OBJECTIVES: The main objective was to study the impact of in-hospital bacteraemia caused by Staphylococcus aureus on mortality within 90 days after admission. We compared methicillin-resistant S. aureus (MRSA) with methicillin-susceptible S. aureus (MSSA).

PATIENTS AND METHODS: The study population consisted of adult residents of Tayside, Scotland, UK, from 1 January 2005 to 30 September 2006 who had a new admission to Ninewells Hospital between 1 July 2005 and 30 June 2006. All patients (n = 3132) in the same wards as the patients infected with S. aureus were included. We addressed key weaknesses in previous studies by using a cohort design and applying a multistate model, which addressed the temporal dynamics. Critically, the model recognized that death and discharge from the hospital are competing events and that delay in discharge independently increases the risk of death.

RESULTS: The cohort included 3132 patients, of whom 494 died within 90 days after admission, 34 developed MRSA bacteraemia and 26 MSSA bacteraemia in the hospital. In comparison with patients without S. aureus bacteraemia, the death hazard was 5.6 times greater with MRSA [95% confidence interval (CI) 3.36-9.41] and 2.7 times greater with MSSA bacteraemia (95% CI 1.33-5.39). After adjustment for co-morbidity, hospitalization, age and sex, the death hazard was 2.9 times greater with MRSA (95% CI 1.70-4.88) and 1.7 times greater with MSSA bacteraemia (95% CI 0.84-3.47).

CONCLUSIONS: Time-dependent models such as the proposed multistate model are necessary to address the temporal dynamics of admission, infection, discharge and death. The impact of S. aureus bacteraemia on mortality should be considered on two levels: the burden of disease, i.e. nosocomial infection with S. aureus bacteraemia, and the burden of resistance to methicillin.