Objectives To review the epidemiology of native septic arthritis to establish local guidelines for empirical antibiotic therapy as part of an antibiotic stewardship programme. Methods We conducted a 10 year retrospective study based on positive synovial fluid cultures and discharge diagnosis of septic arthritis in adult patients. Microbiology results and medical records were reviewed. Results Between 1999 and 2008, we identified 233 episodes of septic arthritis. The predominant causative pathogens were methicillin-susceptible Staphylococcus aureus (MSSA) and streptococci (respectively, 44.6% and 14.2% of cases). Only 11 cases (4.7%) of methicillin-resistant S. aureus (MRSA) arthritis were diagnosed, among which 5 (45.5%) occurred in known carriers. For large-joint infections, amoxicillin/clavulanate or cefuroxime would have been appropriate in 84.5% of cases. MRSA and Mycobacterium tuberculosis would have been the most frequent pathogens that would not have been covered. In contrast, amoxicillin/clavulanate would have been appropriate for only 75.3% of small-joint infections (82.6% if diabetics are excluded). MRSA and Pseudomonas aeruginosa would have been the main pathogens not covered. Piperacillin/tazobactam would have been appropriate in 93.8% of cases (P < 0.01 versus amoxicillin/clavulanate). This statistically significant advantage is lost after exclusion of diabetics (P = 0.19). Conclusions Amoxicillin/clavulanate or cefuroxime would be adequate for empirical coverage of large-joint septic arthritis in our area. A broad-spectrum antibiotic would be significantly superior for small-joint infections in diabetics. Systematic coverage of MRSA is not justified, but should be considered for known carriers. These recommendations are applicable to our local setting. They might also apply to hospitals sharing the same epidemiology.