Impact of Methicillin Resistance on the Outcome of Patients With Bacteremia Caused by Staphylococcus aureus FREE

METHICILLIN-resistant Staphylococcus aureus (MRSA) has become a worldwide problem, adding to the overall burden of nosocomial infections.^1- 2 Bloodstream infections (BSIs) due to MRSA may account for up to 50% of all staphylococcal BSI.^3- 6 Factors predisposing to MRSA BSI include previous antibiotic treatment, prolonged hospital stay, intravascular catheters, severe underlying conditions, and MRSA nasal carriage.^6- 11

Despite ongoing controversy about the relative virulence of MRSA in vitro and in animal models,^12- 13 most studies have concluded that infections due to MRSA are probably similar in virulence, as measured by the duration of fever, infectious complications, length of hospital stay, and mortality, as infections caused by methicillin-sensitive S aureus (MSSA).^14- 17 Methicillin-resistant S aureus BSI, in particular, can be associated with a crude case-fatality rate ranging from 10% to almost 60%.^3- 4,10,18 Nevertheless, the real effect on mortality of methicillin resistance in staphylococcal bacteremia is still unknown.

Romero-Vivas et al¹⁰ showed that nosocomial bacteremia due to MRSA was associated with a 3-fold higher mortality than MSSA BSI after adjustment for several risk factors. The authors determined that this difference was unaffected by age, the severity of the underlying condition, the length of hospital stay, preinfection antibiotic treatment, and previous surgery, but they hypothesized that it might have been related to an inaccurate adjustment for the severity of illness and underlying diseases.

Because uncertainties remain about the mortality and the effect of MRSA BSI compared with MSSA BSI, we evaluated the contribution of methicillin resistance to the morbidity and mortality associated with S aureus bacteremia using 2 different and complementary approaches: a population-based retrospective cohort study and a matched case-control study.

This study was undertaken in the University Hospital of Geneva, Geneva, Switzerland, a 1500-bed health care center providing primary and tertiary care for Geneva and the surrounding areas (500000 inhabitants). About 40000 patients are admitted annually for a mean length of stay of 10 days.

At this center, the incidence rates of MRSA colonization or infection increased significantly between 1989 and 1995 from 0.05 cases per 100 admissions in 1989 to 0.81 cases per 100 admissions in 1995 (R² = 0.94, P<.001).¹⁹ After the implementation of infection control measures in 1993, the annual attack rate (±SD) of MRSA colonization or infection remained stable: 0.55±0.05 cases per 100 admissions (range, 0.51-0.62).

The objective of the study was to evaluate the effects of methicillin resistance on the morbidity and mortality associated with S aureus bacteremia. All patients with clinically substantial episodes of S aureus bacteremia between January 1, 1994, and December 31, 1995, were included in a retrospective population-based cohort study. The main outcome measure was the in-hospital mortality after staphylococcal BSI. In addition, a pairwise-matched (1:1) case-control study concerning 38 patients with MSSA and MRSA bacteremia was performed to confirm findings from the population-based cohort study.

All episodes of MSSA or MRSA bacteremia from January 1, 1994, through December 31, 1995, were identified using 3 sources of information: (1) database of the computerized clinical microbiology laboratory, (2) prospective surveillance records of patients with nosocomial BSIs from the infection control program, and (3) follow-up data of cases by infectious diseases consultants. When a patient had more than 1 episode of S aureus bacteremia or more than 1 hospital admission, only the first episode was considered. Patients younger than 16 years at the time of the onset of infection were excluded.

Staphylococcus aureus was identified using standard laboratory procedures.^19- 20 Methicillin resistance was determined according to methods recommended by the National Committee for Clinical Laboratory Standards for disk diffusion testing and the use of an oxacillin (0.85%) agar screening plate.²¹

Patients were considered to have staphylococcal BSI if cultures of 2 or more blood specimens were positive for S aureus within 24 hours and the clinical course was consistent with staphylococcal infection. Organisms of S aureus isolated from the bloodstream were considered to be contaminants if culture of only 1 blood specimen yielded the organism, if the clinicians or the infectious diseases consultant judged the organism to be a contaminant, and if antibiotic therapy directed against the organism was not administered. Bloodstream infection was considered community-acquired if the blood specimen for the first positive culture was obtained within 72 hours of admission or in the presence of S aureus infection at another body site at the time of hospital admission.²² An infection that was neither present nor incubating at the time of admission was considered nosocomial.²³

In the case of secondary bacteremia, a primary focus of infection was determined using the following definitions: The term "pneumonia" was retained for patients with clinical signs of lower respiratory tract infection associated with radiographic evidence of pulmonary infiltrates not attributable to other causes. An intravenous catheter was considered as the source of BSI if the catheter had been in place for at least 72 hours, culture of a quantitative catheter specimen yielded more than 100 colonies of S aureus,²⁴ or culture of a specimen of purulent drainage from the insertion site grew S aureus. Endocarditis was considered in patients with S aureus bacteremia and 1 or more of the following characteristics: surgical or autopsy findings consistent with endocarditis, echocardiographic evidence of valvular vegetation, and the presence of septic emboli. Urinary tract infection was considered if the patient had urinary symptoms and S aureus (>10⁵ colony-forming units per milliliter) was identified as the sole pathogen from urine. Soft tissue infection was considered in the case of a patient who had a pure culture of S aureus from a tissue or drainage specimen from the affected site and signs of infection. "Surgical wound infections" were diagnosed following the standard definitions from the Centers for Disease Control and Prevention.²⁵ "Primary bacteremia" defined conditions in which no primary focus could be determined.²⁶

In-hospital mortality reflects the mortality of the underlying illness and the mortality attributable to BSI.²⁶ The mortality associated with methicillin resistance in staphylococcal bacteremia was defined as the difference between the mortality among case-patients with MRSA BSI and the mortality among control patients with MSSA BSI.

Patients were observed from the day of hospitalization until hospital discharge or death. Death was attributed definitely to staphylococcal BSI in the presence of at least 1 of the following criteria: (1) blood cultures positive for S aureus at the time of death; (2) a persistent focus of staphylococcal infection associated with clinical signs of sepsis such as fever, leukocytosis, or hypotension; (3) death within 14 days of the documentation of BSI without another explanation; or (4) autopsy findings indicating staphylococcal infection as the cause of death. Staphylococcal BSI was considered the probable cause of death if there was a strong clinical association between the BSI and death that occurred within a week after the infection, in the absence of proof by autopsy and signs of ongoing sepsis. Death possibly related to staphylococcal BSI was considered when it occurred within a month after BSI, but was in occurrence not related to infection or a complication of the BSI.

Underlying diseases were classified according to McCabe and Jackson.²⁷ Patients' comorbidities were recorded according to predetermined definitions.^28- 29 In brief, comorbidities were defined as "active" if they required a recent treatment intervention or if they caused a permanent functional deficit. "Passive" comorbidities were defined as causing illness in the past without effect on patients' current health status. The number of cases of both types of comorbidities were added up, with active disorders accounting for 2 points and passive disorders accounting for 1 point.^28- 29

The systemic inflammatory response syndrome, sepsis, and end-organ dysfunction were defined as previously described.^30- 32 Patients' clinical condition during a bacteremic episode was classified according to the criteria of the sepsis severity score introduced by Skau et al.³³ Complications were considered infection-related if they were not associated with the patients' underlying diseases. Metastatic infections due to BSI and the need to drain the primary or the metastatic focus of infection were assessed.

All antimicrobial agents received for more than 48 hours during the previous month and following the onset of staphylococcal bacteremia were recorded. In our institution, the Division of Infectious Diseases' treatment policy recommends administering intravenous floxacillin (flucloxacillin sodium) as first-choice therapy for MSSA BSI and intravenous vancomycin hydrochloride for MRSA BSI. In a patient with possible or proven endocarditis, the additional use of an intravenous aminoglycoside is recommended during the first 5 days of therapy. No changes in antibiotic or infection control policies were instituted during the study.

Each case-patient was matched to the most suitable control patient on a subject-to-subject basis. Patients in whom MRSA BSI developed were designated as "cases" and those with MSSA BSI as "controls." The best control patient was selected in a stepwise manner using a scoring system similar to described methods.^34- 35 Control patients with staphylococcal bacteremia during a previous hospital stay were excluded. Matching criteria included sex, age, number of comorbidities, the McCabe and Jackson classification of the underlying disease, and the length of stay from admission to BSI (stratified in 4 categories: <72 hours, 3-7 days, 8-28 days, and >28 days).

To select the most suitable control for every patient with MRSA, we used a 16-point scoring system based on the above-listed matching variables: matching for the same length of stay to BSI (3 points if in the same category as mentioned above, no point if not); matching for sex (2 points in case of concordance, no point if discordant matching); matching for age (3 points if age difference ± <5 years, 2 points if age difference ± <10 years, no point if age difference >10 years); matching for comorbidities (5 points if complete concordance in comorbidities, 4 points if difference of 1-3 comorbidities, 3 points if difference of 4-5 comorbidities, no point if >5 comorbidities discordance); and matching for the same McCabe and Jackson classification of the underlying disease (3 points for concordance, no point in cases of discordance). When several control patients with MSSA were available with the same point score, 1 of them was chosen at random. Controls were selected without knowing the patients' survival status.

Of the 39 patients with MRSA bacteremia included in the population-based study, 38 (97%) were completely available for the case-control study. Thus, 38 case-patients with MRSA bacteremia were matched with 38 control patients with MSSA bacteremia. All case and control patients had their hospital stay during the study period. The average age was 65 years in case and control patients; 30 (79%) of the 38 case-control pairs were successfully matched for sex. Of the 38 case-control pairs, 36 (95%) were matched for the McCabe and Jackson classification of severity of illness, and 29 (76%) of 36 were matched for the length of hospital stay before BSI. All except 7 case-control pairs (31/38; 82%) were matched for the number of preexisting comorbidities. Among these 7 case-control pairs not successfully matched for the number of comorbidities according to our criteria, the difference in comorbidities exceeded 6 in only 3 cases. Overall, on the 16-point matching scale, the 38 controls had an average score of 12.7 points.

The results were analyzed using a commercially available software package (SPSS, version 4.0, SPSS Inc, Chicago, Ill). Means were compared by the Student t test and survival by the method of Kaplan-Meier and the log-rank test.³⁶ A Cox proportional hazard model was used to compare survival in both groups, adjusted for confounding variables.³⁷ Results of these comparisons were expressed as the hazard ratio. The McNemar test with continuity correction was performed to test the comparison of mortality and other outcome variables for the case and control patients. Conditional logistic regression analysis was used to calculate the odds ratio (OR) of mortality and confidence intervals (CI) in the matched case-control study. All tests of significance were 2-tailed; P values of less than .05 were considered significant.

From January 1, 1994, through December 31, 1995, 311 episodes of staphylococcal BSI occurred among 220 patients at the University Hospital of Geneva. Methicillin-resistant staphylococcal bloodstream isolates accounted for 22% (68/311). Only first episodes of infection were considered for each patient (MRSA, n=41; MSSA, n=152). Following medical records revision, we excluded 2 patients (4.9%) with MRSA and 7 (4.6%) with MSSA bloodstream isolates considered as contaminants. Thus, 184 patients with first episodes of staphylococcal bacteremia (MRSA, n=39; MSSA, n=145) that were considered clinically significant were included in the analysis.

The mean (±SD) age of the 39 patients with MRSA BSI was 67±18 years compared with 62±19 years for the 145 patients with MSSA BSI. In both groups, there was a predominance of male patients: 72% (n=104) in the group with MSSA, and 69% (n=27) in the group with MRSA. The mean (±SD) length of stay of patients with MRSA BSI from the time of admission to the onset of bacteremia was 32±34 days (median, 15), and it was 8±15 days (median, 4) for patients with MSSA bacteremia. In 70 patients (48%) with MSSA bacteremia, the infection developed within 72 hours after admission, compared with only 10 patients (26%) with MRSA bacteremia. All those 10 patients had been hospitalized in the previous 12 months in long-term care facilities, nursing homes, or in our institution. The total length of hospital stay differed in both groups (P=.01): patients with MRSA bacteremia had a mean (±SD) length of stay of 75±54 days (median, 73), compared with 37±38 days (median, 27) in those with MSSA infection. The in-hospital mortality was 28% (40 deaths) for patients with MSSA BSI vs 36% (14 deaths) in those with MRSA infection.

The mean (±SD) length of hospital stay from the time of the onset of BSI to discharge or death differed between the 2 groups (P=.03): 42±38 days for patients with MRSA and 29±33 days for patients with MSSA. This difference, however, was no longer significant after adjustment for age and length of hospital stay before bacteremia. Thus, to control for a possible difference in age and length of stay before the development of BSI in the survival analysis, we adjusted for these 2 confounding variables using a Cox proportional hazards model. Patients' survival following staphylococcal bacteremia due to MSSA or MRSA was similar (P=.30, log-rank test), with a relative hazard of death of 1.1 (95% CI, 0.5-2.1). Figure 1 shows the results of the survival analysis using the Cox model.

We compared 38 case-patients with MRSA bacteremia with 38 matched control patients with MSSA bacteremia. Table 1 shows characteristics of both groups of patients; as shown, the 2 patient populations were similar in the primary diagnosis for admission, exposure to surgery, the days in intensive care, and the use of medical devices before the occurrence of staphylococcal bacteremia. Patients in whom MRSA bacteremia developed had more frequent admissions to the hospital during the preceding 5 years than control patients with MSSA bacteremia (29/36 [81%] vs 14/36 [39%]). The use of antimicrobial agents before staphylococcal bacteremia occurred was more frequent among patients with MRSA (76%) than among those with MSSA (29%); however, the duration of exposure to antibiotics before the infection did not differ significantly between the 2 groups of patients (9.8 vs 8.1 days).

The sources of staphylococcal bacteremia in the 2 groups of patients included in the case-control study are listed in Table 2. Almost all patients had clinical signs of sepsis (37 of 39 case-patients and 35 of 39 control patients). Among the 4 patients without documented systemic inflammatory response syndrome, 2 had not been monitored for all variables within 24 hours after the onset of BSI, 1 had neutropenia following chemotherapy, and 1 had hypotension without complete systemic inflammatory response syndrome. The number of blood specimens that cultured positive for S aureus and positivity rates were similar in both groups of patients.

Metastatic infections were frequent in both groups (8 case-patients [21%] and 15 control patients [39%]; Table 2). Three patients in the MRSA group and 5 in the MSSA group had evidence of endocarditis with various other systemic foci of staphylococcal infection, including 1 episode of staphylococcal meningitis in each group. Three patients with MRSA had osteoarticular and 2 had splenic metastases. In the MSSA group, 4 patients had osteoarticular sites of dissemination, 4 had pulmonary empyema, 1 had a pacemaker infection, and 1 had disseminated pyomyositis.

Appropriate antibiotic treatment was received by 35 case-patients and 37 control patients. In the subgroup of patients with inadequate antimicrobial therapy, 1 patient with MRSA died before receiving any antibiotic treatment. In 1 case-patient and 1 control patient, physicians did not start adequate treatment within 72 hours despite the fact that both patients had signs of systemic infection; both patients survived sepsis without complications. Nine patients with MRSA (24%) and 7 patients with MSSA (18%) had blood specimens that cultured positive for the same pathogen for more than 1 day. All of these patients received adequate antimicrobial therapy; persistently positive blood cultures expressed ongoing infection like endocarditis or other deep-seated infectious foci. The median duration of antimicrobial treatment for the 2 groups was 17 days (range, 5-64) in the MRSA group vs 14 days (range, 6-93) in the MSSA group, excluding patients who died within 5 days after bacteremia developed.

In-hospital mortality was identical in case-patients with MRSA bacteremia and in control patients with MSSA bacteremia (13/38, 34%). Twenty matched pairs had a concordant outcome (16 lived and 4 died). Eighteen pairs had a discordant outcome, and in half of these pairs, either the case-patient or the control patient died. Of the 13 deaths occurring in case-patients, 11 (85%) occurred within the first 14 days after the onset of bacteremia, compared with 10 (77%) of 13 deaths in the control group. After conditional logistic regression analysis, methicillin resistance was not associated with an increased risk of death in patients with staphylococcal bacteremia (OR, 1.0; 95% CI, 0.4-2.5).

Autopsy was performed for 5 case-patients and 6 control patients (42% of patients who died). Considering major clinical and autopsy findings, 7 (54%) of the 13 case-patients with MRSA and 9 (69%) of the 13 control patients with MSSA died probably or definitely because of staphylococcal BSI. In the 7 fatal MRSA cases, the cause of death was as follows: 5 patients died of complicated septic shock, 1 had complicated endocarditis with meningitis, and 1 had pulmonary empyema with prolonged sepsis. In the group with MRSA BSI, 6 deaths were unrelated or only possibly related to the infection. In this group, 3 patients died of prolonged and complicated septic shock, 3 died of complicated endocarditis (1 patient with additional staphylococcal meningitis and renal abscesses), 2 of pulmonary empyema with prolonged sepsis, and 1 of bacterial septic shock due to bacteremia with multiple intra-abdominal abscesses. The other 4 deaths were probably unrelated to staphylococcal disease.

After excluding patients who died of unrelated conditions, S aureus was the definite or highly probable cause of death in 7 (18%) of the 38 patients with bacteremia due to MRSA, compared with 9 (24%) of 38 patients with MSSA BSI. The crude in-hospital mortality associated with nosocomial MSSA bacteremia was 46% (11/24 patients died) and that associated with community-acquired MSSA bacteremia, 14% (2/14 patients). Patients with MRSA bacteremia in whom the infection developed within 72 hours after admission had a better outcome than those in whom the infection developed later (2 deaths [22%] among 9 patients compared with 11 deaths [40%] among 29 patients).

The incidence of staphylococcal bacteremia is increasing in hospitals and in the community.^6,38 We investigated a cohort of 184 patients with S aureus BSI to understand the clinical importance of methicillin resistance and to clarify its impact on survival. Crude in-hospital mortality in patients with bacteremia due to MRSA was 36%, a rate similar to that in previous reports.^{5,7,9,39- 40} Mortality as high as 60% in patient populations with more severe underlying diseases or for nosocomial infections only has been reported.^3- 4,10,41 The in-hospital mortality for patients with MSSA bacteremia was 27% (40 deaths in 145 patients), consistent with that of other series.^26,39 After adjustment for major confounders, the outcome of patients with bacteremia due to MRSA and MSSA was similar: in-hospital mortality was 34%.

When patients who died of unrelated diseases were excluded, MRSA was the definite or highly probable cause of death in 18% (7/38) of the patients with bacteremia due to MRSA, compared with 24% (9/38) in closely matched control patients with MSSA infection. Besides death, both groups of patients with staphylococcal bacteremia had major systemic illness associated with high morbidity. This indicates that MRSA and MSSA may cause serious disease in patients with bacteremia and that previous studies^17,39 may have underestimated the occurrence of severe complications. In an important number of patients in our study with staphylococcal bacteremia (either MSSA or MRSA), signs of shock or metastatic foci of infection developed comparable to that in previously published series.^42- 45

The results of a population-based cohort and a case-control study did not show any difference in mortality associated with methicillin resistance in patients with staphylococcal bacteremia. Our results confirm and extend those of previous studies that failed to identify significant differences in associated mortality between MRSA- and MSSA-infected patients.^4,14,17,46

Patients' preexisting comorbidities are associated with prolonged hospital stays, an increased risk of infectious and noninfectious complications,^47- 48 and an important effect on short- and long-term survival in patients with bloodstream and other severe infections.^{28- 29,49- 51} The severity of underlying disease and preexisting comorbidities may distort the comparison of outcomes between patients with MRSA and MSSA BSI. A study by Romero-Vivas et al¹⁰ did not carefully adjust for the severity of illness and associated comorbidities. Their conclusion that methicillin resistance is independently associated with death due to S aureus bacteremia and that the risk of death is 3-fold higher for patients in the MRSA group than in the MSSA group may represent an overestimation of the virulence of methicillin resistance due to an inappropriate adjustment for underlying conditions. From our study, when patients with MRSA and MSSA bacteremia are carefully matched for the severity of underlying illness, the impact of methicillin resistance on the outcome is negligible. Nevertheless, 95% CIs surrounding the relative risk of death (population-based cohort study: 0.5-2.1; matched case-control study: 0.4-2.5) in our study overlap those reported by Romero-Vivas et al (OR, 3.0; 95% CI, 1.44-6.25).

Slightly more infectious complications after the onset of sepsis were observed in patients with MSSA bacteremia in the present study, consistent with nonsignificant trends observed by other investigators.^9,14 The observed increase may reflect the possible influence of the source of bacteremia. As described in other studies, the source of infection may play an important role in the severity of illness and associated mortality of staphylococcal bacteremia and BSIs due to other organisms.^46,52- 53 Higher proportions of patients with staphylococcal pneumonia in the MSSA group and urinary tract infections in the MRSA group may explain the observed differences concerning patients' condition after staphylococcal sepsis.

As mentioned by Marty et al,¹⁶ the comparison made between the outcomes of patients treated with glycopeptides for MRSA BSI and those treated with penicillinase-resistant β-lactam antibiotics for MSSA BSI may be questionable because differences in efficacy may exist when different classes of antibiotics are used to treat the same pathogen. It is difficult to evaluate the importance of different treatment regimens in this study, but this confounder might not be of crucial influence on the main outcome because deaths directly attributed to the infection occurred in similar proportions in both groups, as seen in other studies.^54- 55

A surprising finding in our study was the high number of patients admitted to our hospital with MRSA BSI. All these patients, however, had ongoing contact with the health care system in the year before infection or received care at home. We agree with reports by others^6,56 that MRSA BSI is becoming an important problem in nonhospital settings. Such an increase in the number of patients receiving care at home emphasizes the need for the surveillance of infections in this setting.⁶

Several criticisms might be raised concerning this study. First, the medical record of 1 case-patient who died could not be located. In any event, the exclusion of this patient from our analysis would not have changed the overall conclusion derived from the case-control study. Second, the primary objective of our matching procedure was to control for biased comparisons between case-patients and control patients. Major confounders of outcome in patients with bacteremia include older age, severity of underlying disease, length of hospital stay at the time of infection, and patients' preexisting comorbidities.²⁶ Matching was successful in 97% of patients in the described cohort. Because matching prevents a confounding effect of the matched variables on outcome, however, the effect of patients' comorbidities, in particular, could not be evaluated in this study. Moreover, overmatching may reduce the validity or the statistical efficacy of a case-control comparison. Consistency in the results obtained through the population-based and case-control approaches used in this study excludes a major selection bias in the choice of the control group. Therefore, overmatching cannot explain the lack of difference seen between the 2 study populations concerning survival. Third, one might challenge the validity of the scoring systems used to evaluate the severity of underlying illness (McCabe and Jackson) and the severity of sepsis at the onset of BSI. The appropriateness of the McCabe and Jackson scoring system has been recognized in several studies of survival of patients with sepsis.^28- 29,57 In contrast to the sepsis severity score, the use of the Acute Physiology and Chronic Health Evaluation II score might allow a more adequate assessment of the severity of S aureus bacteremia, as described recently.^45- 46 Although this scoring system might be highly predictive for both clinical course and final outcome in patients with S aureus bacteremia,⁴⁵ it was neither designed nor has it been validated for use in patients outside intensive care units. Therefore, we used the simpler sepsis severity score because it does not require a blood gas analysis for assessing patients' respiratory status. Further validation of the later scoring system in patients with bacteremia is needed.

In conclusion, our results do not indicate that methicillin resistance in S aureus has a significant impact on the mortality of patients with staphylococcal bacteremia after adjustment for major confounders.

Accepted for publication May 29, 1997.

Nadia Colaizzi provided technical assistance, and Sinclair Wynchank, MD, carefully read the manuscript.

Reprints: Didier Pittet, MD, MS, Infection Control Program, Division of Infectious Diseases, Department of Internal Medicine, Hôpital Cantonal Universitaire, 1211 Geneva 14, Switzerland (e-mail: dpittet@dminov1.hcuge.ch).