This study shows that SAB is an important cause of morbidity and mortality in northeast Thailand. The all-cause and S. aureus
-attributable mortality rates in our study, 52% and 44% respectively, are considerably higher than mortality rates for SAB reported from temperate industrialised countries 
. A study of SAB conducted in the USA by Fowler et al 
found an all-cause and attributable mortality of 22% (157/724) and 12% (86/724), respectively, but this excluded deaths that occurred before culture results were available. The addition of these cases would have given an all-cause and S. aureus
-attributable mortality rate of 33% and 23%, respectively. Published studies of all-cause bacteraemia in Asia have identified S. aureus
as a major cause of bacteraemia, accounting for both community-acquired and hospital-acquired disease 
. If our data showing that S. aureus
accounts for 1% of all in-patient deaths is representative of this populous region (half the world's population lives within 2000 miles of northeast Thailand), then S. aureus
would be a major contributor to preventable mortality worldwide. To our knowledge this is the first published prospective study to focus on SAB in tropical Asia. This is noteworthy because papers describing all-cause bacteraemia in Asia rarely give details by causative organism, resulting in an under-appreciation of the morbidity and mortality burden due to S. aureus
in the tropics. Sappasithiprasong Hospital is the regional hospital for the province with referrals from clinics and hospitals in the province, but also provides primary care services to a large local population so accurate estimates of disease incidence were not possible.
In our study, SAB was most common at the extremes of age, a similar pattern to that described in a temperate industrialised country 
. Among the 17 patients under 1 year of age, 16 (94%) had hospital-acquired infections and nearly a third (29%) were premature or very low birth weight babies. This suggests the particular vulnerability to acquiring nosocomial SAB of children less than 1 year of age requiring multiple interventions and prolonged stays in hospital. A higher incidence in those aged under 1 year and the predominance of nosocomial infections in this age group has been described in Denmark 
. The increase in numbers of cases and mortality with age also mirror the rise in co-morbidities with age, as noted in Denmark 
The broad range of clinical manifestations seen in patients with SAB in industrialised countries was observed in this series. Our findings suggest that the burden of S. aureus
disease in the tropics exceed current perceptions, and demonstrate that serious invasive infections are common. This has an important bearing on antimicrobial therapy because management of deep infections and bacteraemia requires effective antimicrobial therapy whereas drainage of superficial pus collections can result in cure irrespective of antibiotic therapy 
The rapid deterioration and short median time to death (3 days) of patients dying from S. aureus
septicaemia meant that echocardiograms could not be performed in half of the patients. However, the prevalence of echocardiographically-confirmed endocarditis of 14% (7/49) is comparable to that seen in industrialised countries 
, which has important implications for many tropical countries where restricted or delayed access to echocardiography is liable to be an issue. Our prevalence may be an underestimate since those patients who died earlier may have been more likely to have endocarditis. A quarter of the patients with endocarditis died, which is at the lower end of the mortality range reported for S. aureus
endocarditis (25–47%) 
and may be a further indication that a number of patients who died prior to echocardiography had undiagnosed endocarditis. The heart valves affected by endocarditis were predominantly left-sided, which is in keeping with the low number of intravenous drug users in our study. Half the patients with endocarditis were teenagers, which is a younger age group than typically seen in industrialised temperate countries 
but is described in other tropical countries 
, often as a result of rheumatic heart disease and uncorrected congenital heart disease. However, none of our patients with endocarditis had known valvular abnormalities.
MRSA was responsible for nearly one third of cases of SAB, all of which were healthcare-associated. Although community-associated MRSA is a major problem elsewhere, we find no evidence for this in our setting where a clone defined by multilocus sequence typing as sequence type 239 predominates 
. Over half of our patients had healthcare-associated infections which suggests that SAB in Thailand is strongly related to healthcare, mirroring industrialised countries 
. Although there are comprehensive hospital infection control guidelines in Sappasithiprasong hospital, implementing these is difficult due to a bed occupancy rate that often exceeds 100% and a lack of infrastructure, such as a scarcity of isolation rooms and only two hand wash basins on each of the general wards. However, alcohol hand rub is available in the intensive care units. Addressing such infection control measures would require a significant increase in investment, although this would be offset by reducing the expense of nosocomial infections. Our finding that over half the infections were healthcare-associated indicates that hospital infection control is an important area for clinical, microbiological and economic research if improvements are to be made.
Significant delays in receiving effective antibiotic therapy were seen with MRSA bacteraemia compared with MSSA bacteraemia. Vancomycin is available in this setting but is not used in the empiric regimen for a patient with suspected bacterial sepsis unless they are already known to be MRSA positive. Monitoring of vancomycin levels is not possible at Sappasithiprasong Hospital, which may lead doctors to give lower doses to reduce the perceived risk of toxicity. The mortality rates for patients with MRSA who received effective and ineffective empirical antibiotics were 0% and 69%, respectively, indicating the importance of early effective treatment. There was a median delay in effective antibiotic therapy of 3 days for MRSA patients. Delays in treating SAB are known to have an adverse effect on outcome 
. The common usage of ampicillin and gentamicin as empirical therapy in children under the age of 1 year should be revised in light of our finding that 53% of SAB cases in this age group were MRSA and all MRSA strains were resistant to gentamicin. Although alternative antibiotics to vancomycin may be appropriate therapy for MRSA, in our setting the high rates of resistance found on susceptibility testing indicate that these alternatives would not be effective.
This study has demonstrated that S. aureus is a significant pathogen in northeast Thailand, with comparable clinical manifestations and a similar endocarditis prevalence but higher mortality than industrialised countries. The factors contributing to this high death rate, such as delayed presentation to hospital and the early management of sepsis, require further evaluation. The majority of infections were associated with exposure to healthcare settings and MRSA was associated with a considerable burden of disease and a high mortality. Revisions to the empirical prescribing practices to include MRSA therapy could be associated with significant benefit. An initiative to raise the profile of infection control is needed, together with work to characterise the burden and causes of hospital-acquired infections, including patient-to-patient transmission of MRSA, such that cost-effective solutions can be devised appropriate to this setting.