Recent observational studies have suggested possible reductions in mortality in patients receiving cefazolin versus antistaphylococcal penicillins. We examined 90-day mortality in patients receiving cefazolin compared to nafcillin for methicillin-susceptible Staphylococcus aureus (MSSA) bloodstream infection (BSI). We identified persons with MSSA BSI admitted to San Francisco General Hospital from January 2008 to July 2013 through a hospital-wide infection surveillance system and confirmed 90-day mortality using U.S. national vital registries. We included persons receiving cefazolin or nafcillin as the predominant intravenous antimicrobial agent; all participants received inpatient Infectious Diseases service consultation. We estimated the association between receipt of cefazolin and 90-day risk of death by multivariate logistic regression, including a propensity score for receiving cefazolin as the second predictor. Of 230 MSSA BSI cases, 30 received nafcillin and 70 received cefazolin as the predominant antimicrobial; 10 died within 90 days, 5 from each group. Unadjusted analysis showed substantial but not statistically significant reduced odds of death in those receiving cefazolin (odds ratio, 0.38; 95% confidence interval [CI], 0.10 to 1.44). Multivariate analysis with propensity scores found a similar adjusted odds ratio (0.40; 95% CI, 0.09 to 1.74; P = 0.22). We found a large reduction in 90-day mortality in those receiving cefazolin compared to nafcillin for MSSA BSI, but this finding was not statistically significant. The magnitude of effect seen in this and other studies justifies further study.