His study was performed to compare the methods of detection and to estimate the prevalence of extendedspectrum β-lactamases (ESBL) among Klebsiella spp and E. coli in a university hospital in southern Brazil. We also used a molecular typing method to evaluate the genetic correlation between isolates of ESBL K. pneumoniae. Production of ESBL was investigated in 95 clinical isolates of Klebsiella spp and Escherichia coli from Hospital de Clínicas de Porto Alegre, using Kirby-Bauer zone diameter (KB), double-disk diffusion (DD), breakpoint for ceftazidime (MIC CAZ), increased zone diameter with clavulanate (CAZ/CAC) and ratio of ceftazidime MIC/ceftazidime-clavulanate MIC (MIC CAZ/CAC). Molecular typing was performed by DNA macrorestriction analysis followed by pulsed-field gel electrophoresis. The KB method displayed the highest rates of ESBL (up to 70% of Klebsiella and 59% of E. coli), contrasting with all the other methods (p < 0.05). The confirmatory methods (DD, MIC CAZ, CAZ/CAC and MIC CAZ/CAC) showed a range of ESBL production from 8 to 13% for E. coli and from 33 to 40% for Klebsiella species. Therefore, the KB method was useful only as a screening method as it provided several false positive results. Molecular typing of 17 ESBL K. pneumoniae indicated that the isolates had no clonal relation. We found a good correlation among the confirmatory methods for ESBL detection although the methods which evaluate inhibition of the β-lactamase by clavulanate appeared to be more specific. The high prevalence of ESBL Klebsiella in our hospital is probably due to individual selection of resistant strains rather than the transmission of a common strain.