Methicillin-resistant Staphylococcus aureus (MRSA) is a major nosocomial pathogen worldwide. To investigate an association between antimicrobial use and MRSA, a case control study of 121 patients infected with MRSA compared with 123 patients infected with methicillin-susceptible S. aureus (MSSA) was carried out. Antimicrobial use was analysed by three different logistic regression models: all beta-lactam antibiotics, beta-lactam antibiotics grouped in classes and antimicrobial use in grammes. Patients infected with MRSA tended to have more co-morbidities, longer lengths of stay (LOS) and greater exposure to antibiotics than MSSA-infected patients. Multivariate analysis identified levofloxacin [odds ratio (OR) 8.01], macrolides (OR 4.06), previous hospitalization (OR 1.95), enteral feedings (OR 2.55), surgery (OR 2.24) and LOS before culture (OR 1.03) as independently associated with MRSA infection. All models were concordant with the exception of macrolides, which were not significant based on the number of grammes administered. There were no significant differences in the types of infection or the attributed mortality in either group. MRSA-infected patients had a significantly longer LOS before infection [18.8 +/- 18.2 compared with 8.4 +/- 6.9 (P < 0.001)] and a significantly longer post-diagnosis LOS [27.8 +/- 32.9 compared with 18.6 +/- 21 (P = 0.01)] than MSSA-infected patients.