JSTOR

Setting and Patient Population

All patients in the Veterans Affairs Maryland Health Care System (VAMHCS) who had S. aureus infections complicated by bacteremia during the period October 1, 1995, through December 31, 2003, were identified from microbiology records. Infection control nurses abstracted information on demographic characteristics, medical history, treatment, laboratory data, and outcomes from the comprehensive electronic medical record. We included veterans with a healthcare encounter at the VAMHCS who had at least 1 blood culture positive for S. aureus, which represented their first S. aureus infection. We excluded polymicrobial infections in which 2 or more organisms were isolated.

Outcome

The primary outcome of interest was death within 90 days after the date of the first blood culture result positive for S. aureus, regardless of whether the patient was an inpatient or outpatient. If the patient died within 90 days after the infection was identified, we assessed whether the S. aureus infection caused the death (ie, whether there was evidence of ongoing infection at the time of death) or contributed to the death (ie, whether the patient died of an underlying condition that was exacerbated by the infection). All other deaths were classified as not related to infection due to S. aureus.

Patient Characteristics

The Charlson comorbidity index was retrospectively calculated for each patient from recorded information so that it represented the patient’s condition at the time of admission^8,9 on the basis of comorbidities listed in the admission history and physical examination reports. The Charlson comorbidity index is a method for scoring comorbid conditions that was developed to predict mortality within 1 year in hospitalized patients.

Infection Characteristics

An episode of S. aureus infection complicated by bacteremia was considered to be hospital acquired if the first blood culture samples positive for S. aureus were obtained 48 hours or more after admission, if the infection was vascular catheter associated, or if the patient was transferred from another facility and the first blood culture samples positive for S. aureus were obtained within 48 hours after admission to the VAMHCS. All other cases of S. aureus infection were considered community acquired. Sepsis at the time of bacteremia was defined according to the American College of Chest Physicians and Society of Critical Care Medicine criteria.¹⁰ To identify the cause of the bacteremia, the source was classified as primary if there was no secondary source. Primary bacteremia cases were further classified as central line–associated if a central line was in place in the 48 hours before the first blood culture samples positive for S. aureus were obtained. A secondary source of infection (eg, pneumonia, urinary tract infection, and/or soft tissue infection) was defined using Centers for Disease Control and Prevention (CDC) criteria for secondary nosocomial infections.¹¹ For example, patients with pneumonia had to meet the CDC definition of pneumonia and have S. aureus isolated from a sputum culture.¹¹ The duration of bacteremia was calculated from the date of the first blood culture result positive for S. aureus to the date of the last culture result positive for S. aureus during antibiotic treatment for infection. Patients who met the Duke criteria for definite or possible endocarditis were presumed to have endocarditis.¹²

Treatment Characteristics

The number of days to effective antibiotic therapy was calculated from the date of the first blood culture positive for S. aureus to the date of the first receipt of an effective antibiotic treatment. We also recorded whether the central line was removed from patients with central line–associated primary bloodstream infection and whether other implanted hardware was removed from patients in whom the implanted hardware was the secondary source of infection (eg, an infection associated with a total knee prosthesis). In cases of infection associated with either a central vascular catheter or hardware, if the catheter or hardware was not removed, then the focus of infection was considered not to have been removed.

Statistical Analysis

Categorical variables were analyzed using the χ² test and the Fisher exact test, and continuous variables were analyzed using the Student t test for normally distributed continuous variables. Continuous variables that were nonnormally distributed were categorized. Effect modification was defined by a Breslow‐Day test statistic for homogeneity with statistical significance set at a P value of less than .01.¹³ Confounding was defined as a change in the relative risk (RR) of 10% or more relative to the unadjusted RR.

We used multivariable Cox proportional hazard regression to evaluate the association between MRSA infection and time to S. aureus infection–related death, while adjusting for identified effect‐modifying and confounding variables. Independent variables were evaluated for adherence to the proportional hazards assumption and examined for effect modification by including appropriate interaction terms.¹⁴ All tests were 2‐sided and P values less than .05 were defined as statistically significant. Statistical analyses were performed using SAS statistical software, version 8.1 for Windows (SAS Institute).

General Data

Of the 484 patients with S. aureus infection complicated by bacteremia, we excluded 46 patients who had polymicrobial infections. Thus, 438 patients remained in the study population. The study patients were predominantly male (99%) and African American (58%), and had a mean age of 63 years. Forty‐four percent of the study patients had MRSA infection. A total of 250 patients died in the 90‐day period after their infection was identified; however, only 114 (26%) died of causes attributable to S. aureus infection.

Methicillin Resistance

Table 1 compares patients with MRSA infection and patients with MSSA infection. Compared with patients with MSSA infection, patients with MRSA infection were significantly older, more likely to be white, and had a higher risk of death within a year as judged by a higher Charlson comorbidity index. Patients with MRSA infection were more likely to have severe sepsis, a hospital‐acquired infection, secondary bacteremia, pneumonia, and a longer interval before receipt of effective antibiotic treatment compared with patients with MSSA infections. Patients with MRSA infection were also less likely to have endocarditis than patients with MSSA infection.

Table 1.  Comparison of Patients With Methicillin‐Resistant and Methicillin‐Susceptible Staphylococcus aureus Infections Complicated by Bacteremia From 1995 to 2003 at the Veterans Affairs Maryland Health Care System

Open New Window

S. aureus–Infection Related Mortality

Table 2 compares patients who did and patients who did not die of their S. aureus infections within 90 days after the infection was identified. Patients who died of their S. aureus infection were more likely to have severe sepsis, had a higher Charlson comorbidity index, had more days of bacteremia, and were more likely to have pneumonia compared with those who survived. They were also older and more likely to be white. Overall, no difference was found in the time to receipt of effective therapy between those who died and those who survived (0.98 vs 0.99 days; , t test). Among patients with hospital‐acquired infection, the time to receipt of effective therapy was significantly shorter among those who died, compared with those who survived. In comparison, among patients with community‐acquired infection, the time to receipt of effective therapy was longer among patients who died than among those who survived, although the difference was not statistically significant ( ). Overall, patients with MRSA infection had a 1.7 times greater risk of dying, compared with patients with MSSA infection (RR, 1.7 [95% confidence interval (CI), 1.3‐2.4]; ).

Table 2.  Comparison of Patients Who Died From Staphylococcus aureus Infection Complicated by Bacteremia From 1995 to 2003 at the Veterans Affairs Maryland Health Care System and Patients Who Survived

Open New Window

Identification of Effect Modification and Confounding Variables

Pneumonia was the only variable that modified the effect of MRSA infection on mortality ( ; Breslow‐Day test). Among patients without pneumonia, the risk of mortality for patients with MRSA infection was 2.4 times greater than the risk of mortality for patients with MSSA infections (RR, 2.4 [95% CI, 1.6‐3.8]; ). Among patients with pneumonia, the risk of mortality was not significantly different between patients with MRSA infection and patients with MSSA infection (RR, 0.7 [95% CI, 0.4‐1.3]; ). We found that age was a potential confounder of the association between MRSA infection and 90‐day S. aureus infection–related mortality (data not shown).

Survival and Cox Regression Analysis

Table 3 gives the sequential Cox proportional hazard models that assess the association between MRSA infection and infection‐related mortality among patients as we added variables to the model. After adjusting for patient characteristics, such as age, that were independently associated with mortality, in patients without pneumonia the hazard ratio for dying from MRSA infection remained significantly greater than that for death from MSSA infection. We did not include the Charlson comorbidity index, type of infection (hospital acquired vs community acquired), or source of infection (primary vs secondary infection) in the final model, because their presence (individually or as a group) did not alter the hazard ratio between MRSA infection and S. aureus infection‐related death (data not shown).

Table 3.  Cox Proportional Hazard Models of Risk Factors for Staphylococcus aureus Infection–Related Mortality in Patients With S. aureus Infections Complicated by Bacteremia

Open New Window