Professional Documents
Culture Documents
To determine whether routine antibiograms (summaries reporting resistance of all tested isolates) reflect resistance
rates among pathogens associated with hospital-acquired infections, we compared data collected from 2 different
surveillance components in the same 166 intensive care units (ICUs). ICUs reported data during the same months
to both the infection-based surveillance and the laboratory-based surveillance. Paired comparisons of the per-
centage of isolates resistant were made between systems within each ICU. No significant differences existed
(P 1 .05) between the percentage of isolates resistant from the infection-based system and laboratory-based system
for all antimicrobial-resistant organisms studied, except methicillin resistance in Staphylococcus species. The
mean difference in percentage resistance was higher from the infection-based system than the laboratory-based
system for S. aureus (mean difference, +8%, P ! .001 ) and coagulase-negative staphylococci (mean difference,
+9%, P ! .001). Overall, hospital antibiograms reflected susceptibility patterns among isolates associated with
hospital-acquired infections. Hospital antibiograms may underestimate the relative frequency of methicillin
resistance among Staphylococcus species when associated with hospital-acquired infections.
Susceptibility statistics, consisting of the cumulative and tant to various health care practitioners, including phy-
ongoing summary of the patterns of antimicrobial sus- sicians, pharmacists, infection control personnel, and
ceptibility of clinically important bacteria, are impor- microbiologists [1, 2]. Currently there are no standards
by which isolates should be included in these summary
data, although such guidelines are in development by
Received 2 October 2000; revised 7 December 2000; electronically published the National Committee for Clinical Laboratory Stan-
21 June 2001. dards (NCCLS). Many issues regarding generating these
Financial support: Grants to the Rollins School of Public Health of Emory reports, such as handling duplicate isolates, incorpo-
University for phase 2 and 3 of Project ICARE by AstraZeneca Pharmaceuticals
(Wilmington, DE) and Pfizer, Inc. (New York, NY). Full sponsors: Roche Laboratories ration of only select sites of cultures, or isolates from
(Nutley, NJ). Partial sponsors: Aventis Pharma (formerly Rhone-Poulenc Rorer; only confirmed infections are currently being discussed
Collegeville, PA), the National Foundation for Infectious Diseases (Bethesda, MD),
the American Society for Health Systems Pharmacists Research and Education by NCCLS (J. Hindler, personal communication). The
Foundation (Bethesda, MD), Kimberly-Clark Corporation (Roswell, GA), and Bayer frequency of antimicrobial resistance among isolates
Corporation, Pharmaceuticals Division (West Haven, CT).
from patients with infections acquired in the health care
Reprints or correspondence: Dr. Scott K. Fridkin, Hospital Infections Program,
MS E-A35, Centers for Disease Control and Prevention, 1600 Clifton Rd., Atlanta, setting may be different from that among all isolates
GA 30333 (skf0@cdc.gov). processed by the clinical microbiology laboratory [3].
Clinical Infectious Diseases 2001; 33:324–30
2001 by the Infectious Diseases Society of America. All rights reserved.
If such differences were observed, then separate sum-
1058-4838/2001/3303-0009$03.00 mary reports would be needed for hospital-acquired
infections and community infections to best aid in empiric prevalence of resistance among isolates associated with hospital-
therapy decisions. Such reports would be difficult to generate acquired infections reported through infection-based surveil-
by many clinical microbiology laboratories. Validating the clin- lance (pathogens reported through the NNIS-ICU component).
ical relevance of “all-isolate” summaries for health Validation of the clinical relevance of the routine hospital anti-
care–acquired infections would provide support for the current biogram would increase the confidence level among clinicians
widespread practice of using a single summary report in most that laboratory-based data can guide empiric therapy for pa-
hospitals. tients with hospital-acquired infections.
To augment surveillance for antimicrobial resistance at hos- METHODS
pitals in the United States, the Hospital Infections Program at
Centers for Disease Control and Prevention (CDC), in collab- Surveillance data and sites. Hospitals that participate in the
oration with the Rollins School of Public Health of Emory ICU surveillance component of the NNIS-ICU system were
University, began Project ICARE (Intensive Care Antimicrobial invited to participate in Project ICARE (NNIS-ICARE), and 61
Resistance Epidemiology) in 1996 at a subset of hospitals par- hospitals representing 171 ICUs submitted data to Project
ticipating in CDC’s National Nosocomial Infections Surveil- ICARE and the ICU component of the NNIS system during
lance (NNIS) system. Project ICARE hospitals reported data the study period January 1996–April 2000. The surveillance
on all strains of selected bacteria tested in the clinical micro- methodology and definitions of the NNIS system [4, 5] and
biology laboratory. These data allowed for determination of Project ICARE [6] have been described elsewhere.
prevalence rates of resistance similar to the routine hospital Infection-based reports. Participating hospitals reported
cumulative antibiogram. The cumulative antibiogram or cu- monthly hospital-acquired infection data from at least one ICU
mulative susceptibility test data report is the percentage of iso- to NNIS-ICU. The NNIS-ICU data include information on all
lates of a given species tested in a given institution, or specific hospital-acquired infections (infections occurring at any site)
areas of the institution, that are susceptible or resistant to each detected in patients during the month in which active sur-
of the antimicrobial agents routinely tested. We used data gath- veillance occurred in the ICU (figure 1). The susceptibility in-
ered from hospitals participating in our study to study differ- terpretation (susceptible, intermediate, resistant) for drugs
ences in the prevalence of resistance in participating intensive tested against each pathogen associated with the hospital-ac-
care units (ICUs) among bacteria reported through laboratory- quired infection is reported according to NCCLS breakpoint
based surveillance (as reported through Project ICARE) to the definitions [7–9]. This allows for determining the cumulative
Median
b c
testing rate Ratio of testing rates
Antimicrobial
Organism agenta IBRS LBRS Median 25th–75th percentile
Staphylococcus aureus Oxacillin 1.0 3.1 4.0 2.8–7.2
Enterococcus species Vancomycin 1.0 2.2 4.9 2.9–6.4
Coagulase-negative staphylococci Oxacillin 0.8 3.3 5.5 4.1–9.3
Klebsiella pneumoniae Ceftazidimed 0.6 1.5 3.4 2.1–3.8
Escherichia coli Ceftazidimed 0.6 2.1 4.3 2.7–8.0
E. coli Quinolone 0.6 2.2 4.8 3.5–7.2
NOTE. ICUs reported testing ⭓10 isolates to both the National Nosocomial Infections Surveillance System, ICU component
(infection-based reporting system) or National Nosocomial Infections Surveillance–Intensive Care Antimicrobial Resistance Ep-
idemiology (laboratory-based reporting system), January 1996–April 2000. IBRS, infection-based reporting system; LBRS, lab-
oratory-based reporting system.
a
Quinolone indicates ofloxacin or ciprofloxacin.
b
Testing rate is the average number of isolates tested against a specified antimicrobial per month for all data in an ICU.
c
Median and 25th–75th percentiles of all testing ratios (number of isolates tested by laboratory-based system divided by
the number tested by infection-based system) for the organism listed.
d
Organism tested against ceftazidime, ceftriaxone, or cefotaxime.
similar to that previously reported from the all NNIS hospitals It is of note that the mean difference in the prevalence of
reporting data from ICU patients (http://www.cdc.gov/ncidod/ vancomycin-susceptible enterococci between the 2 systems was
hip/surveill/nnis.htm) [11]. For the 11 antimicrobial-resistant 11%, although this difference still did not reach statistical
organisms evaluated, the number of ICUs testing ⭓10 organ- significance.
isms was consistently higher when we used data from the lab- DISCUSSION
oratory-based reporting system (median value, 110, range,
96–132 ICUs) than the infection-based system (median value, Aggregating cumulative susceptibility data is a common prac-
26, range, 10–44 ICUs). Therefore, comparisons between the tice among clinical microbiology laboratories, but the best
2 reporting systems is limited to those ICUs that reported sus- methodology to analyze and present these data has not been
ceptibility results on a sufficient numbers of organisms by the determined. The most common use of the hospital antibio-
infection-based system. grams data is probably for assisting clinicians with empiric
The prevalence rates of resistance to antimicrobial agents therapy for suspected infections [2]. Our study suggests cu-
among all ICUs were usually higher among isolates reported mulative susceptibility data (laboratory-based reporting) that
to the infection-based system than among isolates reported to clinicians use to guide empiric treatment of hospital-acquired
the laboratory-based system (table 2). However, these differ- infections usually will be representative of the organisms fre-
ences were rarely statistically significant when we used the pair- quently encountered among hospital-acquired infections. Our
wise comparisons, which takes into account the differences study evaluated select patterns of resistance to antimicrobial
observed within each ICU and the consistency with which these agents among the organisms most frequently associated with
differences were observed. There were 2 notable exceptions. By hospital-acquired infections and compared the resistance rates
use of the pairwise comparisons, infection-based prevalence of reported to a laboratory-based monitoring system (NNIS-
resistance was significantly higher than laboratory-based prev- ICARE) to rates reported to an infection-based monitoring
alence for MRSA (average difference, ⫹7.9%; P ! .001) and system (NNIS-ICU). With 2 exceptions, the resistance rates
methicillin-resistant coagulase-negative staphylococci (mean among isolates associated with only hospital-acquired infec-
difference, ⫹8.7%; P ! .001; table 2). Similar comparisons for tions were similar to rates among isolates associated with in-
the other organisms evaluated identified no additional signif- fection (both community and hospital acquired), colonization,
icant differences. Similar analysis comparing the percentage of and contaminants.
isolates susceptible to the antimicrobial revealed similar results. The 2 organisms associated with significantly higher rates of
NOTE. Included are data from ICUs that reported sufficient data (⭓10 isolates) to both laboratory-based and infection-
based reporting systems and the average difference in prevalence by reporting system, National Nosocomial Infections Sur-
veillance System, January 1996–April 2000. IBRS, infection-based reporting system; LBRS, laboratory-based reporting system;
%R, prevalence of resistance.
a
Weighted mean difference for all organisms except S. aureus and coagulase-negative staphylococci (arithmetic mean).
b
P value by Z test of weighted mean difference among paired resistance rates, except S. aureus and coagulase-negative
staphylococci, by Z test of mean difference. Paired t test and Wilcoxon signed-rank test resulted in similar P values (data not
shown).
c
Denotes nonsusceptibility (E. coli or Klebsiella pneumoniae) or resistance (Enterobacter species) to ceftazidime, ceftriaxone,
or cefotaxime.
resistance among isolates reported to the infection-based system coagulase-negative staphylococci. However, the impact of these
compared with the laboratory-based system were MRSA and factors does not appear strong enough to consistently create
methicillin-resistant coagulase-negative staphylococci. Although differences in measured resistance prevalence of any magnitude
several other organisms tended to be associated with higher re- worthy of clinical importance. If the sample of data were larger,
sistance rates in the infection-based system, these differences were we might have identified more significant differences in the
not statistically significant. We would expect that resistance to reporting systems. However, the relative differences in resis-
antimicrobial agents is more frequently encountered among tance prevalence observed were small (!10%) and unlikely to
organisms associated with hospital-acquired infections com- have any clinical impact, even if differences were to become
pared with community-infections for several reasons. First, pre- statistically significant.
vious analysis of data from Project ICARE demonstrated that, We limited our study to ICUs because our surveillance data
with a few exceptions (i.e, Streptococcus pneumoniae), organ- for hospital-acquired infections were limited to the ICU. Cur-
isms with onset of disease in the community (from the out- rently, the NNIS system does not receive reports from hospital-
patient or emergency department setting) are less likely to be wide surveillance, making comparisons to the non-ICU areas
resistant to antimicrobial agents than are those encountered in impossible. Although this study involved data from only ICU
the inpatient setting [6]. Second, patients with longer hospital patients, its conclusions may be applicable to hospital-wide
stays are at increased risk to become infected with antimicro- data. This is because we evaluated the differences in reported
bial-resistant pathogens, and these isolates would be more rep- resistance rates within the same ICUs, and we suspect these
resented in a infection-based system (i.e., nosocomial infection) differences would be of similar magnitude and statistical sig-
than the laboratory-based system [3, 12–14]. Third, there would nificance in the non-ICU areas. The major differences between
be potential for reporting bias if infection control practitioners these patient groups in terms of resistance to antimicrobial
were more likely to report hospital-acquired infections when agents are higher severity of illness and higher rates of resistance
they were associated with an antimicrobial-resistant pathogen, to antimicrobial agents [6]. With this lower prevalence of re-
although this seems unlikely. Some of these hypotheses may sistance to antimicrobial agents in the non-ICU areas, we may
explain the significantly higher resistance rates observed in data expect any observed differences in the prevalence by reporting
from infection-based reporting compared with that from lab- system in these hospital areas to be smaller than that observed
oratory-based reporting among MRSA and methicillin-resistant in our study. A second limitation of our study is the exclusion