Risk Factors for New Detection of Vancomycin-Resistant Enterococci in Acute-Care Hospitals That Employ Strict Infection Control Procedures

Abstract

Accurate assessment of the risk factors for colonization with vancomycin-resistant enterococci (VRE) among high-risk patients is often confounded by nosocomial VRE transmission. We undertook a 15-month prospective cohort study of adults admitted to high-risk units (hematology, renal, transplant, and intensive care) in three teaching hospitals that used identical strict infection control and isolation procedures for VRE to minimize nosocomial spread. Rectal swab specimens for culture were regularly obtained, and the results were compared with patient demographic factors and antibiotic exposure data. Compliance with screening was defined as “optimal” (100% compliance) or “acceptable” (minor protocol violations were allowed, but a negative rectal swab specimen culture was required within 1 week of becoming colonized with VRE). Colonization with VRE was detected in 1.56% (66 of 4,215) of admissions (0.45% at admission and 0.83% after admission; the acquisition time was uncertain for 0.28%), representing 1.91% of patients. No patients developed infection with VRE. The subsequent rate of new acquisition of VRE was 1.4/1,000 patient days. Renal units had the highest rate (3.23/1,000 patient days; 95% confidence interval [CI], 1.54 to 6.77/1,000 patient days). vanB Enterococcus faecium was the most common species (71%), but other species included vanB Enterococcus faecalis (21%), vanA E. faecium (6%), and vanA E. faecalis (2%). The majority of isolates were nonclonal by pulsed-field gel electrophoresis analysis. Multivariate analysis of risk factors in patients with an acceptable screening suggested that being managed by a renal unit (hazard ratio [HR] compared to the results for patients managed in an intensive care unit, 4.6; 95% CI, 1.2 to 17.0 [P = 0.02]) and recent administration of either ticarcillin-clavulanic acid (HR, 3.6; 95% CI, 1.1 to 11.6 [P = 0.03]) or carbapenems (HR, 2.8; 95% CI, 1.0, 8.0 [P = 0.05]), but not vancomycin or broad-spectrum cephalosporins, were associated with acquisition of VRE. The relatively low rates of colonization with VRE, the polyclonal nature of most isolates, and the possible association with the use of broad-spectrum antibiotics are consistent with either the endogenous emergence of VRE or the amplification of previously undetectable colonization with VRE among high-risk patients managed under conditions in which the risk of nosocomial acquisition was minimized.

In Australian hospitals, unlike those in the United States and Europe, colonization with vancomycin-resistant enterococci (VRE) and disease caused by VRE have only recently become significant nosocomial problems. They have occurred in the presence of a low background rate (0.2%) of fecal colonization with VRE in the general Australian community (25). Clonal outbreaks of VRE (mainly vanB strains) in some acute-care hospitals in 1998 prompted the widespread implementation in many Australian states of guidelines (based on those from the Centers for Disease Control and Prevention) to prevent its spread (4, 17, 20, 28; E. E. Cooper and M. O'Reilly, Abstr. 39th Intersci. Conf. Antimicrob. Agents Chemother., p. 643, 1999). Identification of the risk factors associated with colonization with VRE and disease caused by VRE has proved difficult (18). Many studies are potentially confounded by high rates of nosocomial acquisition of VRE, such that differentiation between the factors associated with endogenous emergence, amplification of previously undetectable colonization, or nosocomial spread of VRE has proved difficult. As a result, published rates of colonization with VRE among patients managed in acute-care hospitals appear to vary widely between countries and have variably suggested an association between the acquisition of VRE and recent use of antibiotics, especially vancomycin and/or broad-spectrum cephalosporins (6, 8, 23, 31, 32, 35). However, only a few of these studies have been prospective, and most have had some methodological flaws (19).

It is presumed that VRE may become detectable via one of three means: they may emerge in an individual patient as a result of antibiotic selective pressure and gene transfer into otherwise susceptible enterococci, they may be present in small numbers but become amplified to a detectable level under certain conditions, or they are acquired as a result of nosocomial spread. Thus, to more accurately assess the epidemiology of colonization with VRE in settings where nosocomial transmission is less likely, we elected to investigate the rate of colonization with VRE among high-risk patients in acute-care hospitals which used strict infection control and isolation procedures.

MATERIALS AND METHODS

Study sites and patient eligibility.

The study was performed in 11 high-risk units in three university teaching hospitals in Melbourne, Victoria, Australia (Alfred Hospital [AH], Austin and Repatriation Medical Centre [ARMC], and Monash Medical Centre [MMC]), during a 15-month period commencing in April 1998. Study units included the renal units (n = 3) and the intensive care units (ICUs; n = 5) at all three institutions (at AH this involved three ICUs, the general, road trauma, and cardiothoracic ICUs) as well as two transplant units (at AH, the heart-lung transplant unit; at ARMC, the liver transplant unit) and one hematology-oncology unit (MMC). The three hospitals were selected because they used identical infection control measures based on the guidelines of the Centers for Disease Control and Prevention. These included strict isolation measures for patients colonized with VRE to prevent cross-infection and screening for colonization among contacts with identified cases (4, 20).

Patients were eligible for participation in the study if they were aged >15 years and had been admitted to a study unit for more than 24 h. Basic demographic and health data were recorded, including age; sex; managing unit; and the presence of diabetes mellitus, immunosuppression (use of prednisolone, cyclosporine, azathioprine, mycophenolate, tacrolimus, or anticancer chemotherapy), transplantation, renal failure (serum creatinine level, ≥0.25 mmol/liter), or hepatic failure (clinical diagnoses were made by the treating physician). In addition, we gathered data regarding previous hospital admissions, the duration of the present hospital admission prior to entry into the study unit, and recent antibiotic administration. An episode in a patient who transferred between study wards was considered a single study episode, but episodes in a patient who was absent from a study ward for longer than 1 day and who was then subsequently readmitted to a study ward were recorded as two separate study episodes. Study end points were (i) identification of colonization or infection with VRE (Enterococcus faecalis or Enterococcus faecium) or (ii) departure from the study unit.

Antibiotic monitoring.

The use of antibiotics, especially the use of glycopeptides, broad-spectrum (ceftriaxone, cefotaxime, and ceftazidime), and newer (cefepime and cefpirome) cephalosporins, carbapenems, metronidazole, fluoroquinolones, gentamicin, ampicillin-amoxicillin, ticarcillin, or ticarcillin-clavulanic acid, was recorded prospectively by ward pharmacists during each admission to a study area. The number of days during which patients were likely to have detectable serum antibiotic concentrations following administration of the last dose of a monitored antibiotic were estimated by using the schedule outlined in Table 1. Thus, the total number of days of antibiotic exposure was calculated for each patient, with antibiotic exposures adjusted for renal impairment. Recent antibiotic exposure was considered to have occurred if antibiotics were administered either within 7 days of a study end point or if detectable serum antibiotic levels could be expected within this time period.

TABLE 1.

Predicted time of detectable serum antibiotic levels following dosing in patients with renal impairment^a

Drug	Serum half-life in patients with severe renal failure and undergoing hemodialysis	Predicted time after administration of last dose with detectable levels in serum
Vancomycin	18-20 days	7 days
Teicoplanin	102-347 h	7 days
Imipenem	3 h	Nil
Meropenem	>20 h	24 h
Ceftriaxone	12-24 h	48 h
Cefotaxime	3-11 h	24 h
Ceftazidime	13.5-19 h	24 h
Cefepime	5-13 h	24 h
Cefpirome	3-14 h	24 h
Metronidazole	7-21 h	Nil
Ciprofloxacin	4.4-12.6 h	24 h
Ticarcillin-clavulanate	4.9-8.5 h	Nil
Ampicillin	5-20 h	24 h
Gentamicin	20-60 h	48 h
Piperacillin-tazobactam	4-6 h	Nil

^aSources: British National Formulary, Australian Hospital Formulary System, and Bennett (2).

Protocol for culture of VRE.

The protocol for cultures of rectal swab specimens for VRE specified that a specimen for culture be collected by the managing nurse or doctor at the time of admission to the study (or within 2 days) and then every 7 days (±2 days). Given the busy clinical nature of the study units and in line with earlier studies, it was expected that adherence to this screening protocol was likely to be a challenge and that strict adherence to this screening protocol was likely to decline with an increasing duration of study participation. Hence, the adequacy of protocol adherence for each patient was classified according to the following definitions: “optimal” adherence was complete adherence to the swabbing protocol, and “acceptable” adherence was retrieval of a specimen for culture upon admission, followed by adherence to the protocol except for an occasional lack of collection of specimens for culture. In such cases there could be no greater than 14 (±2) days between cultures of rectal swab specimens (i.e., one missed culture), and for patients who were found to be rectally colonized with VRE, a negative rectal swab specimen for culture was required 7 (±2) days prior to the first positive culture for VRE. Thus, patients who were considered to have undergone “optimal” screening were a subset of those who were considered to have had “acceptable” screening. “Unacceptable” adherence was a protocol violation in excess of those defined as acceptable.

Only patients who underwent either optimal or acceptable screening were evaluated for factors that may potentially be associated with colonization or infection with VRE. All analyses were conducted with these study subpopulations. Patients were censored when they became colonized with VRE.

Microbiology cultures and clonality.

Rectal swabs were directly inoculated onto Enterococcosel agar (BBL, Cockeysville, Md.) containing 6 μg of vancomycin per ml and were incubated at 35°C for 48 to 72 h. Esculin-positive colonies (three isolates of each morphological appearance) were investigated and provisionally identified as E. faecium or E. faecalis on the basis of routine criteria (gram-positive cocci, l-pyrrolidonyl-β-naphthylamide hydrolase positive, nonmotile, catalase negative, and pigment negative) (15). Isolates fulfilling these criteria were assessed for susceptibility to vancomycin by the E-test (AB Biodisk, Dalvagen, Sweden). Isolates for which the vancomycin MIC was ≥2 μg/ml were analyzed for the presence of the vanA, vanB, vanC1, or vanC2-vanC3 gene by PCR; and isolates positive for the vanA or vanB gene were then confirmed to be either E. faecium or E. faecalis by PCR (1, 13). Isolates containing vanC genes were not assessed further or recorded.

The molecular clonality of all VRE isolates was assessed by pulsed-field gel electrophoresis (PFGE), and differences were compared according to the criteria described previously (22, 30). The PFGE patterns of the study isolates were compared with each other and with those of the clinical strains of VRE known to be most commonly responsible for colonization and infection in Melbourne (17, 24). All participating institutions followed identical screening and microbiology protocols.

Ethics approval.

The ethics committee at each of the three institutions approved the study protocol. Since the study protocol represented a relatively minor expansion of the infection control screening program already in place at each institution, only verbal patient consent was required.

Statistical analyses.

The impact of hospital admission during the 3-month period prior to study entry on the likelihood that patients would be found to be colonized with VRE at the time of study admission was assessed by using the Pearson chi-square statistic.

Analyses were performed for patients who were not colonized with VRE on admission and who achieved optimal or acceptable adherence with the screening protocol. The rate of acquisition of VRE was defined as the number of new cases of colonization with VRE divided by the total number of days of patient exposure until a study end point was reached. Among patients who became colonized with VRE, this duration of exposure was from the time of study admission to the day that the patient was first found to be colonized with VRE. For patients who did not become colonized with VRE, exposure time was calculated from the time of study admission until the date of the final negative rectal swab specimen culture. The potential impacts of various factors including, patient sex, comorbidities, managing unit, and recent antibiotic exposure on the rates of acquisition of VRE were assessed by use of the Cox proportional hazards regression model for univariate and multivariate analyses. Variables with P values of <0.2 in the univariate analysis were included in the multivariate analysis. Standard errors for these incidence rates were calculated by the jackknife method (5) to account for a likely correlation between multiple admissions for the same patient. If no events were observed in a group, the Poisson approximation method was used to calculate standard errors (Statcorp 1999; Stata statistical software, release 6.0; Stata Corporation, College Station, Tex.). A P value of <0.05 was considered statistically significant.

RESULTS

During the 15-month study 8,953 rectal swab specimens were obtained from 3,458 patients during 4,215 admissions. Only 16% of patients had repeat admissions during the study. For most patients (n = 2,103; 50%) the result for only one rectal swab specimen was assessed; but for 1,168 (28%) patients two rectal swab specimens were assessed, for 419 (10%) patients three rectal swab specimens were assessed, and for 525 (12%) patients four or more rectal swab specimens were assessed. This swabbing frequency was expected given the typical lengths of stay observed for these patients (the median length of stay at each study site was 7 days). Adherence to the swabbing protocol was generally good, with 93% of the intervals between the collection of swab specimens being ≤8 days and only 1.4% of the intervals being greater than 16 days. The majority of patients (66%) were admitted directly to a study ward. However, 5% of patients were in a hospital for more than 10 days before they entered a study ward.

No clinical infections with VRE occurred at any study site during the trial period. However, 66 patients were found to be colonized with VRE (Table 2; vanB E. faecium, 71%; vanB E. faecalis, 21%; vanA E. faecium, 6%; vanA E. faecalis, 2%). The vancomycin and teicoplanin susceptibilities of these isolates were typical of those for vanA isolates (MIC, ≥256 μg of vancomycin per ml) or vanB isolates (range, 4 to ≥256 μg of vancomycin per ml). Notably, most strains of VRE were nonclonal on PFGE analysis. A total of 56 of the 66 isolates (85%) from patients found to be colonized with VRE were subjected to PFGE analysis, and 33 different PFGE patterns were identified among the 56 isolates. Data for patients with acceptable adherence are shown in Table 3.

TABLE 2.

New cases of VRE colonization

Hospital and unit	No. (%) of:		No. of rectal swab specimens in which VRE were detected on:
	Patients	Admissions	Initial culture	Subsequent culture
AH
ICU, general	496 (14)	513 (12)	9	5
ICU, road trauma	405 (11)	414 (10)	2a	0
ICU, cardiothoracic	821 (23)	842 (20)	3	7
Renal	130 (4)	178 (4)	0	1
Heart-lung transplant	124 (4)	163 (4)	0	4
ARMC
ICU	403 (12)	423 (10)	3	2
Renal	288 (8)	525 (12)	5	11
Liver transplant	34 (1)	39 (1)	2	2
MMC
ICU	187 (5)	191 (5)	1	0
Renal	321 (9)	495 (12)	2	3
Hematology-oncology	285 (8)	432 (10)	4	0
Total	3,494	4,215	31	35

^aFor both specimens, VRE were detected in swabs recovered after a delay. No road trauma patients for whom a rectal swab specimen was cultured (n = 342) within 2 days of admission was colonized with VRE.

TABLE 3.

Timing of acquisition of VRE among patients with both optimal and acceptable adherence to rectal swab specimen culture protocol

Patient no.a	Detection on study dayb:
	2	4	6	8	10	12	14	16	18	20	22	24	26	28	30	32	34	36	38	40	42	44	46	48	50	52	54	56	58	60	62	64	66	68	70	72	74
1	X	•A
2	X	•B
3	X	•C
4	X		•*
5	X		•*
6	X			•D
7	X	X		•E
8	X			•*
9	X			•F
10	X	X		•E
11	X	X				•G
12		X			X				•E
13		X	X				X	X		X	X				X		X	X		X		X		X	•H
14			X	•E
15				X		•C
16			X			•I
17			X				•J
18		X				X			•K
19			X				X				•L
20					X						X			X				•M
21				X	X	X	X	X	X	X				X		X	X	•N
22	X	X							X	X		X		X		X	X		X		•*
23				X			X				X			X				X			X	X		X	•E
24	X	X		X		X	X		X	X			X		X	X		X		X	X						X	X		X		X		X	X		•O

^aPatients 1 to 13 underwent the optimal rectal swab specimen culture protocol.

^bX, no VRE detected on rectal swab specimen culture; •, VRE detected on rectal swab specimen culture; superscripts A to O, PFGE data, in which different letters represent different strains, based on three band differences; *, not tested. Results for 2-day intervals are given for ease of comprehension.

Among the 3,086 cases (2,612 patients) for which a rectal swab specimen was obtained within 2 days of study admission, 19 were colonized with VRE, suggesting an overall baseline prevalence of 0.6%. The risk of colonization with VRE increased with time spent in a hospital prior to entry into a study unit: the prevalence of colonization with VRE was 0.4% for cases admitted directly to a study unit, 0.8% if cases had been in a hospital 1 to 8 days prior to study entry, and 1.9% if cases had been in a hospital for >8 days prior to entry into a study unit. Among acute trauma patients who were admitted from the community directly to the trauma ICU (n = 342), none were colonized with VRE. In contrast, among the 437 cases that had had a previous admission to a study site during the preceding 3 months, 1.4% (95% confidence interval [CI] for relative risk, 1.1 to 7.3 [P = 0.03]) were colonized with VRE at admission.

Interpretation of the status of acquisition of VRE among patients from whom a baseline rectal swab specimen was not obtained (unacceptable screening) was difficult. The limited analysis that could be undertaken suggested that the rate of colonization with VRE increased with the duration of inpatient stay at the study sites. Among 954 cases initially swabbed 3 to 8 days after study admission, 8 (0.8%) were colonized, whereas 4 of 175 (2.3%) cases that were first swabbed more than 8 days after study admission were colonized.

Acquisition of VRE was observed for 35 patients who had negative initial baseline rectal swab specimen cultures. Of these patients, 24 had undergone acceptable screening (and 13 of these had optimal screening adherence) (Tables 2 and 3). Acceptable screening occurred for 1,992 admissions (and optimal screening occurred for 1,553 admissions), accounting for a total of 16,881 days at risk. Thus, the rate of acquisition of colonization with VRE among the group with acceptable screening was 24 of 16,881, or 1.4 cases per 1,000 days of person-time at risk (95% CI, 1.0 to 2.2). This rate was similar if only the patients who had optimal screening were considered (data not shown).

The timing of acquisition of VRE among patients with acceptable screening is shown in Table 3. The results of univariate and multivariate analyses of patient risk factors for the acquisition of VRE are shown in Table 4. Patients managed by renal units appeared to have a significantly higher risk of acquisition than patients managed by other units. Although univariate analysis suggested that previous admission to a study site within the past 3 months was significantly associated with the acquisition of VRE, this was not confirmed on multivariate analysis. Sex, the presence of liver failure or renal impairment, diabetes, or organ transplantation was not associated with the acquisition of VRE. Antibiotic use was common in our study population, especially the use of glycopeptides and cephalosporins. Among the 1,992 admissions that had acceptable adherence to the rectal swab specimen culture protocol, glycopeptides were used in 614 (31%); broad-spectrum cephalosporins were used in 688 (35%); carbapenems were used in 329 (17%); ticarcillin-clavulanic acid was used in 195 (10%); ampicillin, gentamicin, or ciprofloxacin was used in 503 (26%); and metronidazole was used in 328 (17%). Piperacillin-tazobactam (0.1%), cefepime (0.3%), and cefpirome (0.3%) were rarely used. Of the broad-spectrum cephalosporins, ceftriaxone (48%) was the most commonly used, followed by cefotaxime (35%) and ceftazidime (17%). However, we did not show an association between the use of glycopeptides or broad-spectrum cephalosporins and the new detection of VRE (Table 4). Instead, the strongest association with the development of colonization with VRE was noted with the recent use of ticarcillin-clavulanic acid. Administration of this agent was highly associated with colonization by multivariate analysis of both the acceptable and optimal screening groups (for the acceptable screening group, the hazard ratio [HR] was 3.6 and the 95% CI was 1.13 to 11.6; for the optimal screening group, HR was 6.4 and the 95% CI was 3.7 to 36.6 [P = 0.03 and <0.001, respectively]). By multivariate analysis, recent carbapenem administration was also associated with colonization with VRE among patients in the acceptable screening group (HR, 2.8; 95% CI, 1.0 to 8.0 [P = 0.048]).

TABLE 4.

Risk factors for new colonization with VRE among patients adherent to acceptable culture protocol

Risk factor	No. (%) of		Univariate analysis			Multivariate analysis
	New cases of VRE (n = 24)	Controls (n = 1,968)	HR	95% CI	P value	HR	95% CI	P value
Sex, male	15 (62)	1,279 (65)	0.91	0.40, 2.09	0.84
Presence of:
Diabetes	8 (33)	360 (18)	1.73	0.74, 4	0.20
Immunosuppression	4 (17)	401 (20)	0.59	0.19, 1.85	0.36
Dialysis	11 (46)	327 (17)	1.98	0.82, 4.8	0.13	0.94	0.26, 3.32	0.93
Renal failure	10 (42)	443 (23)	1.46	0.63, 3.37	0.37
Liver failure	3 (12)	137 (7)	2.04	0.66, 6.26	0.21
Transplantation	2 (8)	176 (89)	0.60	0.12, 2.8	0.52
Previous admission within 3 mo	9 (37)	305 (15)	2.45	1.05, 5.73	0.04	2.08	0.84, 5.18	0.11
Managing unit:
Intensive carea	7 (29)	1234 (63)	1
Renal	13 (54)	429 (22)	2.32	0.92, 5.86	0.07	4.62	1.22, 17	0.02b
Transplant (liver, heart/lung)	4 (17)	102 (5)	1.27	0.37, 4.38	0.71	1.48	0.43, 5.12	0.5
Haematology/oncology	0	203 (10)
Antibiotic use
Any antibiotic	19 (79)		1.77	0.65, 4.81	0.26
Glycopeptide(s)	13 (54)		2.08	0.96, 4.55	0.06c	2.08	0.82, 5.26	0.12
Broad-spectrum cephalosporin(s)g	4 (17)		0.50	0.17, 1.53	0.23
Imipenem, meropenem	8 (33)		2.04	0.80, 5.18	0.14d	2.84	1.02, 7.96	0.048d
Ticarcillin-clavulanate	4 (17)		2.33	0.80, 6.84	0.12e	3.64	1.13, 11.64	0.03e
Ampicillin, gentamicin, ciprofloxacinf	6 (25)		0.93	0.35, 2.43	0.88
Metronidazole	4 (17)		1.39	0.47, 4.16	0.55

^aManagement by ICU was used as the comparator; therefore, HR was equal to 1.

^bFor the 13 patients adherent to the optimal culture protocol, HR was 4.26 (95% CI, 1.26 to 14.5 [P = 0.02]) by univariate analysis and HR was 10.9 (95% CI, 1.6 to 74 [P = 0.014]) by multivariate analysis.

^cFor the 13 patients adherent to the optimal culture protocol, HR was 0.96 (95% CI, 0.29 to 3.31 [P = 0.94]) by univariate analysis.

^dFor the 13 patients adherent to the optimal culture protocol, HR was 1.44 (95% CI, 0.31 to 6.63 [P = 0.64]) by univariate analysis; therefore, data for this population were not evaluated by multivariate analysis.

^eFor the 13 patients adherent to the optimal culture protocol, HR was 4.86 (95% CI, 1.65 to 14.32 [P = 0.004]) by univariate analysis and HR was 6.44 (95% CI, 3.63 to 36.6 [P = <0.001]) by multivariate analysis.

^fPatients receiving any of these antibiotics were analyzed together because these agents have only limited activities against anaerobic organisms.

^gCeftriaxone, cefotaxime, and ceftazidime.

A post hoc analysis of patients in the acceptable screening group was undertaken to assess any difference in the impacts of antibiotics with activity against anaerobes (ticarcillin-clavulanic acid, piperacillin-tazobactam, meropenem, imipenem, and metronidazole) compared with the impacts of those with no activities against anaerobes. This demonstrated a significant association between the use of antibiotics with activities against anaerobes and the new detection of VRE (by univariate analysis, HR was 2.21 and the 95% CI was 0.82 to 5.96 [P = 0.12]; by multivariate analysis, HR was 3.34 and the 95% CI was 1.4 to 7.73 [P < 0.01]).

DISCUSSION

This is one of the largest prospective studies to assess colonization with VRE in institutions in which the risk of nosocomial transmission of VRE had been minimized by the strict enforcement of infection control and isolation procedures. We especially aimed to accurately identify risk factors that were associated with the new detection of VRE, particularly the new emergence of VRE or the unmasking of preexisting low-level colonization rather than the nosocomial spread of VRE. The prospective cohort design also allowed us to control for the potential confounding effects of duration of exposure.

Overall, we found that 1.56% of cases (1.91% patients) were colonized with VRE, with an acquisition rate of 1.4 patients per 1,000 patient days in study units. Our findings are consistent with previous Australian data and for a number of reasons suggest that we did in fact largely control for nosocomial transmission. First, the 1.56% overall rate of colonization with VRE is substantially lower than that from a previous report, in which 4.6% of high-risk patients were colonized following an outbreak of VRE in one of the study institutions (17). This rate of colonization is substantially less than that described in many U.S. and European studies (6, 31, 35).

Second, consistent with previous Australian reports and unlike the epidemiology of VRE most commonly reported from the United States and Europe, we found vanB strains of VRE, especially vanB E. faecium strains, to be the most commonly isolated species (1, 3, 17, 25). Most importantly, the vast majority of our isolates appeared to be nonclonal by PFGE and to only infrequently be the clones responsible for most cases of colonization with VRE and disease caused by VRE in Melbourne (G. Hogg, personal communication). This observation is consistent with the view that the strict infection control measures in place at each study site were largely effective in controlling the nosocomial transmission of VRE.

The low overall rate (0.6%) of carriage of VRE at admission, the even lower rate of carriage among the subset of patients new to the study, and the fact that no colonization with VRE was detected among any acute trauma patients directly admitted to the major trauma ICU support the view that colonization with VRE is uncommon in the general Australian community. We have previously suggested that the community rate of carriage of VRE is only 0.2%, unlike in Europe, where community colonization rates of 8 to 23% have been reported (14, 16, 21, 25, 33, 34).

Multivariate analysis of our data suggested that the potential risk factors for the new detection of VRE were being managed by a renal unit and being treated with a broad-spectrum agent, ticarcillin-clavulanic acid or a carbapenem. These data are consistent with those from previous reports (26). The association between the recent administration of ticarcillin-clavulanic acid (and possibly carbapenems) and the new detection of VRE provides important in vivo support for humans of the animal studies conducted by Donskey et al. (10, 11), in which drugs with broad-spectrum activities (notably, ticarcillin-clavulanic acid) appeared to be linked to both the establishment and the persistence of colonization with VRE. Similarly, other studies by those investigators (9) have suggested that antibiotic regimens with activity against anaerobes promote the density of colonization with VRE in the human gastrointestinal tract, although those studies did not specifically assess the role of ticarcillin-clavulanic acid. Importantly, our patient-specific antibiotic data confirm the association found between hospital-wide VRE infection rates and hospital purchasing data for ticarcillin-clavulanic acid (12). A number of studies have noted a high rate of colonization with VRE among renal patients (21, 27).

We did not find an association between the new detection of VRE and recent administration of either broad-spectrum cephalosporins or vancomycin. There may be a number of potential reasons for this observation. First, as was well described in the recent review by Harbath et al. (18), the factors associated with acquisition of VRE are often complex, may be confounded by local variables, and may be different depending on whether the patient acquires VRE by nosocomial transmission or by primary in vivo emergence (e.g., gene transfer to previously susceptible enterococci). Second, our attempt to predict and adjust for the likely impact of renal dysfunction on the potential time during which patients had detectable serum antibiotic concentrations may have influenced our analysis. We believe this to be an important feature of any studies such as ours, since it provides a more accurate assessment of the likely duration of exposure during which the patients' normal floras are exposed to selective antibiotic pressure. However, few other studies assessing risk factors for colonization with VRE or disease caused by VRE have controlled for this variable. Finally, our study may have been too small to detect such an association.

Our findings of largely nonclonal strains of VRE, the generally delayed timing of detection of VRE, and the association with recent antibiotic therapy are consistent with our reported hypothesis regarding the emergence of VRE. We have previously identified vanB genes in a number of naturally occurring fecal anaerobes and have proposed that VRE may emerge under appropriate selective conditions when vanB genes are transferred to susceptible enterococci in the human gut. Subsequently, nosocomial transmission may result in the spread of some clones (29). While the findings of this study are consistent with this hypothesis, formal evidence of gene transfer is still required.

Our study has a number of limitations. First, the epidemiology of VRE in Australia appears to be different from that in Europe and the United States (1, 24). Our findings may not be applicable to other regions. In particular, the fact that colonization with VRE appears to be uncommon in the general Australian community and the fact that VRE are not endemic in most Australian hospitals are notable (25). Furthermore, risk factors associated with colonization with VRE among high-risk patients may be different for patients not managed by such specialized units. Second, the number of patients who were colonized with VRE and who had acceptable adherence to the screening program was relatively limited (n = 24). Thus, only a relatively limited proportion of the total number of patients colonized with VRE identified (n = 66) could be accurately assessed for risk factors associated with colonization with VRE. This finding highlights the difficulty in maintaining adherence to such clinical study protocols in busy acute-care hospitals where high-risk patients are usually found. Adherence difficulties may also have affected our assessment of risk factors for colonization with VRE. For instance, recent administration of carbapenems was significantly associated with colonization with VRE among patients with acceptable screening adherence but was not significantly associated with colonization with VRE among patients who had optimal screening, probably due to the smaller number of cases with newly detected colonization with VRE.

We believe that since PFGE analysis of our VRE strains suggested that most isolates were nonclonal, the strict infection control procedures that were in place at each study site were reasonably effective in preventing the nosocomial transmission of dominant clones. Nevertheless, the fact that some strains had similar PFGE patterns suggests that we cannot exclude some hospital-related transmission of VRE.

Given the recent findings by D'Agata et al. (7) regarding the potential rate of false-negative results for rectal swab specimen cultures for the detection of fecal carriage of VRE, we cannot be absolutely certain that some of our patients who appeared to have become colonized with VRE after study admission may not have actually already been colonized at the time of study entry. Apart from the inherent potential sampling error associated with rectal swab specimen cultures, some variability in culture technique is possible, although, it is hoped, limited. Fecal culture for the detection of VRE would probably have been more accurate, but this would have been impractical within the study sites and the patient population assessed (7).

Harbath et al. (18) have highlighted the large number of variables that may be associated with colonization with VRE and the difficulty in assessing and controlling for all potential confounding factors. Nevertheless, given the size and circumstances of our study, we believe that we have accurately identified risk factors associated with colonization with vanB strains of VRE. Our study suggests that strict infection control and isolation procedures are effective in controlling nosocomial transmission.