This retrospective cohort study was conducted at two hospitals in the University of Pennsylvania Health System (UPHS) in Philadelphia: the Hospital of the University of Pennsylvania (HUP), a 725-bed academic tertiary care medical center, and Penn Presbyterian Medical Center (PPMC), a 344-bed urban community hospital. The study was approved by the institutional review board of the University of Pennsylvania.

All hospitalized patients with an episode of MRSA bacteremia occurring between 1 December 2007 and 31 May 2009 were identified through the HUP Clinical Microbiology Laboratory, which processes all specimens obtained from patients at HUP and PPMC. All of these patients were subsequently included in the study. For patients with multiple blood cultures positive for MRSA, only the first culture was included for analysis.

Identification and susceptibility testing of S. aureus was performed and interpreted according to Clinical and Laboratory Standards Institute (CLSI) guidelines. Standard susceptibility testing was performed using Vitek2. SCCmec typing was performed as previously described.¹⁶ The vancomycin minimum inhibitory concentration (MIC) of all isolates was determined by the Etest using Mueller-Hinton agar (BBL, BD Diagnostic Systems, Franklin Lakes, NJ, USA)¹⁷ with reduced vancomycin susceptibility defined as an Etest vancomycin MIC >1.0 μg/ml.^{18, 19} Vancomycin hetero-resistance was screened for using the macro-Etest method using Etest GRD vancomycin/teicoplanin strips with brain heart infusion agar¹⁷ and by growth on vancomycin-containing brain heart infusion agar;²⁰ a positive result for either screening test was confirmed by population analysis using a Spiral Plater (Advanced Instruments, Norwood, MA, USA) and inoculation onto vancomycin brain heart infusion agar (BBL).²¹ Detection of the genes encoding Panton-Valentine leukocidin (PVL) was performed using real-time polymerase chain reaction using previously described methods.²² Isolates were evaluated for accessory gene regulator (agr) dysfunction via delta-hemolysin production using a beta-hemolysin disk.²³

Data were abstracted from the Pennsylvania Integrated Clinical and Administrative Research Database (PICARD),^{24, 25} which includes demographic, laboratory, pharmacy, and billing information. The following data were collected for all patients: baseline demographics, origin at the time of hospital admission (i.e., physician referral, transfer from another facility, or admission through the Emergency Department), previous admission to UPHS in the 30 days prior to the culture date, hospital location at the time of infection (i.e., intensive care unit [ICU] or medical floor), nosocomial infection (date of the culture ≥48 hours after admission), white blood cell count (WBC) on the culture date, and All Patient Refined-Diagnosis Related Group (APRDRG) Risk of Mortality and Severity of Illness scores.²⁶ Infections were classified as healthcare-associated if the date of the first positive blood culture was ≥48 hours from the date of admission, or if the patient had been previously hospitalized at HUP or PPMC in the 30 days prior to the culture date or was admitted as a transfer from another institution. Otherwise, infections were classified as community-onset. Data on the following conditions was collected in relation to the positive blood culture date: diabetes mellitus, HIV infection, malignancy, renal insufficiency (creatinine ≥2.0 mg/dL or the requirement of dialysis), solid organ or hematopoietic stem cell transplantation, neutropenia (absolute neutrophil count <500/mm³), and receipt of an immunosuppressive agent (e.g., corticosteroids, tacrolimus) in the prior 30 days. The Charlson comorbidity index was calculated for each subject.²⁷ Chart review was performed to collect data on the presence of complicated infection (i.e., endocarditis, osteomyelitis, septic arthritis, epidural and/or spinal abscess) and the presence of intravascular devices (i.e., intravascular catheter, pacemaker or defibrillator, arteriovenous fistula or graft) prior to the episode of bacteremia.

Information on all antimicrobial therapy administered during the same hospitalization was collected, including the time of receipt in relation to the culture date. Antibiotics were considered to be appropriate in relation to treatment of the episode of MRSA bacteremia if they were determined to be active in vitro against the isolate via standard susceptibility testing. The primary outcome was crude in-hospital 30-day mortality, defined as death in the hospital from any cause occurring in the 30 days after the date of the first positive blood culture.

Continuous variables were compared using the Student’s t-test or Wilcoxon rank-sum test and categorical variables were compared using the χ² or Fisher’s exact test. Bivariable analyses were performed to determine the association between SCCmec type and 30-day in-hospital mortality, with the primary exposure of interest being SCCmec type II. Stratified analyses were conducted to elucidate where confounding and interaction were likely to exist in multivariable analyses, using the Mantel-Haenszel test for summary statistics.²⁸ In particular, location in the ICU, hospital of admission, and healthcare-associated infection were designated a priori as potential effect modifiers of interest. Effect modification was assumed to be present when the test for heterogeneity between the odds ratios (ORs) for different strata yielded a P value <0.05. The Mantel-Haenszel test for summary statistics²⁸ was used to evaluate the effects of each variable of interest as a possible confounder. Adjusted ORs with 95% confidence intervals (CI) were calculated using multiple logistic regression for the outcome of 30-day in-hospital mortality. A stepwise selection procedure was used for all multivariable analyses, with variables with P values <0.20 on bivariable analyses or noted to be confounders on stratified analyses considered as candidate variables and maintained in the final model if their inclusion resulted in a ≥15% change in the effect measure for the primary association of interest or were statistically significant on likelihood ratio testing.²⁹

For all calculations, a 2-tailed P value <0.05 was considered to be significant. All statistical calculations were performed using commercially available software (STATA version 11.0; StataCorp LP, College Station, Texas, USA).

A total of 184 unique patients with MRSA bacteremia were identified during the study period. The distribution of SCCmec type among isolates was as follows: 109 (59.2%) with SCCmec II and 75 (40.8%) with SCCmec IV. Baseline clinical and demographic characteristics of patients with MRSA bacteremia are shown in Table 1. Patients with bacteremia due to MRSA harboring SCCmec II compared to SCCmec IV were older (mean age 62.8 and 55.1 years, respectively; P=0.002), had a higher APRDRG Risk of Mortality score at the time of MRSA isolation (mean 2.6 and 1.8, respectively; P=0.004), and had a significantly longer length of stay prior to the culture date (11.1 and 2.49 mean days, respectively, P=0.02).

Isolates with SCCmec II versus SCCmec IV were more likely to be characterized by reduced vancomycin susceptibility (51.4% and 22.7%, respectively; P<0.001)³⁰ and less likely to be PVL positive (0.9% and 68.0%, respectively, P<0.001). There were no significant differences in the proportion of isolates with SCCmec II and SCCmec IV that were characterized by vancomycin hetero-resistance (6.4% and 2.7%, respectively; P=0.36) or agr dysfunction (16.5% and 10.7%, respectively; P=0.29).

In regard to antimicrobial susceptibility rates, MRSA isolates with SCCmec IV compared to those with SCCmec II demonstrated significantly higher rates of susceptibility to clindamycin (86.7% and 4.6%, respectively; P<0.001) and levofloxacin (41.3% and 0.92%, respectively; P<0.001). However, there were no significant differences in susceptibility rates for isolates with SCCmec IV versus SCCmec II for tetracycline (94.7% and 98.2%, respectively; P=0.23) and trimethoprim-sulfamethoxazole (94.7% and 99.1%, respectively; P=0.16).

A total of 34 patients died while hospitalized for MRSA bacteremia, resulting in a crude mortality rate of 18.5%. The mortality rate was 23.9% and 10.7% for patients with MRSA isolates characterized by SCCmec II and SCCmec IV, respectively (P=0.03). Results of bivariable analyses of risk factors associated with in-hospital mortality are given in Table 2. The majority of patients received appropriate antibiotics, specifically 99.3% of patients who survived and 96.9% of patients who died during hospitalization, with vancomycin the most commonly administered initial antibiotic (86.7% and 91.2%, respectively; P=0.58).

On multivariable analyses of risk factors for 30-day in-hospital mortality (Table 3), there was no significant effect modification by location in the ICU (P=0.26), hospital of admission (P=0.21), or healthcare-associated infection (P=0.16). The unadjusted OR between SCCmec II and mortality was 2.62 (95% CI 1.06-7.12, P=0.03). On multivariable analyses, independent risk factors for in-hospital mortality included APRDRG Risk of Mortality score (OR 5.33, 95% CI 2.28-12.4, P<0.001), malignancy (OR 3.25, 95% CI 1.17-9.02, P=0.02), and WBC count on the culture date (OR 1.09, 95% CI 1.03-1.15, P=0.002). After controlling for confounders, SCCmec II was not significantly associated with greater 30-day in-hospital mortality (OR 1.85, 95% CI 0.69-4.92, P=0.22).