, is kennis
van de onderliggende bacteriële genetica, bacteriële populatiedynamiek,
interacties tussen bacteriën, gasheer en omgeving en kennis van de klinische
epidemiologie, onmisbaar.
In dit proefschrift ligt de nadruk op het beschrijven van de verschillende
factoren betrokken bij de kolonisatie met antibioticaresistente bacteriën op
IC’s. Hiervoor hebben we gebruik gemaakt van zowel het bestuderen van de
bestaande literatuur op dit gebied, als van studies naar de kolonisatiedynamiek
193
van Staphylococcus aureus en Enterobacteriaceae, met als doel om gaten in de
huidige kennis te achterhalen en deze waar mogelijk aan te vullen.
Uit de huidige literatuur blijkt dat de kolonisatiedynamiek op IC’s vooral
afhankelijk is van de complexe interacties tussen bacterie-gerelateerde en
medewerker-gerelateerde eigenschappen, hoofdstuk 1. Patiënten kunnen
drager worden doordat al aanwezige bacteriën veranderingen in het DNA
ondergaan tijdens antibioticumgebruik (mutatie) en vervolgens wordt
uitgeselecteerd. Antibioticaresistente bacteriën kunnen ook bij het starten van
antibiotische therapie al in kleine hoeveelheden aanwezig zijn in het lichaam
bij opname (introductie) en vervolgens verder worden uitgeselecteerd.
Antibiotische selectie is in dit proefschrift gedefinieerd als endogene
kolonisatie. Tevens kunnen de handen van ziekenhuismedewerkers
(verpleegkundigen en artsen) tijdelijk gekoloniseerd raken op die manier van
patiënt naar patiënt worden overbracht. Dit proces wordt exogene kolonisatie
of kruistransmissie genoemd en is over het algemeen het gevolg van het falen
van infectiepreventiemaatregelen; Figuur 1, hoofdstuk 1.
Contact rates, cohorting (de mate van één op één verpleging) en de naleving
van handhygiëne voorschriften zijn belangrijke factoren die in de cascade die
kan leiden tot kruistransmissie (of het kunnen voorkomen). Kennis van deze
factoren en van de belangrijkste kolonisatieroute (endogeen of
kruistransmissie) en antibioticumgebruik is onmisbaar voor het
implementeren van de juiste controlestrategieën ter preventie van
verspreiding van antibiotica- resistentie. Vooral in (interventie)studies is het
daarom van groot belang dat alle determinanten zorgvuldig worden gemeten
voor (baselineperiode) en na het implementeren van een interventie, om het
werkelijke effect van de interventie te bepalen. Tevens is de meest ideale
194
situatie dat slechts een van de determinanten wordt aangepast en het effect
ervan te bepalen, terwijl alle anderen gelijk blijven of in ieder geval zorgvuldig
worden gemeten, hoofdstuk 1.
Om dit illustreren, hebben we een beschouwing gemaakt van de mate waarin
bovengenoemde methodologische aspecten aanwezig waren in 19 studies die
het effect van een interventie in het antibioticumgebruik op de
kolonisatiedynamiek op IC’s hebben onderzocht. Deze beschouwing laat zien
dat de meerderheid van de studies slechts gedeeltelijk voldoen aan de
eerdergenoemde methodologische voorwaarden voor interventiestudies. Als
gevolg hiervan lijkt de validiteit van de conclusies getrokken in deze studies,
beperkt (hoofdstuk 1).
In hoofdstuk 2 wordt de noodzaak van isolatiemaatregelen op geleide van
kweekuitslagen in de preventie van kruistransmissie besproken. Het effect
van microbiologische surveillance als enige interventie (dus zonder
rapportage van de kweekuitslagen en daaropvolgende isolatie) op kolonisatie
met methicilline-gevoelige en methicilline-resistente Staphylococcus aureus
(MSSA en MRSA) werd gedurende 10 weken bestudeerd in een IC in Cook
County Hospital, Chicago. Van alle met MSSA of MRSA gekoloniseerde
patiënten, verwierven slechts twee patiënten kolonisatie tijdens verblijf op de
IC (MSSA). In beide gevallen bleek na genotypering dat dit niet het gevolg
was van kruistransmissie.
Rapportage met daaropvolgende isolatie, zoals geadviseerd in vele richtlijnen,
zou in deze setting onterecht tot de conclusie hebben kunnen leiden dat zij
succesvol waren in het voorkomen van kruistransmissie terwijl er in
werkelijkheid zonder deze maatregelen geen kruistransmissie plaatsvond. Het
195
baseren van richtlijnen op basis van de locale epidemiologie geniet
waarschijnlijk de voorkeur.
Aangezien er een complexe interactie bestaat tussen contact rates, cohorting
en handhygiëne (hoofdstuk 1) en de kwantitatieve waarde van de individuele
factoren verschillend is voor artsen en verpleegkundigen, is het moeilijk te
bepalen wie de belangrijkste vector is in de overdracht van
antibioticaresistente bacteriën. Verpleegkundigen hebben waarschijnlijk meer
lichamelijke contacten met patiënten dan artsen en dus meer kans op tijdelijke
kolonisatie van hun handen. Over het algemeen hebben zij contact met
minder verschillende patiënten gedurende hun diensten (hogere
cohortingsgraad) dan artsen. Daarnaast houden verpleegkundigen zich over
het algemeen ook beter aan handhygiëne voorschriften. In hoofdstuk 3
hebben we bovengenoemde factoren zorgvuldig gemeten door middel van
observaties van patiënten en personeel en vervolgens het risico op
kruistransmissie voor zowel artsen als verpleegkundigen berekend. Hieruit
bleek dat artsen, op deze IC, een 1.6 keer grotere kans hadden om bacteriën
te verspreiden dan verpleegkundigen, mede door het grotere aantal patiënten
waarmee artsen contact hadden en doordat zij zich minder goed hielden aan
de handhygiëne voorschriften (43% vs. 59%). Uit deze observationele studie
blijkt dus eens te meer de complexiteit van de interacties tussen de
determinanten van kolonisatie.
Overdracht van antibioticaresistentie kan via verschillende wegen
plaatsvinden: de overdracht van de gehele bacterie (kruistransmissie) of
slechts van de elementen die resistentiegenen bevatten (horizontale
overdracht). Dit laatste gebeurt zowel tussen bacteriën behorend tot dezelfde
196
soort (intraspecies) maar ook tussen verschillende soorten (interspecies).
Integronen zijn een voorbeeld van mobiele genetische elementen die
resistentiegenen bevatten en van de ene bacterie naar de andere kunnen
worden overgedragen. In hoofdstuk 4 hebben we de prevalentie van
integronen in en de bijdrage van kruistransmissie en horizontale overdracht
binnen verschillende leden van de Enterobacteriaceae met verminderde
gevoeligheid voor cefalosporinen (ERSC) bepaald op twee IC’s. Hiertoe
werden de resultaten van surveillance, genotypering (AFLP) in combinatie
met ‘epidemilogische linkage’ en integronenanalyse aangewend.
Tijdens de studieperiode waren 121 van de 457 patiënten gekoloniseerd met
ERSC (27%) en in 34 isolaten van 31 patiënten werden integronen
aangetoond. Deze konden worden onderverdeeld in negen verschillende
epidemiologische clusters. Overdracht van de gehele bacterie
(kruistransmissie) was verantwoordelijk voor 38% tot 73% van de gevallen
van resistentieoverdracht, intraspecies overdracht voor 8% en interspecies
voor 4%. Op basis van surveillance en genotypering hadden we deze setting
qua ERSC gekarakteriseerd als een low-level endemisch met slechts enkele
circulerende clonen, maar na het combineren met integronenanalyse bleek er
sprake te zijn van meerdere mini-epidemieën. Hoewel overdracht van
resistentie vooral door kruistransmissie plaatsvond, waren niet alle gevallen
aantoonbaar met genotypering alleen. Een gecombineerde aanpak is dus
zeker relevant voor het beschrijven van resistentie-epidemiologie.
Bacteriële genotypering wordt tot op heden beschouwd als de gouden
standaard voor het aantonen van verwantschap van bacteriële isolaten. Het
nadeel van deze methode is echter dat het tijdrovend, arbeidsintensief en
kostbaar is. Recent is er een Markov model ontwikkeld waarmee, op basis van
197
verschillen tussen lineaire processen (zoals endogene selectie) en non-lineaire
processen (zoals kruistransmissie), kan worden voorspeld wat het aandeel van
beide routes van kolonisatie is in een bepaalde setting zonder de noodzaak
van genotypering.
In hoofdstuk 5, evalueren wij de accuraatheid van een aangepast Markov
model voor het schatten van het aandeel endogene selectie en
kruistransmissie in kolonisatie met derdegeneratie cefalosporinen-resistente
Enterobacteriaceae (CRE) op twee IC’s. Op basis van opname- en
ontslagdata en longitudinale surveillance werden de proporties endogene
selectie en kruistransmissie eerst door het model geschat en vervolgens
vergeleken met die berekend met behulp van de gouden standaard
(surveillance en genotypering).
De bijdrage van beide routes aan kolonisatie met CRE op deze twee IC’s
werd door het model accuraat gekwantificeerd, met endogene selectie als de
belangrijkste route, en dus lijkt deze methode veelbelovend voor het
analyseren van kolonisatiedynamiek in de ziekenhuissetting zonder de
noodzaak van een kostbare en tijdrovende techniek als genotypering.
Verschillende mechanismen kunnen leiden tot resistentie tegen β-lactams
waaronder de productie van Extended-spectrum Bèta-lactamases, oftewel
ESBLs. Dit zijn plasmide-gecodeerde enzymen die β-lactams, inclusief
derdegeneratie cefalosporinen, kunnen afbreken. Binnen de
Enterobacteriaceae zijn zij veelvuldig beschreven in Escherichia coli en Klebsiella
pneumoniae, maar worden ook geproduceerd door andere leden van deze groep.
Het aantonen van ESBLs blijkt tot op heden geen eenvoudige taak. De twee
meest gebruikte methoden zijn de double disk diffusion test (DDT) en de
ESBL E-test. De eerste is simpel in gebruik en goedkoop, de tweede is meer
198
tijdsintensief (m.n. de interpretatie van de resultaten) en duurder. In
hoofdstuk 6 vergelijken we de accuraatheid van beide testen in het aantonen
van ESBLs in 404 potentieel ESBL-producerende Enterobactriaceae-isolaten.
DDT identificeerde 56% en E-test 51% van de isolaten als ESBL-
producerend, met een concordantie van 69%. Er werd geen moleculaire
identificatie uitgevoerd en dus blijft de werkelijke prevalentie van ESBLs in
deze populatie onbekend. De interpretatie van de resultaten met E-test bleek
vaak niet mogelijk, een praktisch probleem dat ontbreekt bij DDT en dus lijkt
de laatste, zeker ook met het oog op de gebruikersvriendelijkheid en de
kosten, de voorkeur te hebben voor het identificeren van ESBLs.
Hoofdstuk 7 is een beschrijving van de prevalentie van β-lactam resistentie
in Enterobactericeae in Europa. De gevoeligheid voor 15 verschillende β-
lactam antibiotica werd bepaald in 5000 isolaten (Escherichia coli, Klebsiella
pneumoniae, Klebsiella oxytoca, Enterobacter cloacae en Proteus mirabilis) afkomstig
uit 25 Europese ziekenhuizen in 1997-1998.
Uit deze analyse blijkt dat de meerderheid van de Enterobacteriaceae in
Europa nog goed gevoelig zijn voor piperacilline-tazobactam (>78%) en
derdegeneratie cefalosporinen (>76%), die frequent als eerste keus empirische
therapie voor ziekenhuisinfecties veroorzaakt door deze groep, worden
gebruikt. De productie van ESBLs werd bepaald zoals beschreven in
hoofdstuk 6 en de prevalentie hiervan lag rond de 4.9%. Echter, er bestaan
grote regionale verschillen binnen Europa en voornamelijk de landen in Zuid-
Europa laten hogere prevalenties zien dan de landen in Noord- en West-
Europa (>13% versus <9%).De gevoeligheden van Enterobacteriaceae
(inclusief ESBL-producerende isolaten) voor cefepime en carbapenems lagen
rond de 95% en 99%, wat deze antibiotica zeer bruikbare alternatieven maakt
in de behandeling van infecties veroorzaakt door Enterobacteriaceae met een
199
verlaagde gevoeligheid voor de eerdergenoemde middelen. Hoewel deze
analyse een aardig beeld geeft van de situatie in Europa, is het regionale
resistentiepatroon van bovengenoemde pathogenen meest indicatief voor de
keuze van de juiste therapie.
In hoofdstuk 8 beschrijven we de resultaten van een studie met een
tweeledig doel: ten eerste het bepalen van de kolonisatiedynamiek van
derdegeneratie cefalosporine-resistente Enterobacteriaceae (CRE) op twee
intensive care-afdelingen in het UMC Utrecht (baseline periode) en ten
tweede het effect van een interventie in het antibioticumgebruik op het
verwerven van kolonisatie met CRE (interventieperiode).
Tijdens de baselineperiode bleek kolonisatie voornamelijk via endogene
selectie verworven te zijn met als belangrijkste risicofactor het gebruik van
amoxicilline-clavulaanzuur. De geïmplementeerde interventie had als doel de
blootstelling aan β-lactam antibiotica in het algemeen en aan amoxicilline-
clavulaanzuur in het bijzonder, te verminderen. Alle relevante determinanten
van kolonisatie werden zorgvuldig gecontroleerd en gemeten. Een heterogeen
(3 maanden) en een homogeen (3 maanden) restrictiebeleid werden in een
gerandomiseerd cross-over design geïmplementeerd. Tijdens het heterogene
beleid was de eerste keus voor empirische therapie in week 1 ceftriaxon, in
week 2 amoxicilline-clavulaanzuur en week 3 een quinolon. Na drie weken
begon deze cyclus weer opnieuw. Tijdens het homogene beleid was de eerste
keus een quinolon. Een stapsgewijze vermindering in het gebruik van
amoxicilline-clavulaanzuur en ceftriaxon, ten koste van een toename in het
quinolon-gebruik maar met gelijktijdige controle van alle relevante
confounders, leidde niet tot een vermindering van het aantal verworven
gevallen van kolonisatie met CRE. Echter, het toegenomen gebruik van
200
quinolonen resulteerde wel in een aanzienlijke toename in ciprofloxacine-
resistente CRE-isolaten.
In de literatuur zijn de resultaten van studies die het effect van antibiotische
cycling en rotatie designs op kolonisatiedynamiek bestuderen, tot op heden
teleurstellend. Dit is mede het gevolg van methodologische tekortkomingen
zoals beschreven in hoofdstuk 1 van dit proefschrift. Ondanks het feit dat
bovenstaande studie ook enige beperkingen ondervindt, komen we tot de
conclusie dat een voordelig effect van cycling/rotatie op kolonisatie met β-
lactam resistente bacteriën ook hier niet definitief kan worden aangetoond.
Dit in acht genomen, is er dus geen onomstotelijk bewijs voor het
implementeren van cycling/rotatie designs als strategie in de strijd tegen
antibioticaresistente pathogenen.
Inmiddels is waarschijnlijk duidelijk dat de vraag met welke strategieën
antibioticaresistentie te controleren zou zijn, niet eenvoudig maar ook niet
onmogelijk te beantwoorden is. Om deze vraag te kunnen beantwoorden is
het belangrijk dat er betrouwbare data beschikbaar komen uit goed opgezette
observationele en interventiestudies. Microbiologische surveillance,
genotypering, registratie van antibioticumgebruik en observatie van
infectiepreventiemaatregelen zijn hierin onmisbaar. Modelering zoals
beschreven in dit proefschrift zou een waardevolle toevoeging hieraan
kunnen zijn.
201
Dankwoord Tja, en dan is er het dankwoord…. Is dit het makkelijkste of juist het
moeilijkste hoofdstuk van het hele proefschrift? In de afgelopen 6 jaar heb ik
niet alleen kennisgemaakt met de vele facetten van wetenschappelijk
onderzoek, maar vooral ook met een heel aantal personen zonder wie ik het
schrijven van mijn proefschrift niet voor elkaar had kunnen boksen.
Op de eerste plaats wil ik mijn beide promotoren prof. dr. M.J.M. Bonten
(Marc) en prof. dr. I.M. Hoepelman (Andy), bedanken.
Beste Marc, zonder jou was dit proefschrift nooit geworden wat het nu is. Je
enorme toewijding, drempelloosheid, eindeloze geduld en nooit afnemende
enthousiasme zijn zeer bewonderenswaardig en erg waardevol voor een
promovendus en zeker voor mij. Eerlijkheid gebiedt te zeggen dat we elkaar
ook regelmatig tot wanhoop hebben gedreven. Vooral de legendarische
uitspraak: ‘Het is nooit af…’, veroorzaakte menig slapeloze nacht. Maar dan
waren er altijd nog de twee andere legendarische uitspraken van professor
Weinstein (die ik hier niet gaat herhalen) en je goede gevoel voor humor
waardoor ik deze momenten snel weer vergeten was en dat ik nu trots kan
zijn op dit boekje. Ik heb veel van je geleerd en hoop in de toekomst nog
regelmatig met je van gedachten te kunnen wisselen over de wetenschap en
van alles en nog wat, onder het genot van een borrel. Bedankt voor alles!
Beste Andy, bedankt voor je vertrouwen in mij toen ik als groentje bij de
infectieziekten binnenstapte. Ik heb je betrokkenheid bij mijn onderzoek en
persoon altijd zeer gewaardeerd. Van een afstandje heb je met een scherp oog
203
mijn vorderingen gadegeslagen en, waar nodig, bijgestuurd. Daarmee heb je
voor mij de juiste voorwaarden weten te creëren om in een prettige omgeving
te kunnen werken en dat is heel wat waard! Ik heb veel geleerd van de
werkbesprekingen, congressen en de persoonlijke gesprekken die ik met je
heb gehad. Bedankt!
Dr. A.C. Fluit, beste Ad, als co-promotor ben je vanaf de eerste dag
betrokken geweest bij het schrijven van dit proefschrift. In het lab heb ik
geheel onder jouw bezielende leiding (samen met de anderen van de
ENARE-groep uiteraard) en met vallen en opstaan geleerd te werken met de
hoofdrolspelers in dit boekje (bacteriën welteverstaan). Onmisbaar voor mijn
toekomstige opleiding en ben je daar zeer dankbaar voor.
Maar wat me het meest is bijgebleven is je vermogen om te relativeren en
alles weer even in de juiste context te plaatsen. Als de resultaten van de
proeven weer niet waren wat ze zijn moesten, maar ook in het schrijfproces
gedurende de laatste periode. Altijd tijd voor een praatje waarin je met een
heldere, praktische en nuchtere kijk op de wetenschap en groot gevoel voor
humor alle obstakels even relativeerde en ik uiteindelijk na een half uurtje
uitrazen weer met hernieuwde energie aan het werk ging. En niet te vergeten
de gezellige momenten tijdens congressen en andere gelegenheden. Bedankt!
Dear prof. R.A. Weinstein, I would like to thank you for your contribution to
this thesis. Without any experience in research or any knowledge of infectious
diseases/microbiology, I came to Chicago and took my first steps in research
at Cook County Hospital under your supervision. Every morning in the car,
on the way to the hospital, you made sure I would not drown and I have
204
learned an awful lot during that period. Many thanks also to your family for
having me at many family parties, that was very special and so much fun.
Bob, Kay and Tom, thank you for all the work you did in the lab and for the
many good times we had during work, lunch and after work. You were really
wonderful!
Many thanks to the staff of the Infectious Disease Department and the
medical ICU of Cook County Hospital for their contributions and their
kindness.
Vanaf de eerste dag zat ik ook op het lab van de ENARE-groep. Graag wil ik
Alice, Stefan, Mirjam en Karlijn van de ENARE groep bedanken die, letterlijk,
stinkend veel hebben gedaan. Gedurende mijn hele onderzoeksperiode
hebben jullie honderden rectumkweken uitgewerkt, nooit was het teveel.
Daarnaast hebben jullie mij vanaf het begin in jullie groep opgenomen en heb
ik heel veel aspecten van de dagelijkse praktijk in een microbiologisch lab van
jullie geleerd. En wat hebben we een lol gehad! Een gezellige sfeer en heel
veel lachen, maakten dat ik dan ook altijd met plezier bij jullie zat te werken,
zelfs als het resultaat voor de zoveelste keer niet was wat ik hoopte.
Inmiddels zitten jullie allemaal op een andere plek, maar ik denk nog
regelmatig met plezier terug aan die tijd!
Binnen het Eijkman-Winkler instituut gaat verder mijn dank uit naar: Petra
Wijnhoven-Vroege, Maurine Leverstein-van Hall, Rob Willems, Janetta Top,
Helen Leavis, Annet Troelstra, Ellen Mascini, Titia Kamp, Hetty Blok, David
van de Vijver, Camiel Wielders, Menno Vriens.
205
De ICARE-studie was niet mogelijk geweest zonder de grote inzet en
betrokkenheid van alle intensivisten en verpleegkundigen van intensive care 1
en 2. Jullie hebben je volledig ingezet om het studieprotocol toe te passen,
gedurende 14 maanden iedere keer weer alle kweken op tijd af te nemen en je
regelmatig door ons op de vingers laten kijken tijdens onze ‘observaties’. De
kritische noten van jullie bij de uitvoering van de studie en de koffie tijdens
de observaties waren zeer waardevol. Bedankt voor jullie bijdrage.
‘De vrouwenvleugel’, bestaande uit Fieke (Grafieke), Marianne (Verkloot),
Irene (Prinsesje), Ilja (Pielja), Helga (schwester Helga), Mirelle (vele namen
zijn voorbij gekomen), Ruby, Jan Jelrik (‘prutser wat kun je wel?’) en Stefan(o)
(ja deze laatste twee ‘dames’ horen er ook bij), is inmiddels een begrip bij
velen en dat is niet voor niets. Dit unieke clubje mensen dat op een zekere
dag bij elkaar in een kamer werd gezet, heeft heel erg veel voor mij betekend
de afgelopen jaren, misschien wel meer dan ze zelf weten. Jullie zijn de meest
bijzondere collega’s die ik ooit heb gehad (in mijn enorm lange loopbaan,
haha!). We hebben veel met elkaar meegemaakt en lief en leed met elkaar
gedeeld. Wat er ook gebeurt, iedereen staat altijd voor elkaar klaar, niets blijft
onopgemerkt. Een bakkie leut (en leuten kunnen we!) met iets lekkers op het
juiste moment, een goede grap op z’n tijd (vaak uitmondend in de slappe
lach), een opbeurende opmerking en vooral geen geroddel zijn de bijzondere
eigenschappen van dit groepje.
Ik vond het heerlijk om met jullie een kamer te delen, te lunchen, te borrelen,
en vooral ook veel te lachen. Door jullie heb ik het in mijn mindere periodes
(waren die er dan???) wel volgehouden en daar ben ik jullie heel dankbaar
206
voor! Het gaat jullie goed en ik hoop dat we met elkaar nog heel lang
doorgaan!
Binnen de afdeling Acute Geneeskunde & Infectieziekten wil ik infectiologen
Karin Schurink en Margriet Schneider bedanken voor het meedenken en de
opbeurende woorden tijdens mijn promotietijd. Jullie inbreng op de
werkbespreking was zeer waardevol.
Tevens wil ik Martin Bootsma bedanken die als AIO bij de faculteit
Wiskunde betrokken was bij de beschreven wiskunde in dit proefschrift. Snel,
maar niet minder kritisch.
Secretaresses Jeanette en Els, bedankt voor alle ondersteuning vanuit het
secretariaat infectieziekten en voor de gezelligheid natuurlijk in de aflopen
jaren.
Buiten het UMC, had ik ook een privé-leven (jaha!!) en in die tijd heb ik
vooral veel gesquasht. Ik wil daarom mijn squashmaatjes Joan, Saskia, Esther,
Diana, Carola, Arenda en Susan bedanken voor de mogelijkheid tot
ontspannen, afreageren (zowel op de baan als aan de bar), de gezellige
competitiedagen en toernooien en de interesse die jullie hebben getoond in de
vorderingen van mijn proefschrift. Dames 3, op naar de eerste divisie
komend jaar!
En als ik niet op de squashbaan was te vinden, niet aan het hardlopen of
spinnen was, dan was ik vaak aan het koken voor mezelf en vrienden. Marlein,
Gwen en Sander, ik heb jullie vaak aan mijn kookuitspattingen onderworpen,
207
maar ik heb ook vaak van jullie kookkunsten mogen genieten! Bedankt voor
jullie interesse, steun en vooral jullie vriendschap van meer dan 25 jaar….
Sander, inmiddels vallen we onder het kopje exen, maar dat neemt niet weg
dat je de eerste drie jaar (en misschien wel de moeilijkste) van mijn
promotietijd in alle opzichten mijn maatje bent geweest. Ik wil je bedanken
voor je liefde, vriendschap, steun, trouw, zorgzaamheid en vooral je grote
gevoel voor humor: relativeren op het hoogste niveau!
Alarick, je hebt je opgeworpen om de lay-out van dit proefschrift te doen
terwijl ik op reis was. Alles verliep soepel, eeuwige dank!
Tenslotte natuurlijk mijn familie………
Pa en ma, bedankt dat ik altijd bij jullie kon aankloppen voor wat dan ook.
Jullie zijn er altijd en onvoorwaardelijk voor mij. Als een warm bad waar ik
altijd zo in kan stappen. Zonder jullie liefde, zorg en zeker ook financiële
steun zou ik niet geworden zijn wie ik nu ben. Ik ben trots op het feit dat
jullie de stap hebben durven nemen om samen naar het verre Veelerveen te
vertrekken en ons hier achter te laten. Het was een zware tijd, maar met veel
nieuwe vrienden en talloze bezigheden is het er volgens mij alleen maar
leuker op geworden. Fijn dat jullie zo genieten, maar wel mis ik de spontane
borrels op elk moment van de week met kaas, worst, noten en chips die
overal vandaan getoverd werden!
Maar gelukkig is er nog iemand die ook goed voor mij zorgt: mijn lieve
zusje….
208
Evelyne, de laatste jaren zijn we steeds dichter bij elkaar gekomen en ik heb
veel aan je gehad de afgelopen periode. Je openheid en eerlijkheid, altijd
gebracht met de nodige humor en zelfspot, maken je een bijzonder mens! Je
hebt met veel doorzettingsvermogen en geheel op eigen kracht je opleiding
tot kinderverpleegkundige afgerond, wat geen kattenpis was. Ben trots op je
en wat je ook gaat doen, het gaat je zeker lukken. En verder: vergeet vooral
niet ook te genieten!
209
Curriculum Vitae Saskia Nijssen is geboren op 9 september 1976 te Maarssen. Van 1988 tot
1994 volgde zij het Atheneum aan het Niftarlake College te Maarssen. Van
1994 tot 1996 maakte zij een begin aan haar opleiding geneeskunde aan de
Rijksuniversiteit van Antwerpen, waarna zij in 1996 de overstap maakte naar
de opleiding geneeskunde aan de Universiteit Utrecht. Op 28 juli 2000
behaalde zij haar doctoraal diploma om vervolgens in september van dat jaar
aan het in dit proefschrift beschreven onderzoek te beginnen in het
Universitair Medisch Centrum Utrecht (promotores prof. dr. M.J.M. Bonten
en prof. dr. I.M. Hoepelman). De eerste 3 maanden van dit
promotieonderzoek (september - december 2000) werden uitgevoerd in Cook
County Hospital te Chicago, onder leiding van prof. R.A. Weinstein. Van
maart 2004 tot en met januari 2006 liep zij haar co-schappen en op 27 januari
2006 behaalde zij haar artsexamen te Utrecht. Op 1 oktober 2006 zal zij
beginnen aan haar opleiding tot Medisch Microbioloog te Tilburg (opleider
Dr. M.F. Peeters).
211
List of Publications Nijssen S, Bootsma M, Bonten M. Potential confounding in evaluating
infection control interventions in hospital settings: changing antibiotic
prescription. Clinical Infectious Diseases; Accepted for publication on
September 1st 2006
Nijssen S, Florijn A, Top J, Willems R, Fluit A, Bonten M. Unnoticed spread
of integron-carrying Enterobacteriaceae in intensive care units. Clinical
Infectious Diseases 2005; 41(1):1-9.
Nijssen S, Bonten M, Weinstein R. Are active microbiological surveillance
and subsequent isolation needed to prevent the spread of MRSA? Clinical
infectious Diseases 2005; 40(3):405-09.
Nijssen S, Florijn A, Bonten M, Schmitz F, Verhoef J, Fluit A. Beta-lactam
susceptibility and prevalence of ESBL-producing isolates among more than
5000 European Enterobacteriaceae isolates. International Journal of
Antimicrobial Agents 2004; 24(6):585-91.
Nijssen S, Bonten M, Franklin C, Verhoef J, Hoepelman A, Weinstein R.
Relative risk of physicians and nurses to transmit pathogens in a medical
intensive care unit. Archives of Internal Medicine 2003; 163(22):2785-86.
Florijn A, Nijssen S, Verhoef J, Fluit A. Comparison of Double Disk
diffusion and E-test. European Journal for Clinical Microbiology and
Infectious Diseases 2002; 21(3):241-43.
213
