Between June and November 2004, a vancomycin-resistant Enterococcus faecium (VRE) strain was isolated from 13 patients in the haematology/bone marrow transplant unit. There were difficulties in identifying the organism, which had low-level, inducible vancomycin resistance, and standard screening methods did not reveal carriage in patients or their contacts. These technical failures led to spread of VRE and delays in providing appropriate management, which might otherwise have been avoided. Therefore, we reviewed our laboratory methods and compared three identification systems to determine which would best identify this VRE strain. The VITEK 2 (BioMerieux) correctly identified, as E. faecium, only two of 16 isolates, whereas API Rapid ID 32 Strep (BioMerieux) and Phoenix 100 (Becton Dickinson and Co.) correctly identified 13 of 15 and 12 of 13 isolates tested, respectively. Isolates from urine, tested by the CLSI disk diffusion method, were apparently susceptible or of intermediate susceptibility to vancomycin, upon primary testing. VITEK 2 and Phoenix 100 identified all isolates as vancomycin-resistant, although the MICs, measured by Etest, were in the susceptible range for three of 16 isolates. Reducing the vancomycin concentration in screening media substantially increased the sensitivity for detection of VRE. Isolates were characterized as genotype vanB2/3 by PCR and were indistinguishable from each other by pulsed-field gel electrophoresis. VRE with low-level inducible resistance can be missed by routine screening methods. Better identification and screening methods for detection of low-level vancomycin resistance are needed to improve surveillance and prevent transmission of VRE.