Vancomycin-Resistant Enterococci in Intensive Care Units: High Frequency of Stool Carriage During a Non–Outbreak Period

Background  We aimed to define the epidemiological associations of vancomycin-resistant enterococci (VRE) in intensive care units (ICUs) during a non–outbreak period by examining prevalence, risk factors for colonization, frequency of acquisition, and molecular strain types.

Design  A prospective cohort design was followed. Consecutive patient admissions to 2 surgical ICUs at a tertiary care hospital were enrolled. The main outcome measures were results of serial surveillance cultures screened for VRE.

Results  Of 290 patients enrolled, 35 (12%) had colonization with VRE on admission. The VRE colonization or infection had been previously detected by clinical cultures in only 4 of these patients. Using logistic regression, VRE colonization at the time of ICU admission was associated with second- and third-generation cephalosporins (odds ratio [OR]=6.0, P<.0001), length of stay prior to surgical ICU admission (OR=1.06, P=.01) greater than 1 prior ICU stay (OR=9.6, P=.002), and a history of solid-organ transplantation (OR=3.8, P=.021). Eleven (12.8%) of 78 patients with follow-up cultures acquired VRE. By pulsed-field gel electrophoresis, 2 strains predominated, one of which was associated with an overt outbreak on a non-ICU ward near the end of the study period.

Conclusions  Colonization was common and usually not recognized by clinical culture. Most patients who had colonization with VRE and were on the surgical ICU acquired VRE prior to surgical ICU entry. Exposure to second- and third-generation cephalosporins, but not vancomycin, was an independent risk factor for colonization. Prospective surveillance of hospitalized patients may yield useful insights about the dissemination of nosocomial VRE beyond what is appreciated by clinical cultures alone.

THE ENTEROCOCCUS is a significant nosocomial pathogen, currently ranked as the fourth most common cause of hospital-acquired infection by the National Nosocomial Surveillance System of the Centers for Disease Control and Prevention.¹ It possesses inherent resistance to several antimicrobial agents and exhibits a propensity to acquire resistance to others.² The emergence of vancomycin-resistant strains has brought particular attention to this organism and has led to intensive efforts at controlling its spread.^3,4 Despite these control efforts, hospitals in the United States have experienced a rapid increase in the percentage of vancomycin-resistant enterococci (VRE).^1,5-7 During the initial period of spread, the incidence of VRE was reported to be much higher among patients located in intensive care units (ICUs) compared with other wards.^1,8,9 Significant associations were also observed between VRE infection and exposure to vancomycin.^5,7,10 However, most studies of nosocomial VRE focused on patients identified only by clinical cultures,^11-14 or were of relatively small size.^8,15 Few studies have performed serial surveillance cultures in a large defined population.^16,17

To address these epidemiological associations, we conducted a prospective surveillance study of patients admitted to 2 surgical ICUs (SICUs) during a non–outbreak period. The goals were to (1) define prevalence of VRE at the time of ICU admission and incidence of acquisition during follow-up, (2) assess risk factors for baseline VRE colonization and acquisition, focusing on antibiotic exposures, and (3) assess patient-to-patient transmission of VRE by correlating molecular typing with epidemiological factors.

The Beth Israel Deaconess Medical Center, West Campus, in Boston, Mass, is a 420-bed tertiary care hospital with approximately 12,000 annual admissions. During the period of this study, there were 2 SICUs, (SICU-East and SICU-West) each with 8 single patient rooms and a nurse-patient ratio of 1:2, with occasional cross coverage. Patients were enrolled in the study between January 15, 1995, and June 15, 1995. This study was performed when the clinical detection of VRE was infrequent, approximately 2 cases per month in the entire hospital. At the time of the study, the prevalence of vancomycin-resistance among clinical enterococcal isolates was 8% to 9%.

Using cotton swabs (Starplex, Etoibicoke, Ontario), the following sites were cultured: rectum, oropharynx, skin, and gastric fluid if a nasogastric tube was present. Rectal swabs coated in stool were considered adequate specimens. The subset of rectal swabs was used for this analysis. Prior to specimen collection, verbal consent was obtained from the patient or family member.

Baseline surveillance cultures were obtained from patients within 24 hours of admission to the SICU. Follow-up cultures were collected from patients who remained in the SICU for 3 days or longer. Cultures were obtained at the following intervals: days 3 to 4, days 7 to 8, days 15 to 26, and every 2 weeks thereafter for as long as the patient remained in the SICU.

Demographic and clinical data were recorded on all patients on the day of their SICU admission, before culture results were available. Underlying diseases were classified according to the organ system involved. The Acute Physiology and Chronic Health Evaluation III score¹⁸ and Charlson criteria¹⁹ were used to measure the severity of illness of the patients admitted to the SICU and to classify co-morbid conditions, respectively. Computerized pharmacy databases were used to determine inpatient exposures to antibiotics. Chart review was used to identify additional demographic data.

Samples were initially processed to identify resistant gram-negative bacilli; then swabs were frozen in separate glass vials containing 2 mL of 0.6% glycerol. The analysis of resistant gram-negative rods in this population has been reported elsewhere.²⁰ To recover VRE, frozen rectal swabs were thawed, and plated on media containing colistin and nalidixic acid with (CNA) and without (CNV) vancomycin, 10 µg/mL (Remel, Lenexa, Kan). The swabs were sequentially inoculated onto CNA and then onto CNV plates. These plates were incubated at 35°C with 5% carbon dioxide and were examined for bacterial growth at 24 and 48 hours. Enterococci were identified by colony and Gram stain morphologic features and by standard biochemical tests.²¹ Enterococci from both CNA and CNV plates were screened for vancomycin resistance by plating on brain heart infusion media agar with 6-µg/mL vancomycin and then vancomycin resistance was confirmed by formal minimal inhibitory concentration (MIC) testing using the agar dilution method. Isolates with an MIC to vancomycin of 8 µg/mL or higher were classified as VRE. Isolates were speciated by use of API-Rapid Strep Strips (bioMerieux Vitex, Inc, Hazelwood, Mo), growth in motility media at 30°C (Remel), and evaluation of pigmentation. The MICs to antimicrobials including teicoplanin, ampicillin, doxycycline, chloramphenicol, and the new quinupristin-dalfopristin combination drug (Synercid; Rhone-Poulenc Rorer Pharmaceuticals Inc, Collegeville, Pa) were determined by serial dilution agar plate method using the National Committee for Clinical Laboratory Standards guidelines.²²

Baseline colonization on SICU admission was defined as the presence of VRE in the initial culture. Acquisition of VRE was defined by at least 1 positive follow-up culture if baseline cultures were negative. Recent antibiotic exposure was defined as use of antibiotics within 30 days prior to study entry. Patients were considered to have epidemiological links if they had overlapping hospital stays on the same floor or in the same room within 1 month prior to culturing. Clinical cultures were defined as clinically directed cultures of sites other than stool.

In patients with more than 1 SICU admission, only the first admission was analyzed. Patients whose initial rectal swab was obtained more than 3 days after SICU admission were also excluded.

Statistical analysis was performed using SAS software (SAS Institute, Cary, NC). Categorical and continuous variables were analyzed using χ² test and either the Student t test or nonparametric tests. Risk factors for colonization, demographics including underlying diseases, hospital and ICU exposures, and antibiotic exposures were evaluated by logistic regression using variables that were statistically significant on crude bivariate analysis (P ≤0.1) and those variables that were considered to be clinically relevant. Confounding was evaluated by assessing changes in β coefficients in models with and without putative confounders. A risk score model was developed using the covariates from the final regression model. The c statistic (area under the receiver operating characteristic curve) was calculated from this model. Survival analytic methods were used to examine acquisition of VRE.

Pulsed-field gel electrophoresis (PFGE) was performed on the first VRE isolate from each patient. Plugs were prepared using standard methods.²³SmaI restriction fragments were resolved using a CHEF-DRIII (Biorad, Richmond, Calif) at 6.0 V/cm for 22 hours with an initial switch time of 0.5 second and a final switch time of 35 seconds. After staining with ethidium bromide, gels were viewed and photographed under UV light. Indistinguishable isolates were assigned to the same subgroup. Isolates differing by 1 or 2 restriction bands, consistent with a single genetic event, were assigned to a subgroup within the same group.²⁴

Between January 15, 1995, and June 15, 1995, there were 416 admissions to the SICUs. One hundred twenty-six admissions were excluded from the study for the following reasons: (1) refusal and inability to collect rectal sample (44 patients), (2) SICU stay of less than 24 hours (63 patients), (3) prior admission to the SICU during same hospitalization (10 patients), and (4) other (9 patients). Thus, a study population of 290 patients was available for analysis (Figure 1).

The study patients had a mean age of 65.7 years; 55% were men and 45%, women. The most common reasons for admission to the SICU were postoperative care 195 (67.2%), cardiac disease 38 (13.1%), and gastrointestinal disorders 20 (6.9%). Eighty-two patients (28.3%) were admitted directly to the SICU and the other 208 patients were transferred to the SICU from other wards. The mean interval from hospital admission to SICU transfer was 5.1 days.

A total of 419 rectal swabs from the 290 study patients were cultured on both CNA and CNV media. Enterococci were identified on CNA plates from 134 swabs (112 patients). Thirty-eight of these enterococcal isolates (from 32 patients) were demonstrated to be vancomycin-resistant on the basis of growth on the vancomycin-containing brain heart infusion agar media plates and MIC testing; the other 96 isolates were vancomycin susceptible. In contrast, enterococci were recovered on CNV plates from 51 swab cultures (41 patients); 50 (40 patients) were confirmed to be VRE. In total, 56 cultures from 46 patients grew VRE. Therefore, the sensitivity for detection of VRE was 67.9% (38 of 56) and 89.3% (50 of 56) for CNA and CNV plates, respectively. Of the 46 patients with VRE, 6 were identified on CNA plates only.

The VRE isolates from 40 patients were Enterococcus faecium, 6 were enterococci species other than E faecium and Enterococcus faecalis. Thirty-nine had the VanA phenotype, 1 the VanB, and 6 the VanC phenotype. Antimicrobial susceptibility results were as follows: vancomycin MIC range: 8 to 512 µg/mL or higher (MIC₉₀≥128 µg/mL); teicoplanin MIC range: 0.05 to 128 µg/mL (MIC₉₀≥8 µg/mL); ampicillin MIC range: 0.5 to 128 µg/mL (MIC₉₀≥64 µg/mL); chloramphenicol MIC range: 2 to 32 µg/mL (MIC₉₀=8 µg/mL); doxycycline MIC range: 0.12 to 32 µg/mL (MIC₉₀≤8 µg/mL); and the combination drug quinupristin-dalfopristin MIC range: 1 to 4 µg/mL (MIC₉₀≤1 µg/mL).

Thirty-five (12.1% of the 290 study patients) were detected to be colonized with VRE on the initial culture. In 7 of these patients, organisms were cultured within 24 hours of admission to the hospital, whereas the other 28 patients with positive baseline cultures were transferred to the SICU from other wards. The 28 transferred patients with positive initial cultures came from 8 distinct non-SICU wards, the most common which was the surgical ward housing solid-organ transplant patients (10 patients). Six of 7 patients admitted directly to the SICU had either been transferred from outside healthcare facilities or had recent prior hospitalizations at our facility. Thus, only 1 patient with VRE colonization detected in the initial culture had no prior hospital exposure.

Four patients with positive baseline cultures had previous clinical cultures that grew VRE, from 1 to 301 days prior to study entry. The sites of positive clinical cultures were urine and wound. Only 2 patients with baseline colonization had subsequent positive clinical cultures during the study admission. One of these patients, a liver transplant recipient, developed an intra-abdominal infection due to VRE and received a prolonged course of treatment with the combination drug quinupristin-dalfopristin.

In the crude analysis, underlying diseases and measures of acuity that were associated with baseline VRE colonization included: history of a gastrointestinal tract disorder, a previous solid-organ transplantation, and a higher Acute Physiology and Chronic Health Evaluation score (Table 1). Indices of hospital exposure that correlated with VRE colonization included: a longer length of stay, hospitalization prior to this SICU admission, and multiple prior ICU stays (Table 1). Specific antibiotic exposures associated with VRE were metronidazole and second- or third-generation cephalosporins. The crude relative risk associated with intravenous vancomycin was 2.2 (P=0.06) (Table 2). None of the patients with baseline VRE colonization had received oral vancomycin.

Most antibiotic exposures were confounded by the duration of hospital stay prior to culturing (Table 3). This confounding was most severe for vancomycin: adjusting for length of stay reduced the β coefficient for vancomycin by 75%. Smaller degrees of confounding were seen when the crude and adjusted β coefficients were compared for other antibiotic categories (Table 3).

In the multivariable logistic regression model, the following 4 variables were identified as having the strongest independent association with VRE colonization: history of prior solid-organ transplantation (OR=3.8, P=.02), second- or third-generation cephalosporin treatment prior to ICU stay (OR=6.0, P=.001), duration of hospital stay prior to ICU admission, in days (OR=1.06, P=.01), and more than 1 prior ICU admission (OR=9.6, P=.002) (Table 4).

To examine the predictive value associated with these risk factors, patients were stratified into risk groups on the basis of the number of risk factors present. In this risk score model, length of stay was dichotomized at 6 days, the 75th percentile. The prevalence of colonization was 2.7% (4 of 143) in patients with 0 risk factors, 13% (13 of 97) in patients with 1 risk factor, and 39% (18 of 46 patients) in patients with 2 or more risk factors. The c statistic (area under the receiver operating characteristic curve) for this model was 0.79.

A total of 136 follow-up cultures were obtained from 84 patients. Including patients with both positive and negative baseline cultures, the proportion of patients with a positive culture increased from 13.9% on the first follow-up culture (days 3-4) to 30.4% on day 11 or later. Among 78 patients whose cultures were negative for VRE at baseline, 11 (14.1%) acquired VRE during follow-up. The Kaplan-Meier estimate of the 25th percentile of time to acquisition was 15 days. Vancomycin was the single most frequent antibiotic exposure in patients who acquired VRE: 8 of 11 patients who acquired VRE received intravenous vancomycin either prior to or during their SICU admission compared with 31 of 67 patients who did not acquire VRE. However, the hazard ratio (relative risk) associated with vancomycin was only 0.7 when length of stay in the SICU was controlled for using Cox proportional hazards regression. The small number of patients who acquired VRE precluded use of a multivariable survival model.

All 9 follow-up cultures from the 6 patients with positive baseline cultures were positive.

The PFGE was performed on the first VRE isolate from each of the 46 patients who had colonization with VRE. Sixteen distinct groups were seen of which 12 were unique (1 patient each). Two groups (designated I and II) accounted for 56.5% of isolates (10 and 16 isolates, respectively). Neither of these strains exhibited temporal clustering. Nonetheless, prior to their positive cultures, 3 of 10 patients with strain I and 4 of 16 patients with strain II were closely linked to another patient who had colonization with VRE with the same strain by virtue of an overlapping hospital stay on the same ward or ICU. Two patients with strain II were in the same hospital room prior to their SICU admission (lanes 2 and 3 on Figure 2). Four patients with strain II were culture positive within 24 hours of admission to the hospital but could not be epidemiologically linked to each other or to other patients harboring strain II prior to their index hospitalization.

Near the end of the study period (June 1995), an overt outbreak of VRE on the transplantation ward occurred, manifested by 5 patients with positive clinical cultures for VRE during a 16-day period. The PFGE typing of isolates from this outbreak revealed strain II to be the culprit.

We examined VRE colonization in a cohort of patients admitted to 2 SICUs during a period when the incidence of VRE in the hospital as detected by clinical cultures was low. Our major findings were (1) colonization with VRE had already occurred in 12% of patients at the time of admission to the ICU, and only 4 of the 35 patients who had colonization with VRE previously or concurrently had clinical cultures positive for VRE; (2) exposure to second- and third-generation cephalosporins but not vancomycin was an independent risk factor for colonization with VRE; (3) 2 predominant VRE strains were observed, suggestive of clonal dissemination.

Vancomycin has been identified as a risk factor for VRE colonization or infection in some but not all studies.^16,17,25,26 Our results demonstrate the dramatic confounding effect of length of hospital stay on the association between vancomycin and VRE, owing to the strong correlation between length of stay and VRE colonization and between length of stay and vancomycin use. Indeed, we have previously found that vancomycin is the single most common antibiotic administered to patients with long length of stay (unpublished data, 1995-1996). In contrast, adjustment for length of stay attenuated the association between VRE colonization and broad-spectrum cephalosporins to a much lesser degree. These results are compatible with other studies demonstrating a role of cephalosporins to promote VRE colonization and enterococcal superinfection.^15,27

Our study also illustrates the limitation of clinical cultures alone to define the true extent of VRE within hospital populations. The greater sensitivity of stool cultures compared with clinical cultures for detection of VRE colonization is well documented.^3,16,17 An unusual feature of our investigation is that the molecular typing revealed predominance of 2 strains at a time when hospital-wide clinical strains were diverse. A previous analysis of contemporaneous VRE isolates from 40 patients who had colonization or infection with VRE on the basis of clinical cultures had revealed 20 distinct PFGE restriction profiles.²⁸ Thus, the extent of transmission of VRE during the period of the study was significantly underestimated both by the low frequency of clinical isolation and by the inability to identify an outbreak strain among clinical isolates. The high prevalence of colonization on admission to the ICU was related, at least in part, to the unrecognized spread of VRE among patients on the transplantation ward who were subsequently transferred to the ICU. At the end of the study period, an overt outbreak detected by clinical cultures was manifested on this ward. Conceivably, if the surveillance culture results had been available during the course of the study, an intervention to reduce transmission could have been implemented and might have circumvented the clinical outbreak.

A potential limitation of our study was that rectal swabs were frozen prior to culturing for VRE. Therefore, it is possible that despite the high frequency of VRE isolation, the true prevalence of VRE colonization was actually greater than observed. However, other data have shown that enterococci remain viable after freezing and so it is unlikely that this procedure caused a significant decrease in sensitivity for detection of VRE.²⁹ We also attempted to maximize our yield of VRE by plating to both selective and nonselective media. Another limitation of this study was that an effective analysis of risk factors for acquisition of VRE was not possible owing to the small number of patients who acquired VRE after ICU admission.

We conclude that VRE acquisition among ICU patients frequently occurs prior to ICU admission and that prospective surveillance may yield useful insights into the dissemination of nosocomial VRE beyond what is appreciated by clinical cultures alone. In addition, these results suggest that screening of high-risk patients at the time of admission to the ICU to identify individuals in whom colonization with VRE had already taken place may be a useful infection control strategy. In concordance with several other studies, we were unable to establish vancomycin use as an independent risk factor for VRE colonization.^16,17 Further studies are needed to assess the impact of antibiotic control programs on the incidence of VRE.

Accepted for publication September 9, 1998.

We acknowledge partial funding provided by Postdoctoral Training grant 440 from the Centers for Diseases Control and Prevention, Atlanta, Ga.

Reprints: Belinda E. Ostrowsky, MD, Centers for Disease Control and Prevention, National Center for Infectious Diseases, Hospital Infections Program, MS-A07, Atlanta, GA 30333 (e-mail: bao6@cdc.gov).