Purpose: The aim of the present study was to determine whether de-escalation guided by blood cultures for patients with a diagnosis of sepsis, severe sepsis or septic shock reduces mortality, and antimicrobial drug resistance (ADR).Methods: A prospective, single-center, cohort study was conducted with adults admitted to the ICU with a diagnosis of sepsis, severe sepsis, or septic shock at a public hospital in Sorocaba, State of São Paulo, Brazil, from January 2013 to December 2013. We excluded patients who had negative blood cultures. Patients who had replaced the initial empirical broad-spectrum antibiotic therapy (EAT) by the antibiotic therapy guided by blood cultures were compared with those who continued receiving EAT. The outcome included mortality and antimicrobial drug resistance. We used the Cox regression (proportional hazards regression) and the Poisson regression to analyze the association between antibiotic therapy guided by blood cultures (ATGBC) and outcomes. The statistical adjustment in all models included the following variables: sex, age, APACHE II (Acute Physiology And Chronic Health Evaluation II) score and SOFA (Sequential Organ Failure Assessment) score.Results: Among the 686 patients who were admitted to the intensive care unit, 91 were included in this study. The mean age of the patients was 52.7 years (standard deviation = 18.5 years) and 70.3% were male. EAT was replaced by ATGBC in 33 patients (36.3%) while 58 patients (63.7%) continued receiving EAT. Overall hospital mortality decreased from 56.9% in patients who received EAT to 48.5% in patients who received ATGBC [Hazard ratio- HR 0.44 (95% CI 0.24–0.82), p = 0.009]. There was no association between ATGBC and ADR [HR 0.90 (95% CI 0.78 – 1.03) p = 0.15].Conclusions: Although the early and appropriate empirical EAT is undoubtedly an important factor prognostic, ATGBC can reduce the mortality in these patients.