OBJECTIVE : Carbapenem-resistant Enterobacteriaceae (CRE) are increasingly widespread in the healthcare system, resulting in infections associated with mortality of up to 50%. Many laboratories use automated systems to identify CRE isolates and determine susceptibility. The aim of this study was to evaluate categorical agreement between the BD Phoenix automated system and the gold standard - broth microdilution - in determining minimum inhibitory concentrations of CRE. METHODS : The activity of amikacin, aztreonam, cefepime, ceftazidime, ertapenem, gentamicin, levofloxacin, meropenem, nitrofurantoin, piperacillin-tazobactam and tobramycin on 125 CRE isolates collected from an academic medical centre was evaluated. Categorical agreement between BD Phoenix and broth microdilution was determined, as well as minor error rates, major error rates and very major error rates. RESULTS : BD Phoenix significantly overestimates susceptibility of CRE isolates to amikacin, aztreonam, cefepime, ceftazidime, gentamicin, levofloxacin, meropenem, nitrofurantoin and tobramycin compared with broth microdilution. Overall, categorical agreement of 76% between testing methods indicates the potential diminished ability of BD Phoenix to predict resistance accurately in highly drug-resistant isolates. All tested antimicrobials had higher major error rates compared with previous literature. CONCLUSIONS : BD Phoenix has diminished ability to determine susceptibility of CRE isolates. Further studies are warranted in order to validate BD Phoenix susceptibility testing in highly resistant CRE isolates. The mechanism by which isolates are resistant to carbapenems does not impact the ability of BD Phoenix to determine susceptibility.