S. aureus bacteremia data were obtained from microbiology departments that prospectively collected information for >12 months on episodes of laboratory-confirmed bacteremia for the hospitals they serviced from January 1, 1999, to December 31, 2002. Information retrieved from existing databases included the total number of episodes of community- and hospital-onset bacteremia, the number of episodes of community- and hospital-onset MRSA and MSSA bacteremia, the total number of hospital separations (defined as completed hospital admissions), and the mean length of stay. Multiple positive blood cultures in the same patient within 14 days were considered a single episode. Episodes were considered to have a hospital onset when the first positive blood culture was collected >48 hours after admission to hospital. All other infections were designated community onset (for example, day-only dialysis related episodes were defined as community onset, as were infections with their onset in nursing homes). Organism identification and susceptibility testing were by standard methods. All these laboratories participate in external quality assurance programs as well as AGAR national surveys (3,12), which have quality control procedures to ensure these laboratories accurately detect methicillin resistance. Published data were used for the details on the number of hospital beds and separations for Australia and for the classification of different types of hospitals (13). The term separations, rather than admissions, is used in the published data because hospital abstracts for inpatient care are based on information gathered at the time of discharge. We have used the more commonly applied term of admissions, however, for these episodes.

In Australia, most healthcare-associated MRSA is caused by 1 clone defined by multilocus sequence type (ST) 239; this clone is characteristically resistant to multiple antimicrobial agents, including gentamicin (3,12). Most of the remaining healthcare-associated infections are caused by a recently introduced strain, ST22, which is indistinguishable from epidemic MRSA-15 in the United Kingdom. It is invariably resistant to ciprofloxacin (12). Thus, in Australia, MRSA that is acquired in the community and is sensitive to both ciprofloxacin and gentamicin is not likely to be associated with healthcare facility acquisition. We used this pattern as a surrogate marker for community acquisition of MRSA without healthcare-associated risk factors.

We detected 12,771 bloodstream infections in the 17 hospitals participating in this study (12 principal referral metropolitan, 3 large metropolitan, 1 private hospital, and 1 medium-sized public hospital, and 1 private hospital with 2,013,534 total separations; see Table 1). There were 3,192 episodes of S. aureus bacteremia identified (i.e., 25% of the total true bloodstream infections). The median rate of S. aureus bacteremia was higher in the principal referral metropolitan hospitals (1.59/1000 admissions) than in large metropolitan hospitals (1.3) or the private hospital (0.6). The range varied from 0.60 to 3.24 (Table 2). The median rate of community-onset bacteremia episodes was 0.80/1000 admissions (range 0.11–0.99). The median rate of hospital-onset bacteremia was 0.72 episodes/1,000 admissions (range 0.13–1.30). The median rate of hospital-onset MRSA episodes was 0.22/1,000 admissions (range 0–0.89). When expressed as MRSA episodes per 1,000 occupied bed days (OBDs), the rates varied from 0 to 0.30 with a median rate of 0.08. If day-only cases are removed from the denominator then the median rate was 0.10 per 1,000 OBDs (range 0–0.39).

Of these 3,192 SAB episodes, 1,621 (51%) were of hospital onset, and 1,571 (49%) had their onset in the community. Of those with a hospital onset, 40% were MRSA in comparison to 12% with a community onset. Of all MRSA bacteremia episodes, 23% had a community onset, and 77% had hospital onset. Of the 193 community-onset episodes of MRSA that occurred, only 47 (24%) had a sensitivity pattern (sensitive to gentamicin and ciprofloxacin) that suggests that they were community acquired.

When both MRSA and MSSA were considered, data were available for 560 community-onset SAB infections (but only from 4 hospitals). The proportions of these episodes that were noninpatient, healthcare-associated were 35%, 42%, 18% and 16%, respectively (from hospitals A, D, E, and N). In those hospitals, the percentage of S. aureus episodes that were healthcare associated overall (i.e., all hospital-onset cases and those community-onset cases associated with healthcare exposure) were 75%, 69%, 64%, and 36%, respectively.

Mortality data were available for 526 patients from 2 hospitals. At hospital E, the mortality rate at day 7 was 10% (27 of 267 patients). When a subgroup of these patients at hospital E (52 patients) was followed for a longer period (2001–2002), the mortality rate was 23% at 30 days and 35% at 6 months. For those 24 patients with a community-onset episode of bacteremia that was not healthcare associated, mortality rates were 6% at day 7, 17% at 1 month, and 21% at 6 months, respectively. At hospital H (259 patients), the mortality rate at 30 days was 19%. At hospital H, the mean length of stay for those with SAB was 25.6 days compared to 6.2 days in matched controls. The mean length of stay was longer for MRSA infections (39.2 days) than for MSSA infections (23.3 days).

The rates of S. aureus bacteremia in different hospital populations were used to estimate the incidence for Australia. Using our median bacteremia rate for S. aureus bacteremia in different types of public hospitals (1.27/1,000 admissions, range 0.68–3.24) and in private hospitals (0.6/1,000 admissions), we estimated ≈6,900 episodes per year nationally (range 3,826–20,658) or 35/100,000 per year (Table 3, Table 4). Some data are available from other countries for comparison,; the lowest annual rates are in Wales (23/100,000) and the highest in the United States (56/100,000; Table 4). However only 2 countries, Denmark and England, appeared to have comprehensive collection systems, and their rates were 29/100,000 and 37/100,000, respectively (19,28,29).

S. aureus bacteremia is very common. Approximately one fourth (26%) of all S. aureus bacteremia episodes were caused by MRSA, and, as expected, the onset of most of these episodes was in hospitals (77%). Notable, however, 12% of all community-onset S. aureus infections were MRSA, which was 23% of all MRSA bloodstream infection episodes. A recent study from the United States similarly showed that 15% of community-onset SAB episodes were MRSA (14). Most of the community-onset strains in our study were multiresistant or phenotypically consistent with UK EMRSA-15 (15) and thus most likely to have been acquired by patients who had previous hospital contact, with nursing home contact a major factor in at least 1 of the hospitals in this study (hospital G). However, approximately one fourth of these community-onset MRSA infections were caused by other phenotypes of non–multiresistant MRSA and thus more likely to be true community-acquired episodes of MRSA bacteremia. Severe cases of MRSA bacteremia not associated with prior healthcare contact have been reported previously in Australia (7,9,16).

Use of the >48 hours postadmission definition of hospital onset underestimates the number of episodes of bacteremia that are healthcare associated. Many patients with chronic conditions are now treated in the community or on a day-only basis. Vascular lines are increasingly used in the community and outpatient settings, providing a potential source of bacteremia. The collection of data on the true association of episodes of bacteremia to health care is time-consuming and was not done by most institutions participating in this study. However, 3 principal referral hospitals (hospitals A, D, and E) did collect these data for 971 episodes, and 64%–75% of their total S. aureus bacteremia episodes were healthcare associated. Only 46%–61% of the episodes were acquired while the patient was an inpatient (i.e., >48 h in hospital). This finding means that in these larger hospitals approximately one third of healthcare-associated episodes were acquired by either outpatients or short- stay patients. These episodes are better defined as “noninpatient, healthcare-associated.” In a recent study in the United States, 62% of their community-onset SAB infections were healthcare related (with intravenous [IV] catheters the most common clinically apparent site of infection) (14). On the basis of our data, we conclude that in Australia approximately two thirds of all SAB episodes were associated with healthcare or medical procedures (i.e., all hospital-onset and approximately one third of community-onset episodes). A similar situation is evident in Denmark (17), where in 2002, at least 59% of all S. aureus infections were associated with healthcare procedures. Clearly, substantial scope exists internationally for interventions in healthcare settings to decrease the numbers of these episodes (especially those related to IV catheters). Interventions to reduce S. aureus bacteremia need to target healthcare-associated infections in the broadest sense and include those following non–inpatient-related medical procedures.

Community-onset infections that have no healthcare association are also common and associated with a high death rate (17% and 19% at hospitals E and H at 1 month, respectively). How best to intervene to decrease these infections is difficult to determine. Vaccination is 1 possibility for the future; a recent trial of a conjugated capsular polysaccharide vaccine in renal dialysis patients estimated efficacy at ≈60% (18). However, vaccination for the general population is unlikely to be available soon. We should therefore concentrate on reducing the number of deaths from established infections. Because the mortality rate associated with community-acquired bacteremia increases with inadequate empiric therapy (19), all efforts should be made to promote compliance with published guidelines for treatment of severe staphylococcal sepsis, including adequate duration of therapy.

Available data suggest that staphylococcal bacteremia is a major global health problem. The median death rate for MSSA infections is 25%, and for MRSA infections, 34% (20). Thus, >1,700 deaths in Australia are likely associated with S. aureus bacteremia per year (assuming 6,900 episodes or a bacteremia rate of 35/100,000/year). This estimate of the rate of SAB is similar to England (21) but much lower than in the United States on the basis of the rate derived from the figures available in the only comparative study (55/100,000) (14). Our estimated rate in Australia is higher than that in Denmark (17,21). It is also higher that those reported from Wales (22) and Ireland (23) (Table 4); however, all episodes from these last 2 countries likely were not reported in their voluntary reporting schemes. England changed recently from a similar voluntary reporting scheme to a compulsory scheme, and the numbers of reported episodes increased by almost 50% (24).

The rate of MRSA bacteremia in England was higher per 1,000 OBDs than in our figures from Australia (0.17 compared to 0.10 episodes per 1,000 OBDs, respectively). MRSA was a substantial cause of episodes of SAB in this study (26%). However, this percentage was lower than that seen in most other countries (e.g., Wales, 47%; Table 4) with the notable exception of Denmark (0.6% in 2002) (17).

We may have overestimated the number of cases of bacteremia occurring in Australia because of the overrepresentation of larger hospitals in our survey. However, these hospitals participated because they had in place surveillance systems for measuring all episodes of bacteremia. The rates of SAB may have been relatively lower in these hospitals because they were also more likely than were hospitals without surveillance systems to have infection control programs in place to try to decrease the numbers of these episodes. If systems were in place that better captured and reported on all bacteremia episodes in well-defined populations (e.g., all of Australia or a state), then this would give a more accurate rate. Such systems appear only to be in place in Denmark and England (17,21,24). Currently, no such systems are operating in Australia. Limited data are available from a voluntary surveillance system in Victoria (25) that captures an estimated two thirds of bacteremic episodes that occur in that state. The extrapolated rate (27 episodes/100,000 persons/year; Table 4) was slightly lower than what we estimated for all of Australia in this study.

Substantial illness and increased medical costs are also associated with staphylococcal bacteremia. S. aureus bacteremia is often related to serious infections, including endocarditis, osteomyelitis, and septic arthritis. It frequently results in prolonged hospital admission and increased costs. In hospital H, the average length of stay for patients with S. aureus bacteremia was 26.5 days. In South Australia, the estimated additional cost of each episode of hospital-acquired S. aureus infection was $22,000 in 1998 (26). Nationally, these South Australian costs translate to additional hospital costs of ≈$150 million dollars ($22,000 x 6,900 episodes).

Treatment of S. aureus infections is complicated by the high prevalence of antimicrobial drug resistance. Although this has long been the case with multiresistant strains of MRSA in hospitals, the spread of hospital strains into the community, as well as the emergence of unique strains of MRSA unrelated to health care, have made this an issue of general importance. At least 3 community strains of MRSA are currently circulating in Australia (10,27,28). Two of these 3 community strains carry the gene for Panton-Valentine leukocidin, which is associated with subcutaneous abscess formation and necrotizing pneumonia. A number of reports have already highlighted the clinical impact of infection due to these strains (9,28–30). Surveillance data show that their prevalence is increasing in our capital cities, but the situation in rural Australia is not well documented (3). This increase will inevitably affect guidelines for empirical antimicrobial drug prescribing for staphylococcal infections and for patients in the community with suspected SAB. Further surveillance of staphylococcal infections, including bacteremia, is warranted to guide recommendations for empirical therapy and infection control interventions.