Community-associated Methicillin-Resistant Staphylococcus aureus Strains in Pediatric Intensive Care Unit 1

Virulent community-associated methicillin-resistant Staphylococcus-aureus (CA-MRSA) strains have spread rapidly in the United States. To characterize the degree to which CA-MRSA strains are imported into and transmitted in pediatric intensive care units (PICU), we performed a retrospective study of children admitted to The Johns Hopkins Hospital PICU, March 1, 2007–May 31, 2008. We found that 72 (6%) of 1,674 PICU patients were colonized with MRSA. MRSA-colonized patients were more likely to be younger (median age 3 years vs. 5 years; p = 0.02) and African American (p<0.001) and to have been hospitalized within 12 months (p<0.001) than were noncolonized patients. MRSA isolates from 66 (92%) colonized patients were ﬁ n-gerprinted; 40 (61%) were genotypically CA-MRSA strains. CA-MRSA strains were isolated from 50% of patients who became colonized with MRSA and caused the only hospital-acquired MRSA catheter-associated bloodstream infection in the cohort. Epidemic CA-MRSA strains are becoming endemic to PICUs, can be transmitted to hospitalized children, and can cause invasive hospital-acquired infections. Further appraisal of MRSA control is needed.

M ethicillin-resistant Staphylococcus aureus (MRSA) frequently infects children. Traditionally, MRSA infections were confi ned to those who frequented healthcare facilities or had predisposing healthcare-associated risk factors. In the 1990s, reports emerged of MRSA infections in healthy children in the community who had no predisposing risk factors (1). Community-onset MRSA infections were caused by MRSA strains belonging to the genotypes USA300 and USA400 (identifi ed by pulsed-fi eld gel electrophoresis [PFGE]), also referred to as the communityassociated MRSA (CA-MRSA) strains (2,3). These CA-MRSA strains are associated with increased production of toxins and are less resistant to antimicrobial drugs than are traditional hospital-acquired MRSA (HA-MRSA) strains (4,5). Although CA-MRSA strains usually cause mild skin and soft tissue infections, they can also cause severe and fatal disease (6)(7)(8).
As the community prevalence of MRSA has risen (9), more children colonized or infected with MRSA have been admitted to hospitals (10)(11)(12), especially with phenotypic CA-MRSA strains. Notably, CA-MRSA strains can cause outbreaks in hospitals (13) and have become a frequent cause of hospital-onset infections (14,15). Aside from ways to manage outbreaks (16) and a report that clinical cultures underestimate MRSA prevalence (17), little is known about the prevalence of MRSA colonization of hospitalized children. The degree to which CA-MRSA strains are imported into and transmitted in high-risk settings such as pediatric intensive care units (PICUs) has not been determined. Understanding the effects of MRSA in hospitalized children is essential to guide, assess, and plan MRSA prevention and control programs among hospitalized children. Our objectives were to measure the prevalence of MRSA colonization at the time of admission to the PICU and to determine the effects of CA-MRSA strains on MRSA colonization, transmission, and hospital-acquired infections in the PICU.

Setting and Design
The Johns Hopkins Hospital is a 920-bed tertiary care academic medical center with an embedded 175-bed children's hospital. The institution serves Baltimore, Maryland, USA, and the surrounding area. The 26-bed PICU admits ≈1,700 medical and surgical patients each year, including patients needing hematopoietic stem cell transplants and organ transplants, as well as cardiac, orthopedic, and neurosurgical patients. Beginning March 1, 2007, as part of a hospital MRSA prevention and control program, the Department of Hospital Epidemiology and Infection Control initiated screening of patients for MRSA colonization at the time of admission to the PICU and weekly thereafter. Nares swab specimens were obtained by PICU nurses and cultured for MRSA as described later. Newly identifi ed patients or those known to be colonized or infected with MRSA were isolated in cohort groups. Compliance with admission screening was reported back to the unit monthly. Hospital policy required strict hand hygiene, use of standard precautions, and contact isolation for all patients colonized or infected with MRSA.
During March 1, 2007-May 31, 2008, we performed a retrospective cohort study to identify all MRSA-colonized patients in the Johns Hopkins Hospital PICU, including those colonized at the time of admission and those who became colonized while in the PICU. If patients were admitted to the PICU multiple times, only the fi rst admission was included. The institutional review board approved this study and waived informed consent to review retrospective data collected during hospital care.

Defi nitions
MRSA colonization at the time of PICU admission was defi ned as having a nasal surveillance culture obtained at the time of admission that grew MRSA or any clinical culture that grew MRSA within 3 days of PICU admission (18,19). Known MRSA carriers were defi ned as any patients with an institutional history of MRSA colonization or infection before PICU admission. Newly identifi ed MRSA patients (incident cases) had no institutional history of MRSA colonization or infection (either had negative cultures on prior admissions or clinic visits or had not previously been tested). Case-patients with incident MRSA became colonized or infected in the PICU and met the following criteria: 1) had a positive screening or clinical culture obtained >3 days after admission to the PICU (19), 2) had no institutional history of MRSA, and 3) had a previous negative culture from the same site during the current PICU admission. Incidence density was calculated as the number of incident MRSA cases per 1,000 patient-days at risk for MRSA acquisition (i.e., patient days during which patients were not known to be colonized or infected with MRSA). Healthcare-associated MRSA (HA-MRSA) infections met the criteria established by the National Healthcare Safety Network's surveillance defi nition for healthcare-associated infections (20). CA-MRSA strains included those belonging to the PFGE genotypes USA300 and USA400 (3). HA-MRSA strains included those belonging to other PFGE genotypes.

Data Collection and Case Identifi cation
We searched a computerized surveillance support system (Theradoc Inc., Salt Lake City, UT, USA) to identify all patients with microbiology cultures of samples from any body site that grew MRSA from March 1, 2007, through May 31, 2008, and to determine compliance in performing screening cultures at the time of PICU admission. Adminis-trative databases were searched to obtain patient characteristics. Medical records were reviewed to determine whether MRSA infections met the National Healthcare Safety Network's surveillance defi nition for healthcare-associated infections.

Laboratory Methods
Nasal surveillance swabs were plated on BBL CHROMagar MRSA plates (BD Diagnostics, Sparks, MD, USA), a selective and differential medium to detect MRSA. Mauve-colored colonies present after 24 or 48 hours of incubation were confi rmed as S. aureus by Gram stain and slide coagulase testing (21). We performed PFGE on available stored isolates. DNA was extracted and digested by using Sma1 (22,23). We used S. aureus subspecies NCTC 8325 as a control strain, and all USA PFGE strain types were included for comparison (3). The USA type strains were obtained through the Network of Antimicrobial Resistance in Staphylococcus aureus program (supported under National Institute of Allergy and Infectious Diseases/National Institutes of Health contract no. HHSN272200700055C). We performed PFGE on the CHEF-DR III (BioRad Laboratories, Hercules, CA, USA). Gels were stained and scanned by using a molecular analysis fi ngerprinting software (Fingerprinting II Version 3.0; BioRad Laboratories). We considered isolates to be related if their patterns had <3 band differences (3) and to be unrelated if they had >3 band differences.

Statistical Analysis
Data were maintained in Microsoft Access 2003 (Microsoft Corp., Redmond, WA, USA) and analyzed by using Stata version 10.0 (StataCorp., College Station, TX, USA). Means, medians, and interquartile ranges (IQRs) were calculated for select demographic variables. Categorical variables were expressed as numbers and percentages. Comparisons were made by using the Pearson χ 2 or Fisher exact test. For continuous variables, the Student t test or the Wil-coxon rank-sum test was used to compare groups, depending on the distribution of the data. Logistic regression was used to calculate odds ratios (ORs) and 95% confi dence intervals (CIs) and to evaluate the strength of associations. A pairwise correlation coeffi cient was calculated to assess an association between incident cases and monthly admission colonization prevalence. A 2-tailed p value <0.05 was considered signifi cant for all statistical tests.

Results
From March 1, 2007, through May 31, 2008, 1,674 children were admitted to the Johns Hopkins Hospital PICU. The median age was 5 years (IQR 1-12 years), and 55% of patients were male. Only 53 patients (3.2%) had an institutional history of MRSA colonization or infection. Screening cultures were performed on nasal swabs from 1,210 children (72%) obtained at the time of PICU admission. When patients that were screened were compared with those that were not screened, screened patients were more likely to have been hospitalized in the previous 12 months (29% vs. 22%, p<0.01). No other signifi cant differences in demographic or clinical characteristics were found between those patients screened for MRSA colonization and those not screened (Table 1).
At the time of admission to PICU, 72 (6.0%) patients were colonized with MRSA: 68 patients (94%) identifi ed by results of nasal screening cultures and 4 patients whose clinical culture grew MRSA within 3 days of PICU admission. Characteristics of patients colonized with MRSA at the time of admission (group 1, currently colonized) were compared with those of patients not colonized with MRSA and never known to be colonized (group 2, never colonized) and patients not colonized with MRSA at admission but who had an institutional history of prior colonization or infection (group 3, previously colonized) ( Table 2). Compared with never-colonized patients, MRSA-colonized patients tended to be younger (median age 3 years vs. 5 years, Most MRSA-colonized patients were <6 years of age (71%) and African American (54%). Eighteen percent were admitted from an in-patient unit, and in the previous 12 months, 58% had been hospitalized and 43% had been admitted to the PICU.
Patients colonized with CA-MRSA were compared with those colonized with HA-MRSA (Table 3). Patients colonized with CA-MRSA strains were less likely to have been admitted to the PICU within the previous 12 months (OR 0.31; 95% CI 0.11-0.84). No signifi cant differences were found between the 2 groups in other demographic, clinical, or outcome characteristics, including median age, sex, number of newly identifi ed MRSA carriers, length of stay in the hospital before PICU admission, admission from home or patient care unit, admission to a medical or surgical service, PICU length of stay, or hospital length of stay.
To identify patients who became colonized with MRSA while in the PICU, MRSA screening cultures from nares specimens were sent weekly, and clinical cultures from patients in the PICU for >3 days were monitored. During the study period, 8 incident MRSA cases were identifi ed (Table 4), an incidence density of 1.01 cases per 1,000 patient days at risk. No correlation was shown between monthly colonization prevalence at time of admission and incident MRSA cases (p = 0.09). Six (75%) of 8 incident cases were identifi ed by a screening culture. Of these 6 patients, 2 had subsequent clinical cultures grow MRSA. If weekly MRSA screening cultures were not performed, only 4 (50%) of 8 incident MRSA cases would have been identifi ed. Patients with incident MRSA cases were in the PICU for a median of 6 days (range 5-24 days) before acquiring MRSA and had a median age of 6 years (range 1-11 years). Seven (88%) of 8 MRSA isolates were available for PFGE analysis. Four (57%) of 7 isolates were identical to or related to PFGE-type USA300, documenting healthcare-associated transmission of CA-MRSA strains. Six (75%) of 8 patients  with incident cases were admitted to a surgical service. Of the patients who became colonized with MRSA, 3 (38%) acquired a subsequent MRSA infection during their stay in the PICU (1 central-line-associated bloodstream infection, 1 case of ventilator-associated pneumonia, 1 case of ventilator-associated tracheitis). Both respiratory infections were caused by PFGE strain A, a strain that was not associated with colonization of any patients at the time of PICU admission. The only HA-MRSA bloodstream infection was caused by a USA300-related strain.

Discussion
These data describe the prevalence of MRSA colonization in patients admitted to the PICU and suggest that CA-MRSA strains may be becoming endemic in hospitalized children. We found that 6.0% of patients screened at the time of admission to the PICU were colonized with MRSA. Most (60%) MRSA-colonized patients would not have been recognized if admission screening cultures had not been performed. Sixty-one percent of MRSA colonized patients harbored CA-MRSA strains, mostly USA300. Our data show that epidemic CA-MRSA strains are endemic to the PICU. These strains can be transmitted to children in the hospital and can cause invasive hospital-acquired infections, including bacteremia.
Aside from how to manage an MRSA outbreak, little research has attempted to characterize the epidemiology of MRSA colonization and transmission in the PICU. The lack of research in PICU patients is surprising, given the abundance of published articles that have characterized the epidemiology of MRSA in adult ICU patients (24,25). Our fi ndings agree with those of studies of adult patients: screening cultures detect a large reservoir of MRSA-colonized patients, cases that would otherwise go undetected (18,24). Most (75%) patients colonized at the time of PICU admission had no institutional history of MRSA colonization or infection. MRSA-colonized patients serve as a reservoir for contamination of healthcare workers' hands and subsequent MRSA transmission to patients. Active detection and isolation of MRSA carriers can reduce MRSA transmission in hospitals (26,27). Given the risk to patients for MRSA acquisition and subsequent infection (28), many hospitals screen high-risk populations to identify MRSA carriers. In an attempt to curb the spread of MRSA in healthcare facilities, some states have passed legislation mandating MRSA screening.
As the community prevalence of MRSA has risen (9), more children infected with MRSA have been admitted to hospitals (10,11). Our data suggest that these patients represent a small fraction of the patients colonized with CA-MRSA strains who enter the hospital. We found that 61% of children colonized with MRSA at the time of PICU admission harbored a CA-MRSA strain. Findings from a previous cohort showed a lower percentage of colonization with CA-MRSA strains (29). Among patients >13 years of age who were admitted to an urban hospital in Atlanta, 7.3% were colonized with MRSA, and 30% of those were colonized with CA-MRSA strains (29).
Several factors may explain the high prevalence of colonization with CA-MRSA strains in our cohort. First, the community prevalence of MRSA colonization in children is increasing nationwide and in Baltimore (30)(31)(32), largely the result of spread of CA-MRSA strains. Second, children generally have less exposure to healthcare settings where they would be exposed to traditional HA-MRSA strains. Third, children frequently are in settings of close personal contact where opportunities for hygiene may be limited, such as day care, schools, and sports teams, settings postulated as high-risk environments for MRSA transmission.
Notably, children in this study who were colonized with MRSA at the time of admission to PICU were more Emerging Infectious Diseases • www.cdc.gov/eid • Vol. 16  likely to be younger and African American. Our fi nding of younger age is consistent with data from the 2003-2004 Centers for Disease Control and Prevention National Health and Nutrition Examination Survey Nasal Swab survey, which found that children 1-5 years of age had the second highest MRSA colonization rates, behind adults >60 years of age (R. Gorwitz, pers. comm.). Similarly, we found that children with previous, but not current, MRSA colonization tended to be older. Why colonization prevalence is different in various pediatric age groups remains unknown. Previous studies have found racial disparities in rates of invasive MRSA disease (6,33). In a 2007 report by Klevens et al., incidence rates of invasive MRSA disease were more than twice as high for African Americans than for whites, and mortality rates were 80% higher for African Americans (7). The reason for these racial disparities is unknown. MRSA colonization is a known risk factor for invasive MRSA, so higher colonization rates in African Americans may in part explain higher rates of invasive disease. CA-MRSA strains are changing the landscape of MRSA infection prevention and control in hospitals. We found that CA-MRSA strain USA300 was the most commonly acquired MRSA strain identifi ed in the PICU. All patients who acquired MRSA had negative nares swab cultures at the time of PICU admission, and all subsequently exhibited MRSA nasal colonization. Of 8 patients, 3 (38%) had subsequent MRSA infection during their PICU stay. We were unable to monitor these patients after PICU dis-charge and likely have underestimated their long-term risk for subsequent MRSA infection. CA-MRSA strains, with the potential to spread rapidly and cause severe disease, have now been shown to cause hospital-acquired infections and hospital outbreaks (7,(13)(14)(15). Hospital outbreaks confi rm that CA-MRSA strains can be transmitted and acquired in the healthcare setting. The role of CA-MRSA strains in endemic MRSA transmission has not been elucidated. The extent to which endemic CA-MRSA strains will affect the epidemiology of HA-MRSA transmission and infections remains unknown. Our data suggest that this topic requires further study.
As CA-MRSA strains become more prevalent in hospitals, the importance of distinguishing between MRSA strains remains unclear. However, CA-MRSA strains appear to be highly transmissible and may have increased virulence (33)(34)(35). When attempting to distinguish between patients colonized with CA-MRSA strains and those colonized with non-CA-MRSA strains, we found that patients colonized with HA-MRSA strains were 3× more likely to have been admitted to an ICU within the previous 12 months. Other demographic characteristics did not differ between the groups. Overall, our fi ndings agree with those of other studies that have shown that demographic data and risk factors may not reliably distinguish between patients colonized or those infected with various MRSA strains (14,36).
Our study has several limitations. First, only nares cultures were performed to identify asymptomatic MRSA  The second limitation is that admission screening cultures were instituted in March 2007, and compliance with screening was only 72%. Our PICU did not have admission order sets, and in the absence of a physician's written order, cultures were not always performed. Compliance improved over time with initiation of a patient order entry system, a visual reminder to perform screening cultures on the front of the patient's chart, and frequent compliance auditing by the nurse manager. However, given the similarities between screened and unscreened patients, we expect that our measured prevalence was representative of the entire cohort. Third, our PICU has low MRSA incidence rates and may have a low prevalence of MRSA colonization at the time of admission compared with other PICUs. These conditions may limit how our fi ndings can be generalized to other institutions.
Overall, we found that epidemic CA-MRSA strains are likely endemic to PICUs. These virulent and transmissible strains are entering the PICU through infected or colonized patients, they are being transmitted to children, and they are responsible for hospital-onset MRSA infections. CA-MR-SA strains often colonize children without healthcare-associated risk factors. Traditional infection-control strategies, in which risk factors are used to target high-risk patients for screening and intervention, may prove insuffi cient for MRSA. Future studies to determine optimal approaches to control MRSA transmission in hospitalized children are needed. As CA-MRSA strains enter the hospital environment, the increased frequency of methicillin resistance and the coexistence of multiple strain types may lead to the selection of novel MRSA strains with enhanced capacity for transmission and infection. These conditions would be especially concerning with regard to children, for whom a more restricted antimicrobial drug arsenal is available. Sound epidemiologic investigation and feasible interventions are needed to control MRSA and protect hospitalized children.