Keimyung University Dongsan Medical Center is a 931-bed tertiary care hospital in Daegu, Korea. The NSICU involved in the outbreak is a 20-bed unit with two private rooms, and is an integral part of a single ward consisting of 61 beds devoted to neurosurgical patients. Our Infection Control Unit with a full-time infection control nurse was established in June 2000 but no Infectious Disease service was established until March 2002. This study was conducted as part of a quality improvement project in the NSICU in 2003.

To determine the trends of isolation of S. marcescens over time, we used stored computerized microbiology data dating from July 1997. Microbiologic data were analyzed by culture specimen, hospital ward, and clinical department.

We examined patient care areas and other areas of the NSICU and reviewed infection control policies and hand antisepsis practices in May 2002. Additionally, we surveyed the infection control practices of health care workers, especially in terms of the manipulation of indwelling urinary catheters.

A case-control study was conducted in May 2002 to identify the risk factors for the acquisition of S. marcescens in urine. The source population for this case-control study was defined as all patients who had been admitted to the NSICU for ≥ 72 hr from January through March 2002. A case patient was defined as any patient in the NSICU with at least one positive culture for S. marcescens in urine. Data were collected until the date of isolation of S. marcescens in urine. For each case, two controls were selected from patients without isolation of S. marcescens in urine as close as possible in time to the case. Collection of data ended after a maximum of 7 days in the NSICU, which was the median length of ICU stay before the isolation of S. marcescens in the case patients.

Environmental cultures were obtained from potential environmental sources, such as, sinks, counter tops, bed rails, medication carts, urinary catheters and urine bags, unused urinals, and disinfectants in February 2003. Cultures were not obtained from healthcare workers. Prospective surveillance cultures were obtained from urine for all patients with indwelling urinary catheters in the NSICU. Urine samples were taken on the day of and 3 days after admission to the NSICU, and thereafter weekly. Surveillance cultures were continued weekly in the NS ward as long as urinary catheters were kept in place after the patients were transferred to the NS ward. Routine surveillance cultures were conducted from February through December 2003, but not thereafter due to limited resources. Additional surveillance cultures to measure point prevalences were done during April 2005. Cultivation and identification of the organism were performed according to standard procedures. Pulsed-field gel electrophoresis (PFGE) was performed for 34 S. marcescens strains isolated from April 2002 to November 2003 as described elsewhere (23) and their fragmentation patterns were interpreted using previously suggested criteria (24).

Univariate analysis was performed using the χ² test or the two-tailed Fisher's exact test (for small expected values) for categorical variables, and using the Student's t-test for continuous variables. Multivariate analysis was performed to demonstrate associations between individual variables and organism acquisition. All variables that were statistically significant by univariate analysis were analyzed using a logistic regression model. Two-tailed p values of <0.05 were considered statistically significant. All statistical analyses were performed using SPSS software (SPSS 10.0 for Windows).

From July 1997 through December 2001, a total of 1,231 S. marcescens duplicate isolates among 525 hospitalized patients were recovered from a variety of clinical specimens. The organisms were most commonly isolated from urine specimens (71.2%) and from the NS (82.0%). Of 389 patients with S. marcescens in urine, 340 (87.4%) were from the NS. For most of these NS patients, acquisition of S. marcescens was identified while they were in the NSICU (81.8%). It was evident that the outbreak was associated with urine, particularly in the NSICU.

The first isolate of S. marcescens from urine in the NSICU was identified in a 51-yr old male patient transferred from other hospital for surgical management of cerebral hemorrhage in July 1999. At the time of admission to the NSICU, indwelling urinary catheter had already been placed at the previous hospital and he revealed no signs of infection. On hospital day 8, the organism was cultured from catheterized urine, and found to be resistant to multiple antibiotics but susceptible to imipenem. Thereafter S. marcescens was increasingly identified in urine of patients in the NSICU and then in the NS ward (Fig. 1). Almost all organisms showed the similar antibiotic susceptibility pattern, although different sets of antibiotics were used for antimicrobial susceptibility testing during the study period. We believe that the above transferred patient was the index case of this outbreak.

The demographic and clinical characteristics of 35 cases and 70 control patients are shown in Table 1. Subarachnoid hemorrhage was the most common diagnosis and no statistical difference was found in the distribution of neurosurgical diseases between cases and controls. Urine sampling for tests and invasive procedures were more frequently performed in case patients. Furthermore case patients received antibiotics for longer periods for surgical prophylaxis or empiric treatment of fever of unknown etiology, and this was especially true for third-generation cephalosporins, such as cefoperazone. Mean length of NSICU stay before the isolation of S. marcescens in urine was 6.9 days (median, 7; range, 3-11 days). All isolates were resistant to multiple antibiotics (including cefoxitin, cefotaxime, cefepime, gentamicin, and amikacin) and only susceptible to imipenem. Only one isolate was susceptible to ciprofloxacin. Multivariate analysis revealed that the use of third-generation cephalosporins, urine sampling for tests, and central venous catheterization were independent risk factors for the acquisition of S. marcescens (Table 2).

Written infection control standards were in place at the hospital, but these were not strictly followed by healthcare workers in the NSICU. Breaks in infection control practices were identified during the manipulation of indwelling urinary catheters. Backflow of urine through a urinary catheter was often found at the time of position changes and during patient movements. Connection tubes were disconnected for urine sampling for urinalysis, culture, or for the measurement of specific gravity. The irrigation of urinary catheters was performed often but not routinely. A few urinals were shared to drain urine collected in urine bags and they were not sufficiently disinfected between usages according to the current infection control standards (25). Furthermore, hand antisepsis between the nursing care of individual patients was inadequate. However, inappropriate practices were not documented at the time of catheter insertion.

In a patient colonized with S. marcescens in urine, environmental cultures isolated S. marcescens, Staphlyococcus epidermidis, and Acinetobacter baumannii from the surface of a urine bag. Results of surveillance cultures are depicted with those of clinical specimens in Fig. 1. Last point prevalence surveillance cultures in April 2005 did not grow S. marcescens from any patient. PFGE of 34 selected strains isolated between April 2002 and November 2003 identified two distinct clones (Fig. 2). One predominant clone distributed throughout the period above, was comprised of two closely related subtypes, and was resistant to all antibiotics tested, except imipenem. The other clone showed different antimicrobial susceptibility patterns (Table 3). This clone had been isolated since October 2003.

After the procedural review and survey, adherence to hand antisepsis was encouraged and infection control measures, especially on the manipulation of indwelling urinary catheters (26), were intensified through in-service educational meetings for nursing staffs in the NSICU and NS ward in January 2003. Individual urinals were not recommended for economic reasons but it was recommended that drainage valves not be touched when draining urine. A pair of new plastic, not sterile, gloves were used each time urine was drained from a urine bag. Disconnection of connection tubes was not allowed thereafter. Urine sampling was performed only by aspiration with a sterile needle and syringe after clamping the distal end of the catheter and disinfection, because no sampling port was present in available catheters. In addition, it was reinforced against irrigation through indwelling urinary catheters. Cohort care was employed for patients acquiring S. marcescens in urine. For residents and attending medical staffs, a series of lectures was given on the state of the S. marcescens outbreak in the NSICU in May 2003. The outbreak lasted despite the reinforcement of infection control measures, but eventually no patients were found to have newly acquired the organism in the NSICU from December 2004.