We performed a retrospective matched-cohort study to determine the risk factors for mortality among patients with Escherichia coli infections. From January 1996 to December 2003, 100 hospitalised patients with extended-spectrum beta-lactamase (ESBL)-producing E. coli infections were compared with patients not infected with ESBL-producing E. coli. These patients were selected according to the same site of infection and the closest date of admission. Comparison of the two groups showed that empirical antibiotic therapy was more often inadequate in patients infected with ESBL-producing E. coli (44% vs 15%; P<0.01), and that early mortality (16% vs 6%; P=0.02) and overall mortality (25% vs 11%; P=0.01) were also significantly higher in patients with ESBL-producing E. coli infections. A multivariate model identified the urinary tract focus as the only independent risk factor influencing early mortality for E. coli infections [odds ratio (OR): 0.1; 95% confidence interval (CI): 0.03-0.7; P=0.01]. All 12 patients with ESBL-producing E. coli urinary tract infections treated initially with an oxyimino-beta-lactam survived. Subsequent analysis of the factors influencing early mortality in the cohort of 130 patients with a non-urinary E. coli infection found inadequate empirical antibiotic therapy as an independent risk factor for mortality only for non-urinary E. coli infections (adjusted OR: 3.0; 95% CI: 1.0-8.6; P=0.03). The study showed that hospitalised patients with ESBL-producing E. coli infections more often receive inadequate empiric antibiotic therapy and have a higher mortality rate than those infected with non-ESBL-producing strains. The site of infection strongly influences mortality. The administration of inadequate empirical antibiotic therapy is independently associated with higher mortality only among patients with non-urinary tract infections.