Empirical antibiotic therapy with a beta-lactam is the standard of care in febrile neutropenia (FN) and is given to prevent early death. The addition of vancomycin is recommended in certain circumstances, but the quality of evidence is low, reflecting the lack of clinical data. In order to characterize the epidemiology of early death and shock in FN, we reviewed all episodes of FN from 2003 to 2017 at University Hospital, Federal University of Rio de Janeiro, and looked at factors associated with shock at first fever and early death (within 3 days from first fever) by univariate and multivariate analyses. Among 1,305 episodes of FN, shock occurred in 42 episodes (3.2%) and early death in 15 (1.1%). Predictors of shock were bacteremia due to Escherichia coli (odds ratio [OR], 8.47; 95% confidence interval [95% CI], 4.08 to 17.55; P < 0.001), Enterobacter sp. (OR, 7.53; 95% CI, 1.60 to 35.33; P = 0.01), and Acinetobacter sp. (OR, 6.95; 95% CI, 1.49 to 32.36; P = 0.01). Factors associated with early death were non-Hodgkin’s lymphoma (OR, 3.57; 95% CI, 1.18 to 10.73; P = 0.02), pneumonia (OR, 21.36; 95% CI, 5.72 to 79.72; P < 0.001), shock (OR, 11.64: 95% CI, 2.77 to 48.86; P = 0.01), and bacteremia due to Klebsiella pneumoniae (OR, 5.91; 95% CI, 1.11 to 31.47; P = 0.03). Adequate empirical antibiotic therapy was protective (OR, 0.23; 95% CI, 0.07 to 0.81; P = 0.02). Shock or early death was not associated with Gram-positive bacteremia; catheter-related, skin, or soft tissue infection; or inadequate Gram-positive coverage. These data challenge guideline recommendations for the empirical use of vancomycin at first fever in neutropenic patients.