Objective: This study aimed to analyze the prognosis of women with breast cancer by molecular subtypes, sociodemographic variables, and clinical and treatment characteristics. Methods: This hospital-based retrospective cohort study analyzed 1,654 women over 18 years of age diagnosed with invasive breast cancer from 2000 to 2018. Data were extracted from Brazil’s Oncocenter Foundation of São Paulo. The variables analyzed were age, histology, molecular subtypes, clinical staging, treatment type, and diagnosis-to-treatment time. Cox regression analysis was applied to estimate death risk. Results: Women with HER-2-positive (nonluminal) and triple-negative molecular subtypes were more than twice more likely to be at risk of death, with adjusted hazard ratio — HRadj=2.30 (95% confidence interval — 95%CI 1.34–3.94) and HRadj=2.51 (95%CI 1.61–3.92), respectively. A delayed treatment associated with an advanced clinical stage at diagnosis increased fourfold the risk of death (HRadj=4.20 (95%CI 2.36–7.49). Conclusion: In summary, besides that interaction between advanced clinical stage and longer time between diagnosis and treatment, HER-2-positive (nonluminal) and triple-negative phenotypes were associated with a worse prognosis. Therefore, actions to reduce barriers in diagnosis and treatment can provide better outcome, even in aggressive phenotypes.