OBJECTIVES To describe the predictive factors for the isolation of fluoroquinolone-resistant or extended- spectrum beta-lactamase (ESBL)-producing Escherichia coli and their impact on bacteraemia outcome. METHODS Analysis of E. coli bacteraemia episodes prospectively collected through a blood culture surveillance programme from January 1991 to December 2007. RESULTS Out of 18 080 episodes, 4758 (26%) E. coli bacteraemias were reported in the period of study. Mortality was noted in 440 cases (9%). Fluoroquinolone-resistant strains were reported in 1300 (27%) cases and ESBL-producing strains in 211 cases (4%). One hundred and seventy-eight strains out of 211 (84%) ESBL-producing E. coli were isolated since 2001. The two main independent risk factors for mortality were shock (OR: 10.28, P < 0.001) and inappropriate empirical therapy (OR: 4.83, P < 0.001). Inappropriate empirical therapy was significantly more frequent for infections caused by fluoroquinolone-resistant strains (n = 203, 16%, P < 0.001) and ESBL-producing strains (n = 110, 52%, P < 0.001). Independent factors associated with the isolation of a fluoroquinolone-resistant strain were: nosocomial origin (OR: 1.61, P < 0.001); urinary catheterization (OR: 2.44, P < 0.001); and previous therapy with a fluoroquinolone (OR: 7.41, P < 0.001). The independent risk factors associated with the isolation of an ESBL-producing strain were: nosocomial origin (OR: 1.68, P = 0.03); urinary catheterization (OR: 1.88, P = 0.001); and previous beta-lactam antibiotic therapy (OR: 2.81, P < 0.001). CONCLUSIONS Inappropriate empirical therapy was the strongest independent factor that we could modify to improve mortality in E. coli bacteraemia and was more frequent in cases caused by fluoroquinolone-resistant or ESBL-producing strains. Nosocomial acquisition, urinary catheterization and previous therapy with a fluoroquinolone or beta-lactam were predictive factors for infection with an antibiotic-resistant strain.