Enterococci Resistant to Multiple Antimicrobial Agents, Including Vancomycin: Establishment of Endemicity in a University Medical Center

Methods

Data were collected from a university hospital (557 beds), cancer center (62 beds), and adjacent shock trauma center (138 beds). The university hospital serves both as a state and regional tertiary referral center and the primary source of care for the immediate West Baltimore community. During the study period (May 1992 to June 1994), 59 196 patients were admitted to these institutions.

Beginning in May 1992, basic demographic and clinical data were collected on all patients from whom vancomycin-resistant enterococci were isolated. Standard CDC definitions were used to differentiate active infection from colonization [14]. Data on blood isolation of methicillin-resistant Staphylococcus aureus, coagulase-negative staphylococci, and vancomycin-susceptible enterococci were obtained by review of data from the hospital microbiology laboratory. Computerized pharmacy databases were used to determine the number of doses administered of selected antimicrobial agents, including vancomycin, ciprofloxacin, ceftriaxone, and ceftazidime.

Point-Prevalence Surveys

To determine the rate of the colonization of patient stool with vancomycin-resistant enterococci, we did a series of point-prevalence surveys. For each survey, we selected a random 30% sample of all persons who were on the inpatient census as of midnight on the dates selected for study (25 July, 16 August, and 27 September 1993). During the 3 days after each randomization date, efforts were made to collect a stool sample for culture from each patient selected in the randomization. Stool samples were also obtained from normal, healthy community volunteers who had been recruited as outpatients for unrelated vaccine trial studies.

Case-Control Studies

Two case–control studies were done in October 1993. In the first study, patients on the surgical intensive care and surgical intermediate care units from whom vancomycin-resistant enterococci had been isolated between May 1992 and October 1993 (case-patients) were compared with an unmatched random sample of patients who had been hospitalized on the same units during the same time period (controls). Controls could only be patients who had been hospitalized for at least 7 days (the minimum number of days between hospital admission and the isolation of vancomycin-resistant enterococci from a case-patient).

In the second study, all adult patients with urine cultures positive for nosocomially acquired vancomycin-resistant enterococci who were identified between May 1992 and October 1993 (case-patients) were compared with an unmatched random sample of patients with urine cultures positive for nosocomially acquired vancomycin-susceptible enterococci (controls).

Both studies included colonized and infected case-patients. Age; sex; diagnoses; and data on instrumentation, antimicrobial therapy, and severity of illness (measured by the Acute Physiologic and Chronic Health Evaluation II [APACHE II] score [15]) were obtained by chart review. For case-patients, data reflected events occurring before the positive culture was collected. For controls, data represented either the entire time the patient was in the intensive care or intermediate care unit (surgical intensive care unit study) or the hospital stay until the positive culture for vancomycin-sensitive enterococci (urinary tract infection study); the midpoint of the intensive care unit or hospital stay was designated as the “day of positive culture” for comparison with case-patients.

Observational Studies

Patients with vancomycin-resistant enterococci were placed in “multidrug-resistant organism” isolation, similar to the CDC's recommended contact isolation [16]. From 1 November to 14 November, compliance with isolation procedures was seen on four wards: two general medical and surgical wards, the surgical intensive care unit, and the surgical intermediate care unit. For twenty 30-minute periods, each room was observed and the occupation, duties, and patient or environment contacts of each person entering the room were recorded. We noted whether each patient's room had a sign indicating isolation for a multidrug-resistant organism, whether gowns and gloves were readily accessible outside the room, whether each person washed his or her hands before and after the visit, and whether gowns and gloves were worn in situations requiring their use.

As part of these studies, hand cultures were obtained from health care workers on these four wards by having volunteers rub a sterile hand wipe moistened with 0.02% polysorbate 80 detergent over their hands for at least 30 seconds [17]. Selected environmental cultures were obtained by using sterile swabs moistened with sterile saline. Environmental and hand-wipe specimens were processed at the CDC and were cultured by using the membrane filter technique [17, 18]. Membrane filters were then cultured on m-Enterococcus agar [18].

Interventions

Infection Control Measures

Beginning in October 1993, routine admission and weekly stool surveillance cultures for vancomycin-resistant enterococci were obtained from all patients in the surgical intensive care and intermediate care units. Patients colonized or infected with vancomycin-resistant enterococci in these units were spatially segregated, and nurses caring for the patients were placed in cohorts. Nurses caring for culture-positive patients in the cancer center were also placed in cohorts.

Restrictions on Vancomycin Use

Beginning in December 1993, vancomycin use was restricted to 1) treatment of documented, culture-proven infections with organisms not susceptible to alternative agents [that is, methicillin-resistant S. aureus or coagulase-negative staphylococci]; 2) treatment of patients in whom there was a high index of suspicion of methicillin-resistant staphylococcal infections on the basis of history, surveillance cultures, or Gram stain; 3) treatment of Clostridium difficile infections in patients who had not responded to an initial course of therapy with metronidazole; 4) treatment of gram-positive infections in patients with documented, severe allergy to β-lactam antimicrobial agents; 5) single doses of vancomycin just before hospital discharge in anephric patients with gram-positive infections; and 6) surgical prophylaxis as part of a previously approved investigational protocol for recipients of solid organ transplants. For all other uses, the patient's attending physician was required to obtain permission from the infectious disease attending physician on service.

Compliance with this policy was monitored by the Department of Pharmacy Services. Vancomycin was not dispensed unless an appropriate indication was documented or the infectious disease attending physician approved use of the drug. If inappropriate therapy was identified, recommendations were made for the use of alternative agents.

Microbiological Analysis

Enterococci were identified by using standard microbiological methods, including hydrolyzing esculin and growth in 6.5% NaCl [19]. Identification of the organisms to the species level was done using the conventional test scheme defined by Facklam and Collins [20]. Antimicrobial susceptibilities to ampicillin, vancomycin, streptomycin, and gentamicin were determined by using the E-test quantitative minimum inhibitory concentration procedure (AB Biodisk, Solna, Sweden). Quality control strains of E. faecalis (ATCC 29212, 51299) were used to ensure the potency of each antimicrobial agent tested. With the exception of vancomycin, susceptibility interpretations followed the guidelines proposed by the National Committee for Clinical Laboratory Standards [21, 22]. For this study, vancomycin resistance was defined as any enterococcal isolate with a minimum inhibitory concentration to vancomycin of at least 16 µg/mL. Vancomycin resistance was confirmed by hybridization to specific gene probes (see below).

We used trypticase soy agar supplemented with 5% sheep blood (BBL, Cockeysville, Maryland) to isolate enterococci from urine, wounds, and sterile body fluids. For stool samples collected as part of the point-prevalence studies and for routine stool surveillance for vancomycin-resistant enterococci, samples were plated on a special selective media (colistin nalidixic acid agar [Difco Laboratories, Detroit, Michigan] supplemented with defibrinated sheep blood (5%), vancomycin (10 µg/mL), and amphotericin [1 µg/mL]). The utility of this media for isolation of vancomycin-resistant enterococci has been previously reported [23].

Selected isolates of vancomycin-resistant enterococci were further characterized by contour-clamped homogeneous electric field electrophoresis after digestion of chromosomal DNA with SmaI [24, 25]. Electrophoretic studies were repeated with digestion with ApaI for strains that appeared to be identical or closely related (differing by one to three bands [26]) after initial analysis. The specific vancomycin-resistant genotype (vanA, vanB, or vanC) was determined with polymerase chain reaction analysis by using specific primers selected from published gene sequences [27].

Statistical Analysis

Analyses of the point-prevalence studies were based on the union of data from three separate sources: the inpatient census at midnight on the days selected for study; a pharmacy database that included records of all antimicrobial agents prescribed during July, August, and September 1993 and April, May, and June 1994; and results of tests of the fecal samples from the randomly selected patients for vancomycin-resistant enterococci. For each randomization date, the patients not sampled, the patients sampled but providing no specimens for testing, and the patients sampled and providing specimens were compared on various demographic and medical factors by using chi-square analysis derived from SAS contingency tables (SAS Institute, Cary, North Carolina). The same methods were then used to compare the characteristics of sampled patients who had positive or negative cultures for vancomycin-resistant enterococci.

In the case–control studies, data were analyzed by using Epi Info version 5.01b software [28]. Categorical variables were compared by either the Fisher exact test or the chi-square test; odds ratios and 95% CIs were also calculated. Continuous variables were compared using the Student t-test or the Wilcoxon rank-sum test. Multivariate analysis was done using PC SAS Proc Logistic (SAS Institute). A stepwise procedure was used to select from among the variables found in univariate analysis to be risk factors for infection and colonization with vancomycin-resistant enterococci; first-order interaction terms were evaluated. Final models included variables significant at P < 0.1 and any covariable that substantially changed a main-effect regression coefficient. The appropriateness of using certain variables on a continuous scale was evaluated by testing a squared term and indicator variables coded for varying levels of the continuous variable (typically with three or four levels). For models in which continuous variables were included, a good fit was obtained with data entered continuously and a linear relation between the logistic function was found. All P values are two-tailed.

Results

Between May 1992 and June 1994, 75 patients with active infections attributable to vancomycin-resistant enterococci were identified (Figure 1). Isolates from 63 patients were available for further analysis; 86% were E. faecium (Table 1). Data on antimicrobial susceptibility are summarized in Table 2.

The primary sources of isolation in the 75 infected patients were blood (31 patients) and urine (21 patients). The 31 blood isolates represented 19% of all nosocomial enterococcal blood isolates (both vancomycin-susceptible and -resistant) recovered during this period. Six of the patients with bacteremia caused by vancomycin-resistant enterococci died; all had serious underlying illnesses. Rates of infection were highest among patients on the organ transplantation service (13.2 infections/1000 admissions); there were 5.6 infections/1000 admissions in the surgical intensive care unit, 4.8 infections/1000 admissions in the medical intensive care unit, and 1.8 infections/1000 admissions on all internal medicine services combined.

Vancomycin-resistant enterococci were isolated from an additional 209 patients who were colonized but showed no evidence of active infection (73 patients) or who were part of our routine stool surveillance (86 patients) or point-prevalence studies (50 patients).

Initial Point-Prevalence Studies

In the point-prevalence studies done from July to September 1993, 20% of patients sampled were colonized with vancomycin-resistant enterococci (Table 3). However, patients from whom stool samples were obtained were not representative of the whole hospital population: They tended to have been in the hospital longer than patients from whom samples were not obtained (stool samples were obtained from 46% of patients hospitalized for more than 3 days before randomization and from 14% of patients hospitalized 3 days or less), were younger, and were more likely to be on the pediatrics service. In contrast, the obstetrics service was under-represented. When prevalence data were weighted according to age, length of stay, and the medical service distribution of the target inpatient population, the resulting estimate of the prevalence of vancomycin-resistant enterococci infection was 16.9% (95% CI, 10.2% to 23.7%). Hospitalized patients were significantly more likely to have vancomycin-resistant enterococci isolated from stool samples than were healthy community volunteers (25 of 125 hospitalized patients compared with 0 of 21 community volunteers; P = 0.03).

When culture-positive and -negative patients were compared, isolation of vancomycin-resistant enterococci was associated with older age (P = 0.012), and a borderline association was seen with longer hospitalization (P = 0.051). Culture-positive and -negative patients did not differ by sex, hospital service, or pattern of antimicrobial administration before the date of randomization. In particular, no association was found between receipt of either oral or intravenous vancomycin or third-generation cephalosporins and colonization with vancomycin-resistant enterococci (relative risk for vancomycin-resistant enterococci colonization after receipt of vancomycin, 1.3; CI, 0.57 to 3.1).

Enterococcus faecium accounted for 13 (52%) of the 25 vancomycin-resistant enterococcal isolates. Forty percent of isolates were E. gallinarum, but this organism was not isolated from patients identified as having active infections with vancomycin-resistant enterococci (Table 1) (P < 0.0001). Results were unaffected by subgroup analysis by species.

Case-Control Studies

Case-patients in the case–control study done in the surgical intensive care and intermediate care units were more likely than controls to have received antimicrobial agents. Case-patients also tended to be more seriously ill as measured by APACHE II scores (Table 4). In a multivariate model, variables with statistically significant adjusted odds ratios included receipt of vancomycin or ciprofloxacin, APACHE II score, and percentage of hospital days during which antimicrobial agents were received (Table 5).

Similar risk factors were identified in the case–control studies of urinary tract infection and colonization (Table 6). Variables that were statistically significant in a multivariate model included receipt of vancomycin and percentage of hospital days during which antimicrobial agents were received (Table 5).

Observational Studies

Of 20 patients placed in isolation because of colonization or infection with vancomycin-resistant enterococci, 19 had one or more health care workers enter his or her hospital room during the observation period. All rooms had a sign indicating that the patient was isolated for a multidrug-resistant organism; 18 had gloves readily available outside the room; and 15 had gowns available. Fifty-seven patient-health care worker interactions were recorded (range of interactions for each patient, 1 to 5; median, 3). In 25 (44%) of these interactions, one or more breaks in contact isolation technique were observed (range of breaks during each interaction, 1 to 4; median, 1). The nature of these breaks varied among the four wards studied. For example, on one general medical and surgical ward, none of the three rooms studied had gowns available, whereas in the surgical intensive care unit, failure to wash hands after patient contact and failure to dispose of gloves before leaving the room were the most common breaks in contact isolation technique.

Vancomycin-resistant enterococci were isolated from none of 29 health care worker hand cultures; vancomycin-susceptible enterococci were isolated from 3 (10%) of these cultures. Enterococci were isolated from 10 of 30 environmental cultures; 4 of these 10 isolates were vancomycin-resistant (2 were E. faecium and 2 were E. faecalis). These resistant isolates were isolated, respectively, from electrocardiograph wires, ventilator tubing, a bedside stand, and an automated medication dispenser serving the entire surgical intensive care unit.

Interventions

In October 1993, collection of routine admission and weekly surveillance cultures for vancomycin-resistant enterococci was initiated in the surgical intensive care and intermediate care units. Patients found to be infected or colonized with the organism were geographically separated (placed in predesignated sections of the intensive care units), and nurses caring for these patients were placed in cohorts. Members of the nursing staff in the cancer center were also placed in cohorts.

Restrictions on the use of vancomycin were imposed on 1 December 1993. Oral vancomycin use decreased from a mean ±SD of 1602 ±490 doses per month in the 11-month period immediately before implementation of these restrictions to 238 ±310 doses per month in the 7 months after the initiation of the policy (85% decrease [CI, 71% to 99%]). Intravenous vancomycin use decreased from 4119 ±562 doses per month to 1698 ±443 doses per month (59% decrease [CI, 51% to 67%]) (Figure 2, top).

The restriction policy was successful at targeting intravenous vancomycin therapy to the appropriate indications. An analysis of 462 intravenous vancomycin orders written during the 7 months after initiation of the vancomycin restriction showed that 40% of the orders were used for culture-documented infections caused by β-lactam-resistant gram-positive organisms; 36% were used for infections in which there was a high index of suspicion of infection with these organisms; 6% were used for infections with gram-positive organisms in patients with allergies to β-lactam antibiotics; 5% were used for a protocol evaluating vancomycin prophylaxis in recipients of solid organ transplants; and 3% were used for infections caused by gram-positive organisms in anephric patients. In 49 instances (10%), use of vancomycin was not considered to be justified; in 43 [88%] of these 49 instances, patients were given an appropriate alternative agent. Use of other antimicrobial agents, including third-generation cephalosporins and ciprofloxacin, remained constant (Figure 2, bottom).

Repeated Point-Prevalence Surveys and Other Outcome Measures

Point-prevalence studies were repeated in April, May, and June 1994. The mean rate of stool colonization with vancomycin-resistant enterococci was 20% (Table 3). When prevalence data were weighted according to age, length of stay, and medical service distribution of the inpatient population at the time of the sampling, the estimated prevalence estimate was 18.7% (CI, 11.6% to 25.8%). Patients from whom vancomycin-resistant enterococci were isolated tended to be older than culture-negative patients (P = 0.002); length of stay was not significantly associated with colonization. Isolation of the organism was associated with receipt of vancomycin (4 of 21 culture-positive persons received vancomycin compared with 3 of 97 culture-negative persons [relative risk, 6.2; CI, 1.3 to 29.6]). A similar number of patients received ceftriaxone (no patient received both vancomycin and ceftriaxone); receipt of ceftriaxone was also associated with a positive culture for vancomycin-resistant enterococci (4 of 21 culture-positive persons received ceftriaxone compared with 3 of 97 culture-negative persons [relative risk, 6.2; CI, 1.3 to 29.6]).

In contrast to the initial point-prevalence studies, specimens obtained in the second study were also cultured for vancomycin-susceptible enterococci. Sixty (48%) of 126 stool samples submitted contained enterococci; 25 (42%) of the 60 enterococcal isolates were vancomycin-resistant. No vancomycin-resistant enterococci were isolated from 16 healthy community volunteers.

The number of infections with vancomycin-resistant enterococci did not notably decrease between December 1993 and June 1994 (the period after implementation of the above interventions) compared with the period before implementation of the interventions (Figure 1). No change was seen in the number of bloodstream infections with organisms for which vancomycin had routinely been used as prophylaxis, including methicillin-resistant S. aureus and coagulase-negative staphylococci.

Further Microbiological Analysis

Eighty-five isolates of vancomycin-resistant enterococci were examined by pulsed-field gel electrophoresis. These isolates featured at least 45 distinct patterns. Several small clusters of patients with identical electrophoresis patterns were identified; an investigation of the largest of these clusters, occurring on our pediatric ward, has been reported [29].

Twenty-five of the strains examined were from the point-prevalence studies. Twenty-four different banding patterns were apparent among these 25 strains (Figure 3). In one instance, identical banding patterns were produced by isolates from two different patients; these patients had been hospitalized in the same room 1 month apart.

Twenty-eight strains were screened by polymerase chain reaction for vancomycin resistance genes. Five strains had the vanA genotype, 6 had vanB1, and 8 had vanB2. Nine of the isolates were E. gallinarum; all of these had the vanC genotype.

Discussion

During the past decade, enterococci have emerged as important nosocomial pathogens: They are the second most common cause of nosocomial infections in the United States and are responsible for approximately 8% of all nosocomial bloodstream infections [30-32]. Even more ominous is the apparent rapid spread of vancomycin resistance among enterococci, particularly E. faecium, resulting in bacterial strains that cannot be effectively treated with currently available antimicrobial agents. Numerous reports have appeared on the serious infections and mortality associated with these strains, especially among immunosuppressed patients or those with underlying illnesses [11, 12, 33-36]. Fortunately, infection is not always fatal. Despite our inability to effectively treat these infections, only 6 of 20 patients with vancomycin-resistant enterococci isolated from their blood died.

Vancomycin-resistant enterococci frequently colonized the stool of our patients, a finding in agreement with data now emerging from other hospital centers [37, 38]. Interestingly, 40% of the vancomycin-resistant isolates recovered in the point-prevalence studies were E. gallinarum. This enterococcal species has traditionally been considered to be nonpathogenic [32], an assumption supported by our finding that, despite its frequent presence in stool samples, the organism caused no active infections. However, even if these isolates are excluded, 10% or more of our hospitalized patients were colonized with potentially pathogenic, vancomycin-resistant strains of enterococci. In the point-prevalence studies, older age was a consistent risk factor for colonization. The point-prevalence studies did not include an indicator for severity of illness; among other possibilities, older age and longer hospital stay (a significant variable in the initial point-prevalence studies) may be markers for more severely ill patients. We did not isolate vancomycin-resistant enterococci from healthy community volunteers, which suggests that resistant enterococci either are not widely distributed in the general (nonhospitalized) population or are present at levels below those that can be detected with our current methods.

We and others [29, 39, 40] have reported isolation of vancomycin-resistant enterococci from environmental surfaces. Although we did not isolate the organism from the hands of staff members, there is clearly the potential for transfer of the organism from patient to patient or from patient to environmental surface to patient [39, 41]. This is supported by results of our molecular typing studies, which indicated the presence of several small clusters, each caused by a distinct set of strains having a common pulsed-field gel electrophoresis pattern. However, in our institution, such clusters appeared to be the exception rather than the rule: Most isolates had unique electrophoretic banding patterns. Vancomycin resistance genes are known to be present on transmissible genetic elements (plasmids, transposons) [1, 8, 40], and it is possible that it is these genetic elements that were being transferred rather than the strains themselves. This hypothesis is associated with some difficulties, however, because the results of our polymerase chain reaction studies suggest that at least four different types of vancomycin resistance genes were present among our vancomycin-resistant strains. Taken together, these observations suggest that resistant isolates have multiple sources: These may include the hospital, other medical facilities (including nursing homes and other chronic care facilities), and environmental or food reservoirs [42-44].

In the face of what appeared to be an increasing rate of colonization with vancomycin-resistant enterococci, and in light of increasing concerns about the possible effect of this organism on our organ transplantation and oncology programs, we initiated vancomycin restrictions and placed nurses in cohorts in the fall of 1993; the procedures we followed were almost identical to those subsequently recommended by the CDC [45]. The resulting decrease in vancomycin use was substantial. Availability of gowns and gloves was good and, despite some breaks in isolation technique, overall compliance with infection control procedures was reasonable. These interventions may have prevented a further increase in the rate of colonization among our patient population or may have reduced colonization in certain high-risk groups. However, the interventions did not result in an overall decrease in colonization or in a clear reduction in the number of infections caused by vancomycin-resistant enterococci.

It is possible that we simply have not waited long enough. Although no data are available, the ability to see an effect on overall carriage rates may require consistent restrictions over a period of years. Even more stringent restrictions on the use of antimicrobial agents may be necessary to produce a decrease in colonization (and, presumably, a corresponding decrease in infections). In support of this latter hypothesis, we still saw an association between colonization of vancomycin-resistant enterococci and vancomycin and ceftriaxone administration in the second set of point-prevalence studies, which were done after “standard” vancomycin restrictions had been implemented. However, further reductions in vancomycin use in our institutions would be difficult, requiring major changes in currently accepted prescribing practices and potentially influencing standards of patient care. In the absence of further studies, we cannot comment on the utility of reducing the use of third-generation cephalosporins, ciprofloxacin (which was implicated as an independent risk factor for vancomycin-resistant enterococci in the surgical intensive care unit study), or other antimicrobial agents.

Currently, vancomycin-resistant enterococcal strains appear to have established stable endemicity within our patient population, with an adjusted prevalence estimate of 16% to 18%. Discussions with infectious disease physicians and infection control practitioners suggest that a similar pattern has emerged in other hospitals and in nursing homes in our area; recent reports suggest that similar patterns are also emerging in other metropolitan areas [10-12, 34-36] as this “epidemic wave” moves across the country. During the past two decades, we have watched the inexorable spread of methicillin-resistant S. aureus, despite similar interventions (and occasional successes), and every indication suggests that vancomycin-resistant enterococci will follow a similar pattern. Although current CDC recommendations for control of vancomycin-resistant enterococci have potential utility in the control of point-source outbreaks, we are not optimistic that their implementation will effectively block the spread of this organism. If we are to limit the transmission and carriage of vancomycin-resistant enterococci, we need further, more sophisticated epidemiologic studies to define routes by which the bacterium enters and is maintained within patient populations. At the same time, we need to reevaluate our standard hospital-based approaches to management of this and other multidrug-resistant organisms.