Introduction

Methicillin-resistant Staphylococcus aureus (MRSA) was first reported in England in 1961, and the current mortality rate associated with serious MRSA infection is estimated to be 20–25% [1]. According to a report by the National Nosocomial Infections Surveillance (NNIS) System in 2004, MRSA infection in intensive care units (ICUs) increased from 11% in 1998 to 60% in 2003 [2]. From July 2004 through December 2005, the Active Bacterial Core surveillance (ABCs) system and the Emerging Infections Program (EIP) of the US Centers for Disease Control and Prevention (CDC) conducted surveillance for invasive MRSA infections. Most MRSA infections were healthcare-associated, but 13.7% were community-associated infections [3]. MRSA is an important pathogen not only for hospital infections, but also for community-associated infections. Furthermore, in the past, MRSA infections have been limited to patients with risk factors such as dialysis, diabetes, or usage of a ventilator or invasive medical devices. But recently, community-acquired MRSA infections have been reported in the absence of identified risk factors [4].

According to one study in Korea [5] which evaluated the nationwide nosocomial infection rate in 2004 and antimicrobial resistance in ICUs, S. aureus was the most frequently identified microorganism (23.2%). MRSA was also the most common pathogenic cause of bacteremia, pneumonia, and surgical site infection, with frequencies of 16.5, 90, and 21%, respectively. The study found that 93% of S. aureus isolates were resistant to methicillin, up from 79% in 1996. This rate was also higher than the 60% MRSA infection rate in US ICUs reported by the NNIS System from 2004. These findings show that MRSA infection has become a serious public health issue in Korea.

In a study conducted by Haley et al. [6], the prevention of nosocomial infections resulted in cost savings of about 95%. Moreover, another study [7] showed that the attributable costs spent by hospitals to prevent MRSA infections were $35,367, affirming the hypothesis that MRSA infection control is important not only for patient health but also economically.

The development of antimicrobial resistance in S. aureus and the prognosis of infected patients has been debated in the study of MRSA infection. Specifically, its effects on the length of hospitalization and hospital cost are controversial.

Careful adjustment for confounding factors is important when comparing outcomes between patients with MRSA and methicillin-susceptible S. aureus (MSSA) bacteremia [8]. The propensity score is a powerful device for constructing matched pairs or matched sets that balance numerous observed covariates, and is sufficient to remove bias regardless of the sample size [9]. In the present study, a propensity score matching method was used to evaluate the impact of methicillin resistance on mortality, length of hospitalization, and hospital costs in patients with S. aureus bacteremia.

Methods

Study design and data collection

This study was conducted at Kyung Hee University Medical Center, a 1,200-bed, tertiary care hospital in Korea. All patients with blood cultures positive for S. aureus were identified from a retrospective review of medical records from between 1st January 2003, and 31st December 2008.

Two situations of S. aureus bacteremia were considered to be clinically significant. One circumstance was two or more positive blood cultures for S. aureus within 24 h. The second was if the patient had clinical signs and symptoms of infection but no concurrent infection with another organism that would explain the displayed signs and symptoms. If a patient had any recurrent episodes of S. aureus bacteremia during the study period, only the first episode was considered. Also, if the antibiotic susceptibility changed in recurrent episodes, only the first susceptibility was included.

Patients with healthcare-acquired S. aureus infection diagnosed  >48 h after hospital admission or those who had resided in a long-term care facility in the 12 months preceding the culture date were included in this study. Those patients with a community-acquired infection that had been diagnosed within 48 h of hospital admission were also included. The source of bacteremia was classified according to clinician assessment as primary or secondary at the time of S. aureus bacteremia based on the evaluation of medical records [10]. A bacteremia was considered to be central venous catheter-associated if the same organisms were observed in at least one percutaneous blood culture, as well as in a culture of the catheter tip or from two blood samples drawn from a catheter hub and a peripheral vein, with no other apparent infection source [11]. Empirical therapy was defined as the first antibiotic regimen following the onset of bacteremia. Concordant treatment was defined as treatment with an antimicrobial agent that was susceptible according to culture results. The route of administration and the timing relative to the first positive blood culture were also considered.

Patient demographics (age and sex), comorbidities (underlying malignancy, diabetes, chronic lung disease, liver disease, or renal disease), previous antibiotic histories, ICU admissions, and risk factors were investigated through their medical records. Susceptibility testing was performed using the MicroScan method (MicroScan Walk-Away, Dade Behring, West Sacramento, CA, USA) and the results were interpreted according to the National Committee for Clinical Laboratory Standards (NCCLS) guidelines [12].

The severity of underlying medical conditions was classified using McCabe and Jackson scores that categorize conditions as fatal, ultimately fatal, or non-fatal [13].

Outcome assessments

The hospital days from the time of admission to the occurrence of S. aureus bacteremia were measured for each patient. The days after the appearance of bacteremia to discharge were also compared between the two groups. Data on the total costs of hospitalization were collected from the central financial service at Kyung Hee University Medical Center and were adjusted to the 2008 US dollar. Hospital costs were categorized as the cost of hospital stay, laboratory tests, care, and treatments. The cost of hospital stay included the costs derived from administration, clerical support, housekeeping, and medical records. The costs associated with care included the cost of physician care, nursing care, and consultations. Treatment costs were based on the total drug costs, cost of materials such as catheters and implanted devices, and costs of procedures, including operations, dialysis, respiratory care, and rehabilitation.

According to the report of the Korean Nosocomial Infections Surveillance System (KONIS), MRSA constituted 70% of isolated pathogens in the healthcare-associated infection (HCAI) study in Korea [14]. Based on these data, we did not isolate the patients with MRSA infection but performed only standard precautions and surveillance methods. Consequently, the costs did not include MRSA quarantine costs.

Death was considered to be related to bacteremia if the patient had positive S. aureus blood cultures and persistent clinical signs and symptoms of sepsis [15]. Vital status 30 days after the onset of bacteremia was also assessed.

Statistical analysis

The results were analyzed using SAS version 9.1 (SAS Institute, Cary, NC, USA) and SPSS for Windows version 17.0 (SPSS, Chicago, IL, USA). The t-test was used to compare the MRSA and MSSA groups. Qualitative variables were compared using Fisher’s exact test and Pearson’s chi-square test. We matched propensity scores to evaluate the health outcomes of MRSA infection compared with those of MSSA infection. To model the probability of MRSA infection, logistic regression was used. Age, sex, admission route, ICU admission, epidemiologic classification (whether the infection was community-associated or healthcare-associated), clinical condition (primary or secondary bacteremia), use of devices (central venous catheter, urinary catheter, or ventilator), and the number of underlying diseases or conditions such as diabetes mellitus (DM), chronic obstructive pulmonary disease (COPD), liver cirrhosis (LC), solid tumor, renal disease with dialysis, or being immunocompromised were considered as independent variables for the propensity scores. Patients who were infected with MRSA were matched to patients in the MSSA group by the greedy matching algorithm. We used a 5–1 computerized greedy matching algorithm, which was developed for the SAS package for this study [16]. In this greedy matching algorithm, the MRSA cases were matched to MSSA controls with regard to five digits of the propensity score, and those that did not match were then attempted to be matched relative to four digits, and so on, down to one digit. If no single-digit match was found for an MRSA case, it was excluded from the analysis. After propensity score matching, McNemar’s test and paired t-tests were used to compare both groups. Also, we compared the appropriateness of therapy between the MRSA and MSSA groups after propensity score matching.

Results

A total of 167,488 patients were admitted to the hospital during the study period and 266 patients with S. aureus bacteremia were analyzed for this study. Among them, 145 patients had MRSA bacteremia (54.3%) and 121 patients had MSSA bacteremia (45.7%). The differential characteristics of patients with MRSA bacteremia and those with MSSA bacteremia are described in Table 1. Patients with MRSA bacteremia were older than those with MSSA bacteremia (mean age, 62.1 vs. 54.8 years; P = 0.03), but the sex distribution for both groups was similar (P = 0.67). More patients with MRSA bacteremia had underlying comorbidities such as COPD (P = 0.01), while more patients with MSSA bacteremia had underlying comorbidities such as LC (P = 0.01) or had undergone immunosuppressive therapy (P = 0.002). Furthermore, patients who were admitted to the ICU were more likely to be having MRSA infection (71.2 vs. 28.8%; P = 0.001). Among patients with community-associated bacteremia, MSSA bacteremia was more prevalent than MRSA bacteremia (78.4 vs. 21.6%; P < 0.001). The use of invasive medical devices such as a central line, Foley catheter, or ventilator was associated with the occurrence of MRSA bacteremia. Of the patients with MRSA bacteremia, 28% compared with 12% of patients with MSSA bacteremia had used a ventilator (P < 0.001), 63% of patients with MRSA bacteremia compared with 46% of patients with MSSA bacteremia had a central line (P < 0.001), and 54% of patients with MRSA bacteremia compared with 19% of patients with MSSA bacteremia had received a urinary catheter (P < 0.001) (data was not shown).

Table 1 Baseline characteristics of patients with Staphylococcus aureus bacteremia, stratified by methicillin susceptibility
Full size table

The case fatality rates in the unmatched total populations of MRSA and MSSA were 21.4 and 28.9%, respectively. This difference was not statistically significant (P = 0.16).

Propensity score matching

In this study, a total of 106 patients, or 53 pairs, were matched on the basis of the propensity score to adjust for potential differences between MRSA and MSSA bacteremia. The result of the Hosmer–Lemeshow test was 0.122, and the C-value was 0.865. Paired t-tests were used to compare characteristics between the matched pairs. The baseline characteristics of both groups were similar after propensity matching (Table 2). Patients with MRSA bacteremia had a median of 0.8 comorbidities and patients with MSSA bacteremia had a median of 0.8 comorbidities upon admission (P = 0.89).

Table 2 Propensity score matching analysis of the impact of methicillin-resistant Staphylococcus aureus on the length of hospital stay and hospital cost
Full size table

Patients with MRSA bacteremia were more likely than those with MSSA bacteremia to have stayed longer in the hospital before the development of bacteremia (mean 25.0 vs. 6.1 days; P < 0.01). But after developing bacteremia, the differences in hospital stay lengths were not statistically significant (mean 35.0 vs. 28.7 days; P = 0.33). The average total length of hospital stay remained significantly longer for patients with MRSA bacteremia as compared to MSSA bacteremia (mean 59.9 vs. 34.8 days; P = 0.01).

After adjustment for age, sex, epidemiologic classification (whether the infection was community-associated or healthcare-associated), clinical condition (primary or secondary bacteremia), use of devices (central venous catheter, urinary catheter, or ventilator), and the number of underlying diseases according to propensity score matching, the hospital costs for the two groups of patients with S. aureus bacteremia were compared. Costs for patients with MRSA bacteremia were higher than those of patients with MSSA bacteremia, but this difference was not statistically significant ($9,369.6 vs. $8,355.8; P = 0.62).

McNemar’s test was used to test the comparison of mortality outcomes between the two groups (Table 3). A total of 35 pairs had a concordant outcome (31 pairs survived and 4 died) and 18 pairs had a discordant outcome. Therefore, methicillin resistance was not considered to be an independent predictor of mortality in patients with S. aureus bacteremia (P = 0.48). Typical empiric therapy in our hospital for patients with MSSA infection was a beta-lactam antibiotic, such as cefazolin or nafcillin, and glycopeptide antibiotics like vancomycin and teicoplanin were used for the patients with MRSA infection. Receipt of concordant empirical therapy was slightly lower in patients with MRSA infection than in those patients with MSSA infection, but the difference was not statistically significant (41 pairs vs. 48 pairs; P = 0.19).

Table 3 Comparison of the outcome for patients with bacteremia caused by methicillin-resistant Staphylococcus aureus or methicillin-susceptible Staphylococcus aureus by propensity score matching analysis
Full size table

Discussion

We have presented evidence that patients with MRSA bacteremia have significantly longer hospitalization stays but are no more likely to pay higher hospital costs or to die due to the infection than are patients with MSSA bacteremia. To the best of our knowledge, this is the first study in Korea using propensity score matching to compare hospital costs between cases of MRSA bacteremia and MSSA bacteremia.

According to the study conducted by Cosgrove et al. [17], MRSA bacteremia was a significant predictor of increased length of hospitalization and the length of hospital stay attributable to MRSA infection was 2.2 days. Other studies that evaluated outcomes related to MRSA bacteremia indicated that MRSA bacteremia was associated with a longer length of hospital stay after the onset of bacteremia [18, 19]. In the current study, MRSA bacteremia affected the total length of hospital stay (P = 0.01) but, interestingly, the difference in the length of stay after bacteremia was not significant in patients with MRSA bacteremia when compared to patients with MSSA bacteremia. This result might arise because of differences among national medical systems. In the United States, each patient is classified into a Diagnosis Related Group (DRG) on the basis of clinical information and hospitals are paid a pre-determined rate for each Medicare admission, which is called the Medicare Prospective Payment System. DRG-based prospective payment systems result in the transition to more outpatient services from inpatient services [20]. But in Korea, where this investigation was conducted, patients pay providers for services under a Fee-For-Service (FFS) reimbursement system that may, in part, be the cause of the lack of difference between the lengths of stays after bacteremia between the MRSA and MSSA groups.

Many previous studies investigating the effects of MRSA infection compared with MSSA infection have shown conflicting results [21–23]. In a previous analysis [24], MRSA infection was the most significant predictor of delayed treatment and delayed treatment was found to be an independent predictor of infection-related mortality. However, in another study, the difference in MRSA bacteremia-related mortality between appropriate and inappropriate empirical treatment was not significant [25]. This disagreement may be the result of methodological differences between studies, small sample sizes, confounding factors, and a broad range of underlying comorbid conditions, etc. Therefore, we used propensity score matching to control for confounding factors as much as possible. After adjustment for the effects of other variables, the mortality rate was not significantly higher in patients with MRSA infection (P = 0.48). These findings agree with those of previous studies [26, 27], which found that mortality due to S. aureus was not more frequent in patients with MRSA bacteremia.

Variables relating to a patient’s prior clinical condition could influence the outcome of bacteremia. In this study, we closely evaluated risk factors for infection with MRSA. Furthermore, the use of propensity score matching allowed us to define the impacts of methicillin resistance on mortality, length of hospitalization, and hospital costs. When we matched the MRSA and MSSA groups by propensity score, methicillin resistance was not a significant predictor of mortality. Our data suggest that the association of underlying disease and host variables could result in an overestimation of MRSA virulence. We believe that strict control of confounding variables, such as the severity of underlying disease, is important because these variables could distort comparisons of outcomes between MRSA and MSSA infections.

In a meta-analysis [17], methicillin resistance was associated with increases in hospital costs of more than $7,000 over costs of patients with MSSA bacteremia. At the national level, annual admission cases for the treatment of patients hospitalized with MRSA was 120,000 and the annual cost for MRSA treatment in the United States was $15 billion [28]. In the current study, Korean patients with infections due to MRSA incurred hospital costs of an average of $9,369.60 per person, but there was no significant difference between this figure and that of MSSA patients (P = 0.62).

A limitation of our study is that only measured covariates were controlled and there may be additional confounding factors not accounted for. This problem is always a limitation in studies without randomized controlled methods, including our study. Furthermore, our results reflect cost data for the hospital where the MRSA infections were identified and the costs may be different in other Medicare and insurance systems.

In summary, this study indicated that MRSA bacteremia was not associated with higher risks of mortality or a substantial increase in hospital costs, but that it was associated with an increased length of hospital stay compared to MSSA bacteremia. These findings remained significant even after adjustment for comorbidities. Because the prevalence of MRSA infection has increased worldwide [29], the information provided by this study may help clinicians and policymakers to derive methods to control the impacts of MRSA infection on medical personnel, health policymakers, and patients.