Clinical and Epidemiological Studies
Are Nosocomial Infection Rates in Intensive
Care Units Useful Benchmark Parameters?
P. Gastmeier, D. Sohr, C. Geffers, A. Nassauer, F. Daschner, H. Rüden
Summary
Background: The objectives of this study were to
determine to what extent the German national
nosocomial infection surveillance system (Krankenhaus
Infektions Surveillance System, KISS) can take into account
the circumstances prevailing in various intensive care
units (ICUs) and to establish whether KISS-ICU infection
rates can serve as useful benchmark parameters.
Methods: The investigation focused on three major
factors: microbiological monitoring, severity of illness
and the duration of surveillance. For each of these
factors separate infection rates were calculated for
various ICU groups and the differences compared.
Results: Significant differences were found for catheter-
associated urinary tract infections (CAUTI) with routine
monitoring, but not for ventilator-associated pneumonia
(VAP). Significant differences were assessed for central
venous catheter-associated bloodstream infections (CVC-
BSI), considering the average ventilator utilization rate in
the ICU as a surrogate parameter for the average severity
of illness in its patient group. Surveillance periods of
about 1 year were necessary to confirm definite outlier
and nonoutlier positions for the majority of the ICUs.
Conclusion: Using KISS data for internal orientation, it is
possible to note important differences between ICUs
when interpreting infection rates; some initial examples of
successful use of surveillance data for the reduction of
infection rates are already available. However, the use of
such data for external assessment is not possible, because
external observers are often unable to fully consider
important factors in the interpretation of infection rates.
Key Words
Nosocomial infections · Intensive care units ·
Benchmarking · Quality management · Surveillance
Infection 2000; 28: 346-350
Introduction
Despite advances in the control and prevention of nosoco-
mial infections (NIs), they remain a major side effect of all
hospital treatment.The rate of occurrence of NIs is affected
346
by the extent of the underlying disease, the degree of inva-
siveness of medical procedures in various patient groups as
well as by the preventative measures taken. The idea of us-
ing NI rates as benchmark parameters for evaluation and
improvement of the situation in intensive care units (ICUs)
has been discussed extensively. There is however some
concern about the possible misuse of NI data by the press,
insurance companies and boards of control to negatively
evaluate hospitals or even to file claims of malpractice.
Despite these concerns, more than 100 German hospi-
tals accepted the invitation by the National Reference Cen-
ter for Hospital Hygiene to participate in the development
of a reference data base for NI in Germany (Krankenhaus
Infektions Surveillance System, KISS). After 30 months of
KISS we are able to demonstrate to what extent the NI
rates of ICUs are useful as benchmark parameters.
Methods
Description of the Surveillance System
KISS was established in 1996 using a surveillance protocol based on
the National Nosocomial Infections Surveillance (NNIS) System
[1]. KISS focuses on the most important ICU infections: ventilator-
associated pneumonia (VAP); central venous catheter-associated
bloodstream infection (CVC-BSI); and catheter-associated urinary
tract infection (CAUTI). Device-associated infection rates were cal-
culated for these types of infections. The number of device days is
derived from recordings which lack information about device use
on individual patients.The data were stratified for the various types
of ICUs (medical, surgical, medical/surgical).
Infection control staff from interested hospitals were invited
to participate in an introductory course, where they were trained
in the application of Centers for Disease Control (CDC) defini-
Petra. Gastmeier (corresponding author), Dorit Sohr, Christine Geffers,
H. Rüden
Institute of Hygiene, Free University of Berlin, Charité Campus, Virchow
Clinic, Heubnerweg 6, Haus II, D-14059 Berlin, Germany; Phone: (+49/30)
4506-1002, Fax: -1900, e-mail: Petra.Gastmeier@charite.de
A. Nassauer
Robert Koch Institute, FB 14, Applied Infection and Hospital Hygiene,
Stresemannstr. 90-102, D-10963 Berlin, Germany
F. Daschner
Institute of Environmental Medicine and Hospital Hygiene, Albert Lud-
wig University of Freiburg, Hugstetter Str. 55, D-79106 Freiburg, Germany
Received: March 24, 2000 • Revision accepted: October 9, 2000
Infection 28 · 2000 · No. 6 © URBAN & VOGEL
 P. Gastmeier et al.: Nosocomial Infection Rates in ICUs
tions for NI [2, 3], and where the main methodological commit-
ments were explained. This included continuous participation of
the ICUs for a period of at least 1 year, regular attendance at an-
nual training courses by infection control personnel and standby
arrangements for questions concerning surveillance data and other
data necessary for the interpretation of NI infection rates.The main
obligations of KISS were the regular (6-monthly) communication
of infection rates to the hospitals (their own data and the reference
data) and a guarantee to treat this data confidentially.
The first ICUs sent data to KISS at the beginning of 1997.At the
end of 1997, 25 ICUs were participating, which increased to 66 at the
end of the following year. In June 1999, 113 ICUs were involved.
Methods for Evaluating the System to Produce Benchmark
Parameters.There are three main factors which influence the use
of NI data as benchmark parameters: accuracy of NI diagnosis; dif-
ferences in severity of illness of the ICU patients; and duration of
surveillance periods, so as to exclude possible random influences.
These factors were investigated using the following methods.
Method for Investigating the Influence of Accuracy of the NI Di-
agnosis. Differences in infection rates can result from variations
among infection control personnel regarding their interpretation
of CDC criteria. However, the major concern of the ICUs are dif-
ferences in infection rates resulting from whether or not routine
microbiological monitoring investigations are carried out in the
unit. Therefore, the ICUs were contacted and asked in a ques-
tionnaire about the number of routine monitoring investigations
of urine and tracheal secretions conducted. 68 ICUs delivered this
information: CAUTI and VAP rates were calculated separately for
those with routine microbiological monitoring (once or twice per
week) and those performing these investigations only when ap-
propriate symptoms were found.The medians of both groups were
compared using the Wilcoxon rank sum test to compare the me-
dians of both groups.
Methods for Investigating the Influence of the Severity of Illness in
ICU Patients. There are no uniform rules in Germany for record-
ing severity of illness scores in ICUs. As a result, if routine scoring
of patients on admission is carried out at all, different scoring sys-
Table 1
Standardized overall nosocomial infection (NI) rates according to the type of intensive care unit (ICU).
Type of Type of ICU (no. of ICUs) Device days
infection
VAP All ICUs (112a) 178,026
Medical/surgical ICUs (55) 96,297
Medical ICUs (23) 19,558
Surgical ICUs (31) 54,069
CVC-BSI All ICUs (113b) 298,443
Medical/surgical ICUs (55) 161,162
Medical ICUs (23) 29,286
Surgical ICUs (31) 95,359
CAUTI All ICUs (113b) 320,069
Medical/surgical ICUs (55) 172,405
Medical ICUs (23) 33,098
Surgical ICUs (31) 101,040
a only 112 ICUs perform routine surveillance of ventilator-associated pneumonia cases; b the remaining four ICUs were three neurosurgical
and one pediatric ICU; VAP: ventilator-associated pneumonia; CVC-BSI: central venous catheter-associated bloodstream infection; CAUTI:
catheter-associated urinary tract infection
Infection 28 · 2000 · No. 6 © URBAN & VOGEL
tems are used. Therefore, stratification of ICUs based on any one
type of severity of illness score is impossible. As the average venti-
lator utilization rate of any one ICU may only describe the average
severity of illness in that ICU [4], CVC-BSI rates were stratified in
two groups: those with a ≥ 40% ventilator utilization rate and those
with a < 40% average ventilator utilization rate.The Wilcoxon rank
sum test was used for comparing infection rate medians.
Methods for Investigating the Influence of the Duration of Sur-
veillance Periods to Exclude Possible Random Influences. We rec-
ommend taking the value of the 75th percentile as an arbitrary
threshold for possible problems of infection control. However,
long surveillance periods are often necessary to detemine whether
there is an outlier (where the lower limit of the 95% CI of the in-
fection rate is above the threshold of the 75th percentile) or no
outlier (where the upper limit of the 95% CI is below this thresh-
old), particularly when the infection rate is near the median or the
90th percentile. In order to demonstrate the distribution of defi-
nite outliers and definite nonoutliers among KISS-ICUs partici-
pating for at least 1 year, these CIs were determined for VAP. For
each ICU it was determined whether there was a definite outlier
position, a definite nonoutlier position or an undetermined posi-
tion, by comparing the CIs with the threshold value.
Results
General Survey of KISS Data
As of June 30, 1999, the number of ICUs participating had
increased to 113, comprising 55 medical/surgical ICUs,
23 medical ICUs, 31 surgical ICUs plus four other types of
ICUs. They were dispersed over the whole country, and
their affiliation grouping according to hospital size was in
accordance with the general distribution of such sized
groups in Germany. There was a survey of about 99,448
ICU patients with 398,792 patient days. Table 1 shows the
data for the three types of device-associated infections as
well as for the different types of ICU.
Device-associated infection rates
Pooled mean Median 75th percentile 90th percentile
11.2 8.0 14.8 21.8
9.1
5.8
9.9
1.8 1.2 2.5 2.5
1.1
1.6
1.3
3.9 2.7 5.3 9.3
1.6
2.9
3.6
347