Extensive Hospital-Wide Spread of a Multidrug-Resi
http://www.100md.com
微生物临床杂志 2006年第2期
ABSTRACT
A hospital-wide increase in the number of patients with aminoglycoside-resistant Enterobacter cloacae (AREC) isolated from clinical cultures was detected in December 2002 using a classical surveillance system (CSS). CSS refers to a strategy based on the recognition of an increased incidence of a species with a particular antibiogram at certain wards in a limited period. Since clonal spread was suspected, hospital records were reviewed for E. cloacae culture-positive patients. Based upon genotyping of 139 clinical E. cloacae isolates from 80 patients, it was concluded that 53 patients had had clinical cultures with a single AREC clone since April 2001. Determinants for unnoticed spread were investigated retrospectively, as was the possibility that a computer-assisted surveillance method would have detected this outbreak at an earlier stage. Determinants associated with late detection of clonal spread were the following: (i) the absence of a hospital-wide increase in incidence of E. cloacae cases for 1.5 years, (ii) the long time interval between cases, (iii) the hospital-wide occurrence of new cases, due to a high number of patient transfers between wards, (iv) the large variety of clinical sites, and (v) the high variability of antibiograms (n = 33). Retrospective application of a recently described computer-assisted surveillance method as well as an "in-house"-developed algorithm resulted in earlier detection of the outbreak of 6 and 12 months, respectively. These findings suggest that computerized tools for surveillance may recognize resistance trends that are too complex to be detected by manual review and indicate the need for prospective evaluation of such algorithms.
INTRODUCTION
In some countries up to 70% of nosocomial infections are caused by microorganisms resistant to at least two antimicrobial agents (6). Infections caused by such pathogens have been associated with increased morbidity, higher mortality, and raised costs for medical care (3). Therefore, it is critical that dissemination of multidrug-resistant microbes is contained. Surveillance of nosocomial infections and antibiotic resistance are important tools for limiting dissemination of antimicrobial resistance (1, 3). Outbreaks are usually recognized upon an increased incidence of specific infections caused by a strain characterized by a particular resistance pattern within a limited time period in a specific ward. This approach could be defined as the classical surveillance strategy. Such combinations of events may be recognized by manual or computerized review of laboratory and hospital data or by an astute clinician or clinical microbiologist.
In January 2003, widespread nosocomial dissemination of multidrug-resistant Enterobacter cloacae in our hospital became apparent. In retrospect, this pathogen was associated with nosocomial infections in 53 patients since April 2001 and possibly even more in the period before. Determinants for unnoticed spread were investigated retrospectively. A recently described computer-assisted surveillance method and an "in-house"-developed method were retrospectively applied to the database of the microbiology laboratory to determine whether this unnoticed outbreak could have been identified at an earlier stage (5).
MATERIALS AND METHODS
Setting, patients, and bacterial strains. The University Medical Center Utrecht in The Netherlands is a 1,042-bed tertiary care hospital with all medical disciplines represented. Each discipline has its own department consisting of different ward(s) and units often sharing the same staff. In addition, there are four adult intensive care units (ICUs) of which three are located next to each other and another is located on a different floor.
At the end of 2002, a hospital-wide outbreak of tobramycin-resistant Enterobacter cloacae (TREC) was recognized, and laboratory records from 1 January 2001 to 28 January 2003 were reviewed to identify patients with aminoglycoside-resistant Enterobacter cloacae (AREC) isolated from clinical cultures. Almost all AREC isolates had been stored and were available for further analysis. In addition, 56 randomly chosen aminoglycoside-susceptible E. cloacae isolates (ASEC), obtained from sterile sites between 1 January 2001 and 1 July 2003, were processed for genotyping.
Identification, antimicrobial susceptibility, and molecular typing. Identification and susceptibility testing were performed using an automated system and software (Phoenix Automated Microbiology System, Becton Dickinson Biosciences, Sparks, MD). Breakpoints were those recommended by the National Committee for Clinical Laboratory Standards (10). Aminoglycoside resistance was confirmed by using Etests (AB Biodisk, Solna, Sweden). Susceptibility patterns of all genotyped isolates were analyzed. Ten antimicrobial agents were included: gentamicin, tobramycin, trimethoprim-sulfamethoxazole, trimethoprim, amikacin, ciprofloxacin, tetracycline, chloramphenicol, nitrofurantoin, and meropenem. Penicillins and cephalosporins were not included, since beta-lactam resistance in E. cloacae often results from potentially reversible induction of the chromosomally located AmpC genes. Consequently, beta-lactam resistance is not used for recognition of epidemic strains. From each patient the isolates with a unique susceptibility pattern were typed by pulsed-field gel electrophoresis analysis (PFGE) as described previously, with the exception that thiourea (50 uM) was used routinely in the running buffer (8). Gels were normalized with 5 references for every 22 isolates tested on each gel. The Jaccard coefficient of similarity was calculated, and the unweighted pair group method with arithmetic averages (UPGMA) was used for cluster analysis in BioNumerics version 2.5 (Applied Maths, Sint-Martens-Latem, Belgium). The optimization was set at 1%, and the tolerance was set at 2% for the smaller fragments and 4% for the larger fragments to cope with interassay variation. Subsequently, the UPGMA clustering was visually checked.
Surveillance policy of the Department of Hospital Hygiene and Infection Prevention. During the period of the study, surveillance of hospital infections in our hospital included (i) active surveillance by infection control professionals consisting of biannual hospital-wide point prevalence surveys for hospital-acquired infections, followed by unit-directed surveillance if indicated, and (ii) clinical microbiology laboratory-based daily surveillance directed at microbial species with specific resistance determinants (aminoglycoside-resistant) Enterobacteriaceae, extended-spectrum ?-lactamases (ESBL)-positive Enterobacteriaceae, vancomycin-resistant Enterococcus faecium, and methicillin-resistant Staphylococcus aureus. If cultured, these isolates were stored at –70°C and automatically reported to the Department of Hospital Hygiene and Infection Prevention. Subsequently, isolation precautions for the patients involved were taken. Once an infection control professional detected at a certain ward an increase in the number of patients with any of these marker microorganisms, genotyping was performed to detect clonal dissemination. Another reason for genotyping was the remark of a consultant about a possible increase in incidence of a particular species or resistance pattern. Written guidelines for decision making regarding typing of microorganisms are present for a limited number of types of microorganisms (i.e., vancomycin-resistant E. faecium, methicillin-resistant Staphylococcus aureus, and Pseudomonas aeruginosa for ICUs).
Computer-assisted surveillance. Recently, Hacek et al. developed two algorithm tools (based on increases of organism number over baseline) to analyze laboratory data for possible clonal dissemination (5). The first algorithm (2SD) defined an alert as two standard deviations (SDs) above the mean monthly number of isolates. The second algorithm (MI) defined an alert as either a 100% increase from the baseline organism number over 2 months or a 50% increase during a three-consecutive-month period. To determine whether the use of these algorithms would have resulted in an earlier recognition of the outbreak, the mean monthly number of E. cloacae isolates in our hospital was determined from 1995 to 2001. Duplicates, defined by Hacek et al. as the same organism with the same susceptibility pattern from the same patient from the same source from the same month, were removed from the database.
Furthermore, a recently in-house-developed algorithm tool was tested which defined an alert as a >200% increase above the median monthly number of new cases. This algorithm was applied to the first TREC isolate of each patient present in the database. To test this algorithm, the median of the monthly number of patients with a TREC isolate in our hospital was determined from 1995 to 2001. In addition, the two algorithms described by Hacek et al. were tested on this same data set.
RESULTS
Descriptive epidemiology. In September 2002 (weeks 36 to 38), an increased incidence of patients with E. cloacae, including TREC, was noticed at different surgical wards. Six isolates obtained from six different patients were genotyped, and PFGE restriction patterns of only two had >90% similarity. Therefore, no specific control measures were reinforced and no screening to detect carriers was performed. However, the hospital-wide incidence of patients with a TREC-positive culture remained high (Fig. 1). In December 2002, all AREC isolates stored since January 2001 were genotyped. These genotyping results revealed a large outbreak of one clonal lineage, involving 53 patients since April 2001. Basic infection prevention measures (including enforcement of adherence to aseptic practices and barrier precautions), a restrictive antibiotic policy, identification of possible environmental reservoirs, and screening for intestinal colonization of contact patients were implemented, as described before (9).
PFGE typing. In total, 168 isolates from 109 patients were analyzed. Seven isolates from six patients were repeatedly nontypeable. The 161 remaining E. cloacae isolates comprised 105 AREC isolates (60 patients) and 56 ASEC isolates (53 patients). Based on the results of the UPGMA clustering and the matching visual analysis of the PFGE restriction patterns, 89 isolates obtained from 53 patients belonged to 1 clonal lineage (called cluster I). Although there was some variation in restriction patterns, all isolates with 80% similarity in PFGE patterns were considered to belong to cluster I (Fig. 2). Preliminary data from molecular analyses demonstrated that at least 94% of the isolates harbored the same 180-kb plasmid containing genes encoding integrons, an ESBL, aminoglycoside resistance, and quinolone resistance, supporting the clustering found in PFGE analysis. This plasmid was not demonstrated for five isolates that had slightly different PFGE patterns in the lower-molecular-weight fragments (called cluster Ia), which were all susceptible to tobramycin (Table 1; Fig. 2).
Incidence and intervals. Monthly incidences of patients with E. cloacae isolated from clinical cultures varied from 0 to 9 (Fig. 1). In all, 349 patients had E. cloacae isolated from clinical cultures; 66 (20%) had TREC, of which 30 patients had ASEC isolates as well. The remaining 283 patients had only ASEC isolates.
From the monthly incidence curve, three findings emerge: (i) incidences of clinical TREC remained low until August 2002, with a single exception in January 2002; (ii) the outbreak strain did not increase the hospital-wide incidence of E. cloacae until September 2002; and (iii) in December 2002 and January 2003, the outbreak strain seemed to replace susceptible E. cloacae isolates.
Furthermore, the mean time interval between detection of cases was 3 weeks, ranging from 1 to 22 weeks.
Susceptibility patterns. Eighty-nine E. cloacae cluster I isolates (53 patients) expressed 33 different susceptibility patterns (Table 1). Of these antibiograms, 30 (91%) represented a tobramycin-resistant variant. Tobramycin-susceptible variants were detected only in genotype IA isolates (five patients).
Patient transfers. Patients colonized with E. cloacae cluster I were identified in 18 different wards or units belonging to 8 different divisions: (i) surgical division, 29 patients at five surgical wards, including 12 patients in the ICU surgery; (ii) neurodivision, 8 at 5 neurological and neurosurgical wards and 8 at the neurosurgical ICU; (iii) division internal medicine, 4 at the 3 medicine wards, including 2 patients on the internal ICU; and (iv) another four patients at four other wards. Forty-three of the fifty-three cases could be linked to each other by overlapping days of hospitalization on the same wards. Eight patients had overlapping days of hospitalization with other cases but were identified at wards where no previous case had been identified. Two cases were identified within 48 h of admission at new wards. To determine to what extent patient transfer between wards had contributed to spread of genotype 1, the hospital database was reviewed for transfer frequency of affected patients. This frequency was compared to frequencies of all hospitalized patients in 2001/2002, as well as in 1991/1992.
The average number of patient transfers between wards at the time of identification of colonization or infection with E. cloacae type I was 3.94 (range, 1 to 7). For the hospital population at large, average numbers of transfers nearly doubled in the last 10 years from 0.37 in 1991/1992 to 0.7 in 2001/2002, while the average duration of hospital stay decreased in the same time interval by 23% (12.5 to 9.6 days). The transfer rate between different departments, however, was very low in both periods.
Infection sites. The outbreak strain caused a variety of hospital-acquired infections, as defined by the criteria of the Centers for Disease Control and Prevention, in 53 patients. The strain was isolated from sputa or bronchoalveolar fluid (20 infections, 5 colonizations), urine (8 infections), catheter urine (2 infections, 4 colonizations), wounds (7 infections), cervix (1 colonization), and from the following sterile sites: abdominal fluid (3 infections), liquor or external ventricular drain (8 infections), blood (4 infections), and an aortic valve (1 infection).
Retrospective evaluation of computer-assisted surveillance on outbreak data. The average monthly number of E. cloacae isolates from clinical cultures in our hospital was 37.2 for the period of 1995 to 2001 (range of the monthly mean, 27.7 to 44.6; range of the monthly SD, 4.0 to 15.9). Application of the 2SD algorithm would have yielded a first alert in March 2002 after analysis of the data obtained in February 2002, while the MI algorithm would not have yielded any alert.
The median monthly number of patients with TREC isolated from a clinical culture in our hospital was 0.9 for the period 1995 to 2001 (range of the monthly median, 0 to 2; range of the monthly SD, 0.5 to 2.4). Application of our in-house algorithm towards the overall incidence of patients with TREC would have yielded an alert 6 months prior to the first alert of the surveillance system by Hacek et al. (Fig. 1). The 3 new cases in September 2001 (September: monthly median, 1.4; SD, 2.1) would have caused an alert by the in-house algorithm (threshold, 2.8), while the 2SD algorithm (threshold, 5.6) would have remained silent until March 2002. The MI algorithm would have yielded an alert after the 3-month interval August-September-October 2002, similar to outbreak recognition based upon the classical surveillance strategy.
DISCUSSION
Our experience with a multidrug-resistant strain of E. cloacae exemplifies the pitfalls encountered in detecting potential outbreaks in a hospital setting. Classically, surveillance for detection of clonal dissemination is based on the recognition of an increased incidence, in a limited period of time, of a certain strain, characterized by a particular resistance pattern in a certain ward, preferentially causing specific infections. Here, we were confronted with prolonged clonal dissemination of a multidrug-resistant strain of E. cloacae with 33 different antibiograms, initially without a manifest increase in total numbers of E. cloacae infections at multiple sites, affecting patients in 18 different wards.
The fact that consecutive cases were identified in different wards or even divisions could partially be explained by the high number of transfers of patients between different wards, being nearly six times as high as the average of hospital-wide number of transfers. The average number of within-hospital transfers nearly doubled in the last 10 years, whereas average length of stay significantly decreased in that period, implicating a substantial increase in patient movement through the hospital. This will catalyze dissemination of bacterial clones through the hospital and hamper rapid recognition of clonal spread.
Another reason for the late detection was the absence of a characteristic susceptibility pattern. Although tobramycin resistance was highly prevalent among outbreak isolates, the patterns of susceptibility to other antimicrobial agents were extremely diverse, further hindering recognition of clonal relatedness between different isolates. To our knowledge, this is the first time that such a high variety in susceptibility patterns within one clonal lineage has been described. Whether this is an emerging trait among nosocomial Enterobacteriaceae in general or is unique to this strain remains to be settled. The molecular basis for the high variety of antibiograms in this strain is currently under study. Horizontal gene transfer seems to play a role. Preliminary results show that the majority of the outbreak isolates carried a large plasmid containing genes encoding integrons, an ESBL, aminoglycoside resistance, and quinolone resistance. In addition to genetic differences within this clonal lineage, natural variations in antibiotic susceptibilities may have caused to some extent the high variety of antibiograms. Therapeutic (CLSI [formerly NCCLS]) breakpoints were used to distinguish resistant and nonresistant isolates in our study. Yet these breakpoints do not necessarily correspond with the optimal MICs for distinguishing between "susceptible" (wild-type) and "resistant" populations. Therefore, some of the observed shifts in susceptibility might be "artificial." Routine antibiotic susceptibility testing of clinical isolates over a wide range of concentrations might, therefore, be preferred. However, the automated susceptibility testing systems available for routine clinical laboratories, such as the Phoenix used in this study, test only a limited range of antimicrobial concentrations around therapeutic breakpoints. Therefore, susceptibility data generated by the clinical laboratory should be interpreted with caution when used for surveillance purposes.
Computerized tools for surveillance might improve our practice (2, 5, 12, 13). Retrospective application of the computer-assisted surveillance system using the algorithms and criteria of Hacek et al. would have produced an alert 6 months before the outbreak was actually detected. The method of Hacek et al. was directed at detection of a hospital-wide increase in the incidence of certain microorganisms, without taking susceptibility patterns, locations, or number of isolates per person into account. Only exact duplicates are removed.
In our opinion, analysis of microbiological data at the patient level of specific genera or species with certain resistance markers prone to causing nosocomial outbreaks would be a more efficient method of surveillance for detecting dissemination of microorganisms between patients. Therefore, an in-house-developed algorithm was retrospectively applied to the first clinical TREC isolate per patient, which resulted in an alert 12 months before the outbreak was actually detected. Application of the more-complicated algorithms of Hacek et al. towards this same data set resulted in an alert 6 months before the actual detection of the outbreak. As a consequence, we prefer the in-house-developed algorithm and currently are evaluating this algorithm for prospective use in surveillance prospectively.
This approach, directed at the combination of specific genera or species with certain resistance markers prone to causing nosocomial outbreaks, is in accordance with the conclusion of Gardam et al. that efficient surveillance might be achieved by performing routine multidrug-resistant Enterobacteriaceae surveillance of clinical isolates (4). This conclusion was made after they had shown that performing multidrug-resistant Enterobacteriaceae surveillance in the absence of an outbreak was a costly exercise that provided little or no benefit for infection control or predicting clinical infection.
Further improvement in the accuracy of the surveillance system for identifying outbreaks could be achieved by using department-specific instead of hospital-wide databases. Patients were transferred frequently between wards or units within the same department (e.g., within the surgical department containing five wards) but only sporadically between departments (e.g., from surgery to internal medicine). This hypothesis, though, needs to be empirically tested.
Well-defined criteria for deciding whether genotyping to assess microbial clonality should be performed are lacking. For instance, should genotyping be performed after every alert, or should different criteria for different species with different resistance traits be used? Our findings reveal that different antibiograms, different sources, long time intervals between cases, and isolation from different wards do not exclude the possibility of clonal spread. Naturally, genotyping can be helpful only if sufficient relevant isolates have been stored. Since it is impossible to store all clinical isolates, choices must be made. These choices should be based on the local ecology and susceptibility patterns with a low threshold to respond to ecological changes. The outbreak described here could be analyzed only because all aminoglycoside-resistant Enterobacteriaceae had been routinely stored.
Our findings also demonstrate the possibility of substantial variation of restriction patterns within a single clone when using PFGE. This finding emphasizes the need to include an adequate number of control isolates in the database to prevent variants of the same clone from being wrongly interpreted as unrelated. Alternatively, a more robust and less labor-intensive typing schemes could enhance the capacity to identify outbreaks, even in nonresearch settings. A recently developed typing scheme based on multiple loci as multilocus sequence typing or multiple locus variable number of tandem repeat analysis would fulfill these requirements (7, 11).
REFERENCES
Association for Professionals in Infection Control and Epidemiology. 1996. Overview of isolation systems, appendix C-6. In R. N. Olmsted (ed.), APIC infection control and applied epidemiology. Mosby Yearbook, Inc., St. Louis, Mo.
Bouam, S., F. Girou, C. Brun-Buisson, H. Karadimas, and E. Lepage. 2003. An intranet-based automated system for surveillance of nosocomial infections: prospective validation compared with physicians' self-reports. Infect. Control Hosp. Epidemiol. 24:51-55.
Burke, J. P. 2003. Infection control—a problem for patient safety. N. Engl. J. Med. 348:651-656.
Gardam, M. A., L. L. Burrows, J. V. Kus, L. D. E. Low, J. M. Conly, and A. Humar. 2002. Is surveillance for multidrug-resistant Enterobacteriaceae an effective infection control strategy in the absence of an outbreak? J. Infect. Dis. 186:1754-1760.
Hacek, D. M., R. L. Cordell, G. A. Noskin, and L. R. Peterson. 2004. Computer-assisted surveillance for detecting clonal outbreaks of nosocomial infection. J. Clin. Microbiol. 42:1170-1175.
Haley, R. W., D. H. Culver, J. W. White, W. M. Morgan, T. G. Emori, V. P. Munn, and T. M. Hooton. 1985. The efficacy of infection surveillance and control programs in preventing nosocomial infections in US hospitals. Am. J. Epidemiol. 12:182-205.
Keys, C., S. Kemper, and P. Keim. 2005. Highly diverse variable number tandem repeat loci in the E. coli O157:H7 and O55:H7 genomes for high-resolution molecular typing. J. Appl. Microbiol. 98:928-949.
Leverstein-van Hall, M. A., A. T. Box, H. E. M. Blok, A. Paauw, A. C. Fluit, and J. Verhoef. 2002. Evidence of extensive interspecies transfer of integron-mediated antimicrobial resistance genes among multidrug-resistant Enterobacteriaceae in a clinical setting. J. Infect. Dis. 186:49-56.
Leverstein-van Hall, M. A., A. C. Fluit, H. E. M. Blok, A. T. A. Box, E. D. J. Peters, A. J. L. Weersink, and J. Verhoef. 2001. Control of nosocomial multiresistant Enterobacteriaceae using a temporary restrictive antibiotic agent policy. Eur. J. Clin. Microbiol. Infect. Dis. 20:785-791.
National Committee for Clinical Laboratory Standards. 2000. Methods for disk susceptibility tests for bacteria that grow aerobically, 7th ed. NCCLS document M2-A7. National Committee for Clinical Laboratory Standards, Wayne, Pa.
Nemoy, L. L., M. Kotetishvili, J. Tigno, A. Keefer-Norris, A. D. Harris, E. N. Perencevich, J. A. Johnson, D. Torpey, A. Sulakvelidze, J. G. Morris, Jr., and O. C. Stine. 2005. Multilocus sequence typing versus pulsed-field gel electrophoresis for characterization of extended-spectrum beta-lactamase-producing isolates. J. Clin. Microbiol. 43:1776-1781.
Sagel, U., R. T. Mikolajczyk, and A. Kr?mer. 2004. Using mandatory data collection on multiresistant bacteria for internal surveillance in a hospital. Methods Infect. Med. 43:483-485.
Wright, M. O., E. N. Perencevich, C. Novak, J. N. Hebden, H. C. Standiford, and A. D. Harris. 2004. Preliminary assessment of an automated surveillance system for infection control. Infect. Control Hosp. Epidemiol. 25:325-332.
Eijkman-Winkler Institute for Microbiology, Infectious Disease and Inflammation, University Medical Center Utrecht, Utrecht, The Netherlands(Maurine A. Leverstein-van)
A hospital-wide increase in the number of patients with aminoglycoside-resistant Enterobacter cloacae (AREC) isolated from clinical cultures was detected in December 2002 using a classical surveillance system (CSS). CSS refers to a strategy based on the recognition of an increased incidence of a species with a particular antibiogram at certain wards in a limited period. Since clonal spread was suspected, hospital records were reviewed for E. cloacae culture-positive patients. Based upon genotyping of 139 clinical E. cloacae isolates from 80 patients, it was concluded that 53 patients had had clinical cultures with a single AREC clone since April 2001. Determinants for unnoticed spread were investigated retrospectively, as was the possibility that a computer-assisted surveillance method would have detected this outbreak at an earlier stage. Determinants associated with late detection of clonal spread were the following: (i) the absence of a hospital-wide increase in incidence of E. cloacae cases for 1.5 years, (ii) the long time interval between cases, (iii) the hospital-wide occurrence of new cases, due to a high number of patient transfers between wards, (iv) the large variety of clinical sites, and (v) the high variability of antibiograms (n = 33). Retrospective application of a recently described computer-assisted surveillance method as well as an "in-house"-developed algorithm resulted in earlier detection of the outbreak of 6 and 12 months, respectively. These findings suggest that computerized tools for surveillance may recognize resistance trends that are too complex to be detected by manual review and indicate the need for prospective evaluation of such algorithms.
INTRODUCTION
In some countries up to 70% of nosocomial infections are caused by microorganisms resistant to at least two antimicrobial agents (6). Infections caused by such pathogens have been associated with increased morbidity, higher mortality, and raised costs for medical care (3). Therefore, it is critical that dissemination of multidrug-resistant microbes is contained. Surveillance of nosocomial infections and antibiotic resistance are important tools for limiting dissemination of antimicrobial resistance (1, 3). Outbreaks are usually recognized upon an increased incidence of specific infections caused by a strain characterized by a particular resistance pattern within a limited time period in a specific ward. This approach could be defined as the classical surveillance strategy. Such combinations of events may be recognized by manual or computerized review of laboratory and hospital data or by an astute clinician or clinical microbiologist.
In January 2003, widespread nosocomial dissemination of multidrug-resistant Enterobacter cloacae in our hospital became apparent. In retrospect, this pathogen was associated with nosocomial infections in 53 patients since April 2001 and possibly even more in the period before. Determinants for unnoticed spread were investigated retrospectively. A recently described computer-assisted surveillance method and an "in-house"-developed method were retrospectively applied to the database of the microbiology laboratory to determine whether this unnoticed outbreak could have been identified at an earlier stage (5).
MATERIALS AND METHODS
Setting, patients, and bacterial strains. The University Medical Center Utrecht in The Netherlands is a 1,042-bed tertiary care hospital with all medical disciplines represented. Each discipline has its own department consisting of different ward(s) and units often sharing the same staff. In addition, there are four adult intensive care units (ICUs) of which three are located next to each other and another is located on a different floor.
At the end of 2002, a hospital-wide outbreak of tobramycin-resistant Enterobacter cloacae (TREC) was recognized, and laboratory records from 1 January 2001 to 28 January 2003 were reviewed to identify patients with aminoglycoside-resistant Enterobacter cloacae (AREC) isolated from clinical cultures. Almost all AREC isolates had been stored and were available for further analysis. In addition, 56 randomly chosen aminoglycoside-susceptible E. cloacae isolates (ASEC), obtained from sterile sites between 1 January 2001 and 1 July 2003, were processed for genotyping.
Identification, antimicrobial susceptibility, and molecular typing. Identification and susceptibility testing were performed using an automated system and software (Phoenix Automated Microbiology System, Becton Dickinson Biosciences, Sparks, MD). Breakpoints were those recommended by the National Committee for Clinical Laboratory Standards (10). Aminoglycoside resistance was confirmed by using Etests (AB Biodisk, Solna, Sweden). Susceptibility patterns of all genotyped isolates were analyzed. Ten antimicrobial agents were included: gentamicin, tobramycin, trimethoprim-sulfamethoxazole, trimethoprim, amikacin, ciprofloxacin, tetracycline, chloramphenicol, nitrofurantoin, and meropenem. Penicillins and cephalosporins were not included, since beta-lactam resistance in E. cloacae often results from potentially reversible induction of the chromosomally located AmpC genes. Consequently, beta-lactam resistance is not used for recognition of epidemic strains. From each patient the isolates with a unique susceptibility pattern were typed by pulsed-field gel electrophoresis analysis (PFGE) as described previously, with the exception that thiourea (50 uM) was used routinely in the running buffer (8). Gels were normalized with 5 references for every 22 isolates tested on each gel. The Jaccard coefficient of similarity was calculated, and the unweighted pair group method with arithmetic averages (UPGMA) was used for cluster analysis in BioNumerics version 2.5 (Applied Maths, Sint-Martens-Latem, Belgium). The optimization was set at 1%, and the tolerance was set at 2% for the smaller fragments and 4% for the larger fragments to cope with interassay variation. Subsequently, the UPGMA clustering was visually checked.
Surveillance policy of the Department of Hospital Hygiene and Infection Prevention. During the period of the study, surveillance of hospital infections in our hospital included (i) active surveillance by infection control professionals consisting of biannual hospital-wide point prevalence surveys for hospital-acquired infections, followed by unit-directed surveillance if indicated, and (ii) clinical microbiology laboratory-based daily surveillance directed at microbial species with specific resistance determinants (aminoglycoside-resistant) Enterobacteriaceae, extended-spectrum ?-lactamases (ESBL)-positive Enterobacteriaceae, vancomycin-resistant Enterococcus faecium, and methicillin-resistant Staphylococcus aureus. If cultured, these isolates were stored at –70°C and automatically reported to the Department of Hospital Hygiene and Infection Prevention. Subsequently, isolation precautions for the patients involved were taken. Once an infection control professional detected at a certain ward an increase in the number of patients with any of these marker microorganisms, genotyping was performed to detect clonal dissemination. Another reason for genotyping was the remark of a consultant about a possible increase in incidence of a particular species or resistance pattern. Written guidelines for decision making regarding typing of microorganisms are present for a limited number of types of microorganisms (i.e., vancomycin-resistant E. faecium, methicillin-resistant Staphylococcus aureus, and Pseudomonas aeruginosa for ICUs).
Computer-assisted surveillance. Recently, Hacek et al. developed two algorithm tools (based on increases of organism number over baseline) to analyze laboratory data for possible clonal dissemination (5). The first algorithm (2SD) defined an alert as two standard deviations (SDs) above the mean monthly number of isolates. The second algorithm (MI) defined an alert as either a 100% increase from the baseline organism number over 2 months or a 50% increase during a three-consecutive-month period. To determine whether the use of these algorithms would have resulted in an earlier recognition of the outbreak, the mean monthly number of E. cloacae isolates in our hospital was determined from 1995 to 2001. Duplicates, defined by Hacek et al. as the same organism with the same susceptibility pattern from the same patient from the same source from the same month, were removed from the database.
Furthermore, a recently in-house-developed algorithm tool was tested which defined an alert as a >200% increase above the median monthly number of new cases. This algorithm was applied to the first TREC isolate of each patient present in the database. To test this algorithm, the median of the monthly number of patients with a TREC isolate in our hospital was determined from 1995 to 2001. In addition, the two algorithms described by Hacek et al. were tested on this same data set.
RESULTS
Descriptive epidemiology. In September 2002 (weeks 36 to 38), an increased incidence of patients with E. cloacae, including TREC, was noticed at different surgical wards. Six isolates obtained from six different patients were genotyped, and PFGE restriction patterns of only two had >90% similarity. Therefore, no specific control measures were reinforced and no screening to detect carriers was performed. However, the hospital-wide incidence of patients with a TREC-positive culture remained high (Fig. 1). In December 2002, all AREC isolates stored since January 2001 were genotyped. These genotyping results revealed a large outbreak of one clonal lineage, involving 53 patients since April 2001. Basic infection prevention measures (including enforcement of adherence to aseptic practices and barrier precautions), a restrictive antibiotic policy, identification of possible environmental reservoirs, and screening for intestinal colonization of contact patients were implemented, as described before (9).
PFGE typing. In total, 168 isolates from 109 patients were analyzed. Seven isolates from six patients were repeatedly nontypeable. The 161 remaining E. cloacae isolates comprised 105 AREC isolates (60 patients) and 56 ASEC isolates (53 patients). Based on the results of the UPGMA clustering and the matching visual analysis of the PFGE restriction patterns, 89 isolates obtained from 53 patients belonged to 1 clonal lineage (called cluster I). Although there was some variation in restriction patterns, all isolates with 80% similarity in PFGE patterns were considered to belong to cluster I (Fig. 2). Preliminary data from molecular analyses demonstrated that at least 94% of the isolates harbored the same 180-kb plasmid containing genes encoding integrons, an ESBL, aminoglycoside resistance, and quinolone resistance, supporting the clustering found in PFGE analysis. This plasmid was not demonstrated for five isolates that had slightly different PFGE patterns in the lower-molecular-weight fragments (called cluster Ia), which were all susceptible to tobramycin (Table 1; Fig. 2).
Incidence and intervals. Monthly incidences of patients with E. cloacae isolated from clinical cultures varied from 0 to 9 (Fig. 1). In all, 349 patients had E. cloacae isolated from clinical cultures; 66 (20%) had TREC, of which 30 patients had ASEC isolates as well. The remaining 283 patients had only ASEC isolates.
From the monthly incidence curve, three findings emerge: (i) incidences of clinical TREC remained low until August 2002, with a single exception in January 2002; (ii) the outbreak strain did not increase the hospital-wide incidence of E. cloacae until September 2002; and (iii) in December 2002 and January 2003, the outbreak strain seemed to replace susceptible E. cloacae isolates.
Furthermore, the mean time interval between detection of cases was 3 weeks, ranging from 1 to 22 weeks.
Susceptibility patterns. Eighty-nine E. cloacae cluster I isolates (53 patients) expressed 33 different susceptibility patterns (Table 1). Of these antibiograms, 30 (91%) represented a tobramycin-resistant variant. Tobramycin-susceptible variants were detected only in genotype IA isolates (five patients).
Patient transfers. Patients colonized with E. cloacae cluster I were identified in 18 different wards or units belonging to 8 different divisions: (i) surgical division, 29 patients at five surgical wards, including 12 patients in the ICU surgery; (ii) neurodivision, 8 at 5 neurological and neurosurgical wards and 8 at the neurosurgical ICU; (iii) division internal medicine, 4 at the 3 medicine wards, including 2 patients on the internal ICU; and (iv) another four patients at four other wards. Forty-three of the fifty-three cases could be linked to each other by overlapping days of hospitalization on the same wards. Eight patients had overlapping days of hospitalization with other cases but were identified at wards where no previous case had been identified. Two cases were identified within 48 h of admission at new wards. To determine to what extent patient transfer between wards had contributed to spread of genotype 1, the hospital database was reviewed for transfer frequency of affected patients. This frequency was compared to frequencies of all hospitalized patients in 2001/2002, as well as in 1991/1992.
The average number of patient transfers between wards at the time of identification of colonization or infection with E. cloacae type I was 3.94 (range, 1 to 7). For the hospital population at large, average numbers of transfers nearly doubled in the last 10 years from 0.37 in 1991/1992 to 0.7 in 2001/2002, while the average duration of hospital stay decreased in the same time interval by 23% (12.5 to 9.6 days). The transfer rate between different departments, however, was very low in both periods.
Infection sites. The outbreak strain caused a variety of hospital-acquired infections, as defined by the criteria of the Centers for Disease Control and Prevention, in 53 patients. The strain was isolated from sputa or bronchoalveolar fluid (20 infections, 5 colonizations), urine (8 infections), catheter urine (2 infections, 4 colonizations), wounds (7 infections), cervix (1 colonization), and from the following sterile sites: abdominal fluid (3 infections), liquor or external ventricular drain (8 infections), blood (4 infections), and an aortic valve (1 infection).
Retrospective evaluation of computer-assisted surveillance on outbreak data. The average monthly number of E. cloacae isolates from clinical cultures in our hospital was 37.2 for the period of 1995 to 2001 (range of the monthly mean, 27.7 to 44.6; range of the monthly SD, 4.0 to 15.9). Application of the 2SD algorithm would have yielded a first alert in March 2002 after analysis of the data obtained in February 2002, while the MI algorithm would not have yielded any alert.
The median monthly number of patients with TREC isolated from a clinical culture in our hospital was 0.9 for the period 1995 to 2001 (range of the monthly median, 0 to 2; range of the monthly SD, 0.5 to 2.4). Application of our in-house algorithm towards the overall incidence of patients with TREC would have yielded an alert 6 months prior to the first alert of the surveillance system by Hacek et al. (Fig. 1). The 3 new cases in September 2001 (September: monthly median, 1.4; SD, 2.1) would have caused an alert by the in-house algorithm (threshold, 2.8), while the 2SD algorithm (threshold, 5.6) would have remained silent until March 2002. The MI algorithm would have yielded an alert after the 3-month interval August-September-October 2002, similar to outbreak recognition based upon the classical surveillance strategy.
DISCUSSION
Our experience with a multidrug-resistant strain of E. cloacae exemplifies the pitfalls encountered in detecting potential outbreaks in a hospital setting. Classically, surveillance for detection of clonal dissemination is based on the recognition of an increased incidence, in a limited period of time, of a certain strain, characterized by a particular resistance pattern in a certain ward, preferentially causing specific infections. Here, we were confronted with prolonged clonal dissemination of a multidrug-resistant strain of E. cloacae with 33 different antibiograms, initially without a manifest increase in total numbers of E. cloacae infections at multiple sites, affecting patients in 18 different wards.
The fact that consecutive cases were identified in different wards or even divisions could partially be explained by the high number of transfers of patients between different wards, being nearly six times as high as the average of hospital-wide number of transfers. The average number of within-hospital transfers nearly doubled in the last 10 years, whereas average length of stay significantly decreased in that period, implicating a substantial increase in patient movement through the hospital. This will catalyze dissemination of bacterial clones through the hospital and hamper rapid recognition of clonal spread.
Another reason for the late detection was the absence of a characteristic susceptibility pattern. Although tobramycin resistance was highly prevalent among outbreak isolates, the patterns of susceptibility to other antimicrobial agents were extremely diverse, further hindering recognition of clonal relatedness between different isolates. To our knowledge, this is the first time that such a high variety in susceptibility patterns within one clonal lineage has been described. Whether this is an emerging trait among nosocomial Enterobacteriaceae in general or is unique to this strain remains to be settled. The molecular basis for the high variety of antibiograms in this strain is currently under study. Horizontal gene transfer seems to play a role. Preliminary results show that the majority of the outbreak isolates carried a large plasmid containing genes encoding integrons, an ESBL, aminoglycoside resistance, and quinolone resistance. In addition to genetic differences within this clonal lineage, natural variations in antibiotic susceptibilities may have caused to some extent the high variety of antibiograms. Therapeutic (CLSI [formerly NCCLS]) breakpoints were used to distinguish resistant and nonresistant isolates in our study. Yet these breakpoints do not necessarily correspond with the optimal MICs for distinguishing between "susceptible" (wild-type) and "resistant" populations. Therefore, some of the observed shifts in susceptibility might be "artificial." Routine antibiotic susceptibility testing of clinical isolates over a wide range of concentrations might, therefore, be preferred. However, the automated susceptibility testing systems available for routine clinical laboratories, such as the Phoenix used in this study, test only a limited range of antimicrobial concentrations around therapeutic breakpoints. Therefore, susceptibility data generated by the clinical laboratory should be interpreted with caution when used for surveillance purposes.
Computerized tools for surveillance might improve our practice (2, 5, 12, 13). Retrospective application of the computer-assisted surveillance system using the algorithms and criteria of Hacek et al. would have produced an alert 6 months before the outbreak was actually detected. The method of Hacek et al. was directed at detection of a hospital-wide increase in the incidence of certain microorganisms, without taking susceptibility patterns, locations, or number of isolates per person into account. Only exact duplicates are removed.
In our opinion, analysis of microbiological data at the patient level of specific genera or species with certain resistance markers prone to causing nosocomial outbreaks would be a more efficient method of surveillance for detecting dissemination of microorganisms between patients. Therefore, an in-house-developed algorithm was retrospectively applied to the first clinical TREC isolate per patient, which resulted in an alert 12 months before the outbreak was actually detected. Application of the more-complicated algorithms of Hacek et al. towards this same data set resulted in an alert 6 months before the actual detection of the outbreak. As a consequence, we prefer the in-house-developed algorithm and currently are evaluating this algorithm for prospective use in surveillance prospectively.
This approach, directed at the combination of specific genera or species with certain resistance markers prone to causing nosocomial outbreaks, is in accordance with the conclusion of Gardam et al. that efficient surveillance might be achieved by performing routine multidrug-resistant Enterobacteriaceae surveillance of clinical isolates (4). This conclusion was made after they had shown that performing multidrug-resistant Enterobacteriaceae surveillance in the absence of an outbreak was a costly exercise that provided little or no benefit for infection control or predicting clinical infection.
Further improvement in the accuracy of the surveillance system for identifying outbreaks could be achieved by using department-specific instead of hospital-wide databases. Patients were transferred frequently between wards or units within the same department (e.g., within the surgical department containing five wards) but only sporadically between departments (e.g., from surgery to internal medicine). This hypothesis, though, needs to be empirically tested.
Well-defined criteria for deciding whether genotyping to assess microbial clonality should be performed are lacking. For instance, should genotyping be performed after every alert, or should different criteria for different species with different resistance traits be used? Our findings reveal that different antibiograms, different sources, long time intervals between cases, and isolation from different wards do not exclude the possibility of clonal spread. Naturally, genotyping can be helpful only if sufficient relevant isolates have been stored. Since it is impossible to store all clinical isolates, choices must be made. These choices should be based on the local ecology and susceptibility patterns with a low threshold to respond to ecological changes. The outbreak described here could be analyzed only because all aminoglycoside-resistant Enterobacteriaceae had been routinely stored.
Our findings also demonstrate the possibility of substantial variation of restriction patterns within a single clone when using PFGE. This finding emphasizes the need to include an adequate number of control isolates in the database to prevent variants of the same clone from being wrongly interpreted as unrelated. Alternatively, a more robust and less labor-intensive typing schemes could enhance the capacity to identify outbreaks, even in nonresearch settings. A recently developed typing scheme based on multiple loci as multilocus sequence typing or multiple locus variable number of tandem repeat analysis would fulfill these requirements (7, 11).
REFERENCES
Association for Professionals in Infection Control and Epidemiology. 1996. Overview of isolation systems, appendix C-6. In R. N. Olmsted (ed.), APIC infection control and applied epidemiology. Mosby Yearbook, Inc., St. Louis, Mo.
Bouam, S., F. Girou, C. Brun-Buisson, H. Karadimas, and E. Lepage. 2003. An intranet-based automated system for surveillance of nosocomial infections: prospective validation compared with physicians' self-reports. Infect. Control Hosp. Epidemiol. 24:51-55.
Burke, J. P. 2003. Infection control—a problem for patient safety. N. Engl. J. Med. 348:651-656.
Gardam, M. A., L. L. Burrows, J. V. Kus, L. D. E. Low, J. M. Conly, and A. Humar. 2002. Is surveillance for multidrug-resistant Enterobacteriaceae an effective infection control strategy in the absence of an outbreak? J. Infect. Dis. 186:1754-1760.
Hacek, D. M., R. L. Cordell, G. A. Noskin, and L. R. Peterson. 2004. Computer-assisted surveillance for detecting clonal outbreaks of nosocomial infection. J. Clin. Microbiol. 42:1170-1175.
Haley, R. W., D. H. Culver, J. W. White, W. M. Morgan, T. G. Emori, V. P. Munn, and T. M. Hooton. 1985. The efficacy of infection surveillance and control programs in preventing nosocomial infections in US hospitals. Am. J. Epidemiol. 12:182-205.
Keys, C., S. Kemper, and P. Keim. 2005. Highly diverse variable number tandem repeat loci in the E. coli O157:H7 and O55:H7 genomes for high-resolution molecular typing. J. Appl. Microbiol. 98:928-949.
Leverstein-van Hall, M. A., A. T. Box, H. E. M. Blok, A. Paauw, A. C. Fluit, and J. Verhoef. 2002. Evidence of extensive interspecies transfer of integron-mediated antimicrobial resistance genes among multidrug-resistant Enterobacteriaceae in a clinical setting. J. Infect. Dis. 186:49-56.
Leverstein-van Hall, M. A., A. C. Fluit, H. E. M. Blok, A. T. A. Box, E. D. J. Peters, A. J. L. Weersink, and J. Verhoef. 2001. Control of nosocomial multiresistant Enterobacteriaceae using a temporary restrictive antibiotic agent policy. Eur. J. Clin. Microbiol. Infect. Dis. 20:785-791.
National Committee for Clinical Laboratory Standards. 2000. Methods for disk susceptibility tests for bacteria that grow aerobically, 7th ed. NCCLS document M2-A7. National Committee for Clinical Laboratory Standards, Wayne, Pa.
Nemoy, L. L., M. Kotetishvili, J. Tigno, A. Keefer-Norris, A. D. Harris, E. N. Perencevich, J. A. Johnson, D. Torpey, A. Sulakvelidze, J. G. Morris, Jr., and O. C. Stine. 2005. Multilocus sequence typing versus pulsed-field gel electrophoresis for characterization of extended-spectrum beta-lactamase-producing isolates. J. Clin. Microbiol. 43:1776-1781.
Sagel, U., R. T. Mikolajczyk, and A. Kr?mer. 2004. Using mandatory data collection on multiresistant bacteria for internal surveillance in a hospital. Methods Infect. Med. 43:483-485.
Wright, M. O., E. N. Perencevich, C. Novak, J. N. Hebden, H. C. Standiford, and A. D. Harris. 2004. Preliminary assessment of an automated surveillance system for infection control. Infect. Control Hosp. Epidemiol. 25:325-332.
Eijkman-Winkler Institute for Microbiology, Infectious Disease and Inflammation, University Medical Center Utrecht, Utrecht, The Netherlands(Maurine A. Leverstein-van)