To find patients with HA-MRSA infections, we identified all cultures obtained >72 hours after hospitalization that grew MRSA, from January 1, 1999, through December 31, 2004, at Harbor-UCLA Medical Center, a tertiary-care, urban, county hospital in Los Angeles County. At this hospital, surveillance cultures for MRSA colonization are not routinely performed; therefore, cultures positive for MRSA are likely to reflect infection rather than colonization. For a given patient, we examined only data from the first positive culture and excluded patients who had positive cultures both ≥72 hours and <72 hours after admission. If a patient had been hospitalized more than once during the study period, only data from the first hospitalization were retained. A standardized instrument was used to abstract data from the medical record of each patient. Information obtained included demographics, admission date and time, hospital location, antimicrobial drug susceptibility of the MRSA isolate, and time, date, and source of the MRSA culture.

We obtained only MRSA blood isolates for molecular typing because the clinical microbiology laboratory discards all other types of isolates after identification is complete. In vitro susceptibilities were reported as minimal inhibitory concentrations and performed with the VITEK system (bioMérieux, Durham, NC, USA), according to the protocols of the Clinical and Laboratory Standards Institute (CLSI). The investigation protocol was reviewed and approved by the Institutional Review Board of Harbor-UCLA Medical Center.

Molecular typing was performed at the University of Chicago by investigators who were blinded to the clinical details and antibiograms of the isolates.

PCR was performed to detect mecA by using the primer pair mecAF/mecAR (22). SCCmec elements were distinguished by the molecular architecture of the ccr and mecA complexes (21,23,24). PCR typing of SCCmec types I–IV was performed under the conditions previously described (24,25). SCCmec type II (ccrAB complex type 2 and mec complex class A), SCCmec type III (ccrAB complex type 3 and mec complex class A), and SCCmec type IV (ccrAB complex type 2 and mec complex class B) were assigned according to published criteria (25). PCR primers used to detect mecI (primers mI3/mI4), the mecR1 membrane spanning region (MS) (primers mcR3/mcR4), and the mecR1 penicillin-binding region (PB) (primers mcR1/mcR5) were originally reported by Suzuki et al. (26). Screening for ccrAB complex types 1, 2, and 3 (ccrAB 1, 2 and 3) was accomplished with a multiplex PCR assay that uses a mixture of 4 primers designed by Ito et al., consisting of a common forward primer (β2) and reverse primers, α2, α3, and α4 specific for ccrAB complexes 1, 2, and 3. Thermocycler conditions used have been described (27). The presence of the ccrAB gene complex allotype 4 (ccr complex 4) was assessed in a separate reaction that used the primer pair ccrA4F and ccrB4R (27). Screening for the ccrC gene (ccr complex 5) was performed with a forward primer (γF) in combination with the reverse primer γR described by Ito et al. (28). Prototype strains used for SCCmec typing were NCTC10442 (SCCmec I), N315 (SCCmec II), 85/2082 (SCCmec III), MW2 (SCCmec IV), and WIS (SCCmec V). The control strain used for detection of ccrAB4 was S. epidermidis strain ATCC 12228, which contains ccrAB4 in the non–mec-containing SCCcomposite island (24).

MLST was performed by PCR amplification and sequencing of 7 housekeeping genes by using the primer pairs designed by Enright et al (29). Denville Taq-Pro Complete (Denville Scientific, Metuchen, NJ, USA) or the Taq DNA Polymerase (Promega, Madison, WI, USA) was used for the PCR reactions. PCR products were evaluated on an agarose gel and purified by using Millipore 96-well Montage (Billerica, MA, USA) plates according to manufacturer’s instructions. The purified templates were sequenced at the University of Chicago Core Sequencing Facility and evaluated with the use of Vector NTI software (Invitrogen, Carlsbad, CA, USA). Each sequence was submitted to the MLST database website (www.mlst.net) for assignment of the allelic profile and sequence type (ST).

Isolates were screened for the lukF-PV and lukS-PV genes encoding the components of the PVL toxin by PCR amplification of a 433-bp product that includes a portion of both the lukS-PV and lukF-PV ORFs by using the primer pair luk-PV-1/ luk-PV-2 (final concentration 0.2 μM) designed by Lina et al. (30). The thermocycler conditions have been described (27).

A standardized definition of CA-MRSA infection was created by the Centers for Disease Control and Prevention (CDC) Active Bacterial Core Surveillance sites (31). Using this definition, we defined HA-MRSA infections as those MRSA infections that did not meet the definition of CA-MRSA infections. Specifically, we defined an MRSA isolate as HA associated if the original entry criteria of hospitalization for >72 hours before culture acquisition was met and if in the year before the present hospitalization, the patient had had any 1 of the following: hospitalization, surgery, residency in a long-term care facility, and hemodialysis or peritoneal dialysis, or at the present admission had indwelling percutaneous devices or catheters. A CA infection was defined as a culture-confirmed MRSA infection without any of the above criteria. However, if the patient did not meet any of the above criteria, had an infection at the time of admission, and the culture of the infection on admission was taken ≥72 hours after admission, then the infection was considered CA. An example of this situation would be a deep tissue infection microbiologically diagnosed from a surgical biopsy specimen 4 days after the patient’s admission.

To validate our definition of HA-associated infection, we reviewed 105 (30%) randomly selected charts of the patients with MRSA infections identified ≥72 hours after hospitalization. The purpose of this validation was to confirm that these cultures did not reflect CA infections that were diagnosed late (>72 hours) in the hospital course. Of note, in the CDC definition, an infection is considered HA if it occurs >48 hours after admission. Yet, we chose >72 hours as a cut-off to more conservatively capture HA infections, i.e., to minimize the miscategorization of CA infections as HA infections.

We then defined MRSA strains as having the SCCmec type IV phenotype if the isolates were resistant to oxacillin and susceptible to gentamicin, clindamycin, and trimethoprim-sulfamethoxazole. All other isolates were considered to be phenotypically non–SCCmec type IV.

Characteristics were compared between patients infected with the non–SCCmec type IV phenotype isolates and those infected with SCCmec type IV phenotype isolates by using a χ² or t test, as appropriate. No adjustments were made for multiple comparisons. Temporal trends in the proportion of the SCCmec type IV phenotype were compared with the Cochran-Armitage test of trends. A multivariate analysis that predicted phenotypically SCCmec type IV isolates was performed by using an unconditional logistic regression model and a backward model selection method. A p value of <0.05 was defined as statistically significant. Data analysis was done with SAS software version 8.2 (SAS Institute Inc., Cary, NC, USA).

We identified 352 patients who had HA-MRSA cultures; 229 (65%) were men, and the median age was 50 years (mean 49.5 years). In the subset of medical records reviewed for validation of HA or CA status, none of the patients’ infections (0/105) fit our definition of a CA infection. The SCCmec type IV phenotype was identified in 128 (36%) of these 352 patients. Compared with the non–SCCmec type IV phenotype, patients with the SCCmec type IV phenotype were younger (median age 48 vs. 54 years, p = 0.02) and had the defining culture taken earlier in the hospitalization (median 8 vs. 15 days, p = 0.01). Finding an isolate with the SCCmec type IV phenotype was more likely if the culture source was from a wound, blood, or source other than sputum (odds ratio [OR] 2.9, 95% confidence interval [CI] 1.7–5.0, p<0.0001; OR 2.6, 95% CI 1.2–5.7, p = 0.02; and OR 1.2, 95% CI 0.6–2.3, p = 0.69) (Table 1).

Of the 352 cultures, 35 were recovered from blood and were potentially available for genetic analysis. We were able to subculture 24 of the blood isolates. We could not perform SCCmec typing on 1 of the 24 growing isolates. The 23 remaining isolates were representative of each year of the 6-year period except 1999, when no isolates could be recovered.

Twelve isolates carried the SCCmec type IV element, and 9 also carried the pvl genes (Table 2). Eleven isolates carried the SCCmec type II element; none carried pvl. The clinical definition of the SCCmec IV phenotype was fulfilled by 11 (92%) of the 12 isolates that carried the SCCmec IV element. The exception was an isolate that contained SCCmec IV that was resistant to gentamicin, clindamycin, and trimethoprim-sulfamethoxozole. The definition of the non-SCCmec IV phenotype was fulfilled by 10 (91%) of 11 isolates carrying the SCCmec II element. Phenotypic case definition of SCCmec type was highly correlated with the genotype confirmation of the SCCmec type phenotype (p<0.0001 by Fisher exact test).

The proportion of MRSA isolates with the SCCmec type IV phenotype increased from 17% in 1999 to 56% in 2003 (p<0.0001, Figure). The proportion of isolates that were of the SCCmec type IV phenotype in 2004 (52%) was little changed from 2003 (Figure). In the multivariate model, independent predictors for having an SCCmec type IV phenotype isolate were wound source of culture (referent group was sputum source, OR 2.6, 95% CI 1.5–4.6, p = 0.001), culture obtained in less time after admission, (OR 0.88 per week, 95% CI 0.8–0.98, p = 0.02), and year of culture acquisition (p<0.0001) (Table 1).