We sought to determine the ability of surveillance cultures and isolation of vancomycin-resistant Enterococcus (VRE)–colonized patients to control nosocomial VRE infection and colonization during a 5-year period (November 1994 through October 1999). During this period, VRE colonization was limited to 0.82% of admissions. The incidence of VRE infection was 0.12 cases per 1000 patient-days (attack rate, 0.07%). Colonized patients were first identified by surveillance (95%) or routine clinical cultures (5%); 14% of colonized patients had a positive clinical culture a median of 15 days after a positive surveillance culture. Ten percent of colonized patients were identified by surveillance at the time of transfer from another health care facility. Identification of these colonized patients was associated with reduction from a peak incidence rate of 2.07% to a rate of 1.25% and stabilization at this lower level. The use of surveillance cultures to identify and isolate patients with asymptomatic colonization can provide sustained control of the spread of VRE within a health care facility.
Since the first clinical isolate of vancomycin-resistant Enterococcus (VRE) was described in 1988 , the prevalence of VRE has increased rapidly. Recent data from the Centers for Disease Control and Prevention (CDC; Atlanta, GA) National Nosocomial Infections Surveillance (NNIS) system indicate that, in 2000, 20%–30% of nosocomial enterococcal infections were resistant to vancomycin . The increasing rate of vancomycin resistance among enterococci is significant for a number of reasons. Infections due to vancomycin-resistant strains of Enterococcus have been associated with higher treatment costs, prolonged morbidity, and greater mortality rates [3–10]. Transfer of vancomycin resistance from enterococci to more-virulent pathogens, such as Staphylococcus aureus, has been demonstrated in experimental conditions , and 2 clinical isolates of vancomycin-resistant S. aureus containing the vanA vancomycin resistance gene from enterococci were recovered from patients in Michigan and Pennsylvania during 2002 [12, 13].
In 1995, the Hospital Infection Control Practices Advisory Committee published recommendations for preventing the spread of vancomycin resistance . These recommendations include prudent use of vancomycin, education of hospital staff about the epidemiology of VRE, use of cohorting and contact isolation precautions for VRE-colonized and -infected patients, use of dedicated equipment (e.g., stethoscopes and blood pressure cuffs) for VRE-colonized and -infected patients, and appropriate hand hygiene practices. These recommendations also state that “intensified fecal screening for VRE might facilitate earlier identification of colonized patients, leading to more efficient containment of the microorganism” [14, p. 7].
The general experience has been failure to control VRE in the health care setting, even though many studies have reported that it can be controlled using active surveillance cultures for identification and contact isolation of colonized patients [4, 15–26]. This report describes the experience from 5 years of control of VRE with this approach after containment of a large hospital outbreak.
Setting. The University of Virginia Hospital is a 600-bed tertiary care hospital that provides all major medical services, including level 1 trauma care, burn care, oncology services, and transplantation of hematopoietic stem cells and solid organs. A program of active surveillance for and isolation of VRE-colonized patients was instituted at this hospital in 1994 in response to an 8-ward outbreak of VRE infection that was associated with an initial point prevalence of 30% on these 8 wards, including an intensive care unit with a 100% prevalence of colonization. The outbreak was terminated by implementation of a program of active surveillance cultures and contact precautions for all colonized patients (figure 1) . Those results, in addition to PFGE data that demonstrated clonality, suggested that transmission from patient to patient was responsible for the outbreak of infection.
VRE surveillance and control program. The program of active surveillance cultures for gastrointestinal colonization with VRE that was introduced in 1994 used perirectal cultures to identify high-risk patients who were asymptomatically colonized with VRE. Perirectal swabs were obtained weekly from high-risk patients, including those located in intensive care units, those receiving any antibiotic, and/or those who had been hospitalized for >3 weeks without meeting other criteria for surveillance cultures. In addition, the clinical microbiology laboratory notified the hospital epidemiology program of all patients for whom VRE was isolated from routine clinical specimens.
A third component of the VRE surveillance program was instituted in 1998 after the prevalence of VRE colonization among hospitalized patients was found to have increased from 0.37% in 1995 to 2.07% in early 1998 (figure 2), suggesting that there was an unidentified reservoir of VRE. Beginning in May 1998, the hospital's computer system was programmed to identify patients being admitted in transfer from other health care facilities and to print a reminder on the appropriate nursing unit to obtain surveillance cultures for VRE and methicillin-resistant S. aureus.
Contact precautions, as recommended by the CDC , were instituted for all patients from whom VRE was isolated on culture. Patients were not isolated until culture results were found to be positive, except for the roommates of patients who were found to be colonized with VRE. These patients were presumptively placed under contact precautions until the results of 2 serial cultures of perirectal samples, which were obtained ⩾1 week apart, were found to be negative.
Microbiological techniques. Perirectal swabs were obtained as part of the VRE surveillance program and were processed at the epidemiology laboratory using methods that were described elsewhere . Specimens were obtained using dry, cotton-tipped swabs and were directly plated onto selective culture media without the use of broth-enrichment techniques. Clinical specimens were processed in the clinical microbiology laboratory using standard microbiological techniques.
Nosocomial infection surveillance and data management. Hospitalwide surveillance for nosocomial infection was performed by infection control practitioners, as described elsewhere , using current CDC definitions [30, 31]. Data regarding nosocomial infections were maintained in the database provided and maintained by the NNIS. Data related to VRE-colonized and -infected patients, including date of VRE culture, type of specimen from which VRE was isolated (i.e., from a perirectal swab specimen obtained at admission or during weekly surveillance or from a clinical specimen), and the type and name of the transferring facility, when appropriate, were entered into an Access database (Microsoft). General hospital demographic data, such as the monthly number of hospital admissions and days of patient care, were obtained from hospital computing services.
Statistical methods. To determine the effect that the admission culture protocol had on the monthly incidence of VRE colonization, Poisson regression was performed. This analysis included data from January 1997, when the VRE colonization rate began to increase, through September 1999. The number of new colonizations in a given month was the dependent variable. Pre- or postintervention status was the primary independent variable. The number of new colonizations during the preceding month was included in the model as a second independent variable. The number of new colonizations in the preceding month served as an index of colonization pressure during the current month. The logarithm of the number of hospital admissions during the month was defined as the offset (i.e., a regression variable with a constant coefficient of 1 for each observation) in the regression. Probability values for each variable were obtained after adjustment for the other independent variable.
From November 1994 through October 1999, 1050 cases of VRE colonization were detected among 128,266 admissions, representing 0.82% of all admissions. From January 1997 through October 1999, for which more-specific data were available, 768 new cases of VRE colonization were detected among 69,672 admissions (1.1% of admissions). Of these, 730 (95.1%) were identified by surveillance methods, and 38 (4.9%) were identified by clinical microbiology specimens. Ten percent of all patients found to be colonized with VRE each month were identified by the admission culture component of the surveillance program. This included patients from 4 (15%) of 26 referring nursing homes and 26 (27%) of 97 referring acute care hospitals. Of the 730 patients initially identified by surveillance methods, 109 (14.9%) had VRE isolated from a subsequent clinical laboratory specimen. The median time between isolation of VRE from a surveillance culture and isolation from a culture of a subsequent clinical specimen was 15 days (range, 0–488 days). Overall, 621 (81%; 95% CI, 78%–84%) of 768 VRE-colonized individuals would have gone completely undetected and unisolated without the use of active surveillance cultures.
During the 5-year observation period, 90 nosocomial VRE infections were identified in 83 patients during 743,956 days of patient care, resulting in an incidence of VRE infection of 0.12 cases per 1000 patient care–days. The identified infections included 60 urinary tract infections, 11 primary bloodstream infections, 10 surgical site infections, 3 skin or soft-tissue infections, 3 gastrointestinal tract infections, 2 reported cases of pneumonia, and 1 otolaryngological infection. Ninety VRE infections occurred among 1050 VRE-colonized patients, for an attack rate of 8.6%; 83 (7.9%) of 1050 colonized patients had ⩾1 VRE infection. Among all hospital admissions, the attack rate of VRE infection was 0.07 cases per 100 patients. These infections occurred in a variety of patient care areas, including: the medical intensive care unit (27%), surgical wards (26%), surgical intensive care units (21%), and general medical wards (14%). Among the infections that occurred in the surgical intensive care units, ∼50% involved solid-organ transplant recipients. The initial outbreak had prominently involved the transplant service. Among the 83 patients who developed nosocomial VRE infections, 69 (83%) had been found to be colonized with VRE by a surveillance culture before the onset of infection. Among those patients, the median time between detection of colonization and the onset of infection was 12 days (range, 0–169 days). Of the remaining 14 patients with infection, 13 were not known to be colonized until VRE was isolated from the site of infection. The remaining patient was known to be colonized because VRE was isolated from a clinical laboratory specimen collected 4 days before the onset of infection.
Thirty-three months of consecutive data were available for the regression analysis. Sixteen of the months were before initiation of the control intervention in May 1998. Thirty-two months of data were analyzed because the first month of data was used to define the number of previous new colonizations before the second month. In the model, the previous number of new monthly colonizations was highly significant as a covariate (P < .0001). The intervention (i.e., the admission culture program) was also a significant independent factor (P = .0321), indicating that this program was associated with reduction in the previously increasing monthly incidence of VRE colonization (i.e., from 2.07% to 1.25%) and stabilization at this lower level.
It is often said that nosocomial VRE infections cannot be controlled. The many examples of control have been discounted because of their relative brevity and the small number of VRE-infected or -colonized patients involved. The control of VRE for 5 years at a major tertiary care medical center documented in this report suggests otherwise, showing that not only can it be controlled initially, but that this control can be sustained over long periods by preventing transmission. In temporal association with the introduction of the admission culture component of the VRE-control program that led to earlier identification and isolation of VRE-colonized patients transferred from other health care facilities, the rate of new VRE colonization among hospitalized patients, which had been increasing linearly from 0.4% in 1995 to 2.07% in early 1998, stabilized at ∼1.25% of all hospital admissions (figure 2). Rates of colonization of <1% were documented during 3 of the last 4 months of the evaluated period. Regression analysis demonstrated that this intervention was associated with a significant reduction in the number of new VRE colonizations occurring each month. Two potential confounding factors were identified. The first factor was the introduction of an antibiotic use program in May 1998; however, there was no appreciable change in antibiotic use during the first 6 months of that program . Automatic stop orders, which discontinued nonindicated antimicrobial therapy 24 h after notice was given to the treating clinicians, were introduced during the seventh month of the program. After implementation of this component of the program, a relative reduction in overall antibiotic use of 10% (in defined daily doses per patient care–day) was seen during the subsequent 4 months. Second, alcohol hand rinse dispensers were placed throughout the facility in September 1998. This could have resulted in improved health care worker hand hygiene practices with a subsequent reduction in patient-to-patient spread of VRE, but observations by unobtrusive observers of nearly 1000 hand washing opportunities on 23 hospital wards from December 1998 through February 1999 revealed an average hand hygiene rate of <40% (data not shown). This would suggest that careful attention to hand hygiene after contact with unisolated patients was not responsible for reducing transmission of VRE within the hospital. Nor can this success be attributed to a lower-than-average severity of illness among hospitalized patients. In fact, this hospital had the second highest overall case-mix index among all 80 acute care hospitals in the state of Virginia, and it had the highest overall case-mix index among hospitals providing obstetrical services during 1999, the final year of the study period (Virginia Health Information [Richmond, VA], unpublished data).
The control of VRE infection and colonization in this hospital supports the many earlier reports demonstrating control of the spread of endemic and/or epidemic VRE within health care facilities using measures recommended by the CDC [4, 15–26]. In the present study, the period prevalence of VRE colonization among hospitalized patients was kept to 0.82% of hospital admissions for >5 years, and an antibiotic control program was not in place for 3 of these years. This suggested that patient-to-patient spread, rather than de novo mutation to vancomycin resistance in the setting of unregulated antibiotic administration, was the major epidemiological factor associated with VRE in this hospital. The significant association between the number of new cases of VRE colonization in one month and the number of new cases of VRE colonization in the preceding month further suggested that “colonization pressure” resulted in VRE being spread from patient to patient. These results accord with the results of other analyses that suggest that de novo development of vancomycin resistance occurs very rarely among enterococci and that virtually all patients infected or colonized with VRE thus acquired it because of transmission [33, 34].
This study identified a number of factors that may be important for understanding the epidemiology of VRE in the acute care hospital and for developing effective control measures. First, if one looks at the population of patients determined by surveillance culture to be colonized with VRE, 86% of these individuals would have gone undetected if clinical specimens alone had been used for identification. The size of this unidentified reservoir is often not considered, despite the fact that several other investigators have reported similar results [35–38]. A previous study found that, during the routine examination of patients colonized with VRE, 67% of examiners' gowns, gloves, and/or stethoscopes became contaminated with the organism . In that study, contamination rates were similar for examinations of clinically infected and asymptomatically colonized patients. The results of that study were supported by those of 2 other studies [40, 41]. Moreover, 4 of 5 studies reporting epidemiologic data have found lower rates of VRE colonization or infection when gowns and gloves were used, compared with when only gloves were used [18, 42–44]. In 1 of these 4 studies , it was concluded that antibiotic restriction was primarily responsible for the control of an outbreak of VRE infection; however, multiple interventions had been introduced simultaneously, which made it difficult to make firm conclusions related to the exact contribution of any single intervention. Thus, failure to identify the population of asymptomatically colonized individuals and to implement contact precautions may be an important cause of the persistence of endemic and epidemic VRE in health care institutions that do not use active surveillance cultures. The large reservoir of VRE-colonized individuals escaping identification in the absence of active surveillance cultures may explain the lack of success reported with attempts to control VRE infection and colonization that have not applied active surveillance methods among the entire high-risk population [36, 42–47].
Among the small minority (19%) of colonized patients in the current evaluation who had VRE isolated from a routine laboratory specimen, VRE colonization was first detected by routine clinical cultures in only one-quarter of them. For the remainder, a median delay of 15 days between a positive surveillance culture result and isolation of VRE from a subsequent clinical culture means that many preventable opportunities for spread would have occurred among these patients if surveillance cultures had not been performed. The 5% of colonized patients initially identified by culture of clinical specimens did not have subsequent surveillance cultures performed because of their prior positive clinical culture result. It is likely, however, that these patients would have been detected at the time of weekly surveillance cultures during the ensuing week.
Another important finding was that 10% of the patients found to be colonized with VRE were identified at the time of transfer to the study facility from another health care facility. This finding was consistent with earlier studies that have reported interfacility patient transfer as a risk factor for gastrointestinal colonization with VRE [48–50]. In the present study, the identification of a substantial number of patients who were colonized with VRE at the time of transfer to the study facility and the subsequent decrease in rates of VRE colonization within the study facility after the introduction of a program of active identification of these patients suggest that this patient group represented an important and previously unrecognized reservoir of VRE within the hospital. Testing these high-risk patients for VRE colonization at the time of admission led to earlier identification and isolation of colonized patients. This resulted in fewer patients being directly and indirectly exposed to these colonized individuals than if detection had not occurred until the transferred patients met criteria for routine surveillance culture. Had all members of this high-risk group of patients been presumptively isolated until screening cultures were found to be negative, spread of VRE within the study facility might have been even more effectively controlled.
The constant introduction into and subsequent spread within a health care facility of different strains of VRE via patients transferred from other facilities, rather than de novo mutation to vancomycin resistance among enterococci within individual patients, may be one explanation for the extensive genetic diversity that has been described among strains of VRE within single facilities [45, 51–55]. Other identified explanations for this phenomenon include transfer of genetic material conferring antibiotic resistance from one bacterium to another .
A potential limitation of this study was the use of directly plated swabs to identify VRE-colonized patients, because this method has been shown to be relatively insensitive for the detection of VRE colonization, compared with quantitative stool cultures . Quantitative stool cultures or broth enrichment of perirectal swabs might have resulted in identification of additional asymptomatically colonized patients and an even greater reduction in the transmission of VRE, but the prevalence of infection did decrease significantly with implementation of the surveillance cultures  and continued for years to be significantly lower than would have been expected without implementation of such effective preventive measures . The VRE infection rate during the current study in the presence of an active surveillance program was 0.7 cases per 1000 hospital discharges. It is also likely that the VRE infection rate in this hospital would have been even lower had surrounding health care facilities been using the same method for preventing the spread of VRE. Tornieporth et al.  reported a VRE infection rate of 4.2 cases per 1000 hospital discharges in 1992 at a New York City tertiary care facility that was not performing surveillance cultures. Although caution must be used in making direct comparisons between these 2 studies, the relative risk of VRE infection was 6-fold higher in the absence of an active surveillance program (95% CI, 2.7–13.3; P < 10-5). The findings of this evaluation support those of Bonten et al. , which suggested that maintaining a low prevalence of VRE colonization should result in a lower incidence of VRE infection.
In conclusion, the primary reservoir for the spread of VRE within a health care facility consists of asymptomatically colonized patients. Implementation of the CDC's recommendations for control of vancomycin resistance in health care facilities can effectively control the spread of VRE within a hospital even after endemicity has been established. These interventions were demonstrated to be successful even in facilities where VRE had become endemic at a high prevalence with polyclonality [59–61]. The recent identifications of clinical isolates of S. aureus that demonstrate vancomycin resistance due to the acquisition of an enterococcal vancomycin resistance gene [12, 13] highlight the importance of controlling the spread of VRE within the health care system.