Abstract Introduction Recommended duration of antibiotic treatment of Staphylococcus aureus bacteremia (SAB) is frequently based on distinguishing uncomplicated and complicated SAB, and several risk factors at the onset of infection have been proposed to define complicated SAB. Predictive values of risk factors for complicated SAB have not been validated, and consequences of their use on antibiotic prescriptions are unknown. Methods In a prospective cohort, patients with SAB were categorized as complicated or uncomplicated through adjudication (reference definition). Associations and predictive values of 9 risk factors were determined, compared with the reference definition, as was accuracy of Infectious Diseases Society of America (IDSA) criteria that include 4 risk factors, and the projected consequences of applying IDSA criteria on antibiotic use. Results Among 490 patients, 296 (60%) had complicated SAB. In multivariable analysis, persistent bacteremia (odds ratio OR, 6.8; 95% confidence interval CI, 3.9–12.0), community acquisition of SAB (OR, 2.9; 95% CI, 1.9–4.7) and presence of prosthetic material (OR, 2.3; 95% CI, 1.5–3.6) were associated with complicated SAB. Presence of any of the 4 risk factors in the IDSA definition of complicated SAB had a positive predictive value of 70.9% (95% CI, 65.5–75.9) and a negative predictive value of 57.5% (95% CI, 49.1–64.8). Compared with the reference, IDSA criteria yielded 24 (5%) false-negative and 90 (18%) false-positive classifications of complicated SAB. Median duration of antibiotic treatment of these 90 patients was 16 days (interquartile range, 14–19), all with favorable clinical outcome. Conclusions Risk factors have low to moderate predictive value to identify complicated SAB and their use may lead to unnecessary prolonged antibiotic use. Risk factors have low to moderate predictive value to identify complicated SAB. Using risk factors as part of the definition of complicated SAB may lead to misclassification and overtreatment.