Persistent Staphylococcus aureus bacteremia (SAB-P) is well known but poorly delineated due to unclear definition. We retrospectively studied 78 patients with SAB-P using a stringent definition (bacteremia for > or = 7 d), in a single teaching hospital, during 1 January 2002 to 30 June 2003 and 1 November 2005 to 31 December 2006 to determine whether the frequency, risk factors and outcome changed over time. SAB was encountered in 354 and 259 instances during the 2 periods, respectively. Patients' characteristics changed with increasing organ dysfunction score (2.9+/-1.7 vs 3.4+/-1.4; p <0.001), patients with invasive devices (27.7% vs 41.3%; p=0.001), hemodialysis dependence (19.2% vs 27.8%; p=0.04), MRSA (50.8% vs 64.5%; p=0.001), and vancomycin treatment (57.9% vs 67.2%; p=0.02). SAB-P frequency increased slightly (11.0% vs 15.1%). Risk (associated) factors for SAB-P (identified by logistic regression) were metastatic infection (OR=5.60; 95% CI 3.00 - 10.47), vancomycin treatment (OR=4.17; 95% CI 2.14 - 8.11), endovascular sources (OR=3.35; 95% CI 1.92 - 5.85) and diabetes (OR=2.14; 95% CI 1.26 - 3.64). SAB- and SAB-P-associated case-fatality did not change (23.2% vs 18.5% and 25.6 vs 30.8%, respectively). All survivors ultimately achieved clearance. These findings suggest that patients with SAB are changing over time. Additionally, SAB-P frequency is higher than previously reported. SAB-P rise is probably due to increasing SAB, MRSA, and patients at risk for complications. Innovative approaches should target novel treatment modalities and risk reduction.