OBJECTIVE: To compare pulsed-field gel electrophoresis (PFGE) typing and O:K-serotyping of Klebsiella in two different epidemiological settings. METHODS: One hundred and four bacteremia isolates without known epidemiological relation and 47 isolates from an outbreak in a neonatal intensive care unit (NICU) were K-typed by countercurrent immunoelectrophoresis (CCIE), O-typed by an inhibition enzyme-linked immunosorbent assay method, and typed by pulsed-field gel electrophoresis (PFGE) using the restriction enzyme XbaI. RESULTS: Typing data for the 104 bacteremia isolates were compared with regard to typability, number of types, maximum number of isolates per type, and the Discriminative Index (DI). O-typing combined with K-typing (DI 0.98) as O:K-serotyping (DI 0.99) gave a very discriminative typing system, whereas O-typing alone was not very discriminative (DI 0.76). PFGE (DI 1.00) was a more discriminative typing method than O:K-serotyping, as it could subdivide 13/22 O:K-serotypes into smaller groups. Isolates with the same PFGE-type had the same O:K-serotype, indicating that isolates with different O- and/or K-types could be expected to be of different PFGE-types. Typing of the 47 isolates from the outbreak in the NICU showed that 38 isolates belonged to a single clone, and that during an epidemic limited in time and space, differences in the electrophoretic patterns of up to five bands between a parental pattern type and a subtype may be found in the PFGE profiles. CONCLUSIONS: Both O:K-serotyping and PFGE typing are highly discriminative typing methods. PFGE is the most discriminative method and is excellent for typing outbreaks with few isolates. If large numbers of isolates are to be typed, a more convenient strategy might be first to K- or O:K-serotype isolates followed by PFGE typing of possible identical isolates. Since K- or O:K-serotyping is a definitive typing method, while PFGE typing is a comparative one, PFGE cannot, for the time being, replace O:K-serotyping for surveillance purposes.