Parallel comparison of accuracy of API 20E, Vitek GNI, MicroScan Walk/Away Rapid ID, and Becton Dickinson Cobas Micro ID-E/NF for identification of members of the family Enterobacteriaceae and common gram-negative, non-glucose-fermenting bacilli.

We compared the API 20E (21 h) (API; bioMérieux Vitek, Hazelwood, Mo.), the Vitek GNI card (4 to 18 h) (Vitek; bioMérieux Vitek), the identification portion of the MicroScan Walk/Away Rapid Neg Combo 3 panel (2 h) (W/A; Baxter Diagnostics, Inc., West Sacramento, Calif.), and the Becton Dickinson Cobas Micro ID-E/NF rotor (21 h) (Cobas; Becton Dickinson Diagnostic Instrument Systems, Sparks, Md.), versus conventional biochemicals for their abilities to identify accurately 252 strains of biochemically typical and atypical members of the family Enterobacteriaceae and common non-glucose-fermenting gram-negative bacilli. All strains used were included in the data base of each product. At the end of the initial incubation, 194 (77.0%), 213 (84.5%), 198 (78.6%), and 192 (76.2%) strains were correct to the genus and species levels with the API, Vitek, W/A, and Cobas systems, respectively. After additional biochemical tests were performed, as directed by each manufacturer's protocol, the numbers of strains correctly identified to the genus and species levels were 241 (95.6%), 234 (92.8%), 243 (96.4%), and 230 (91.3%) with the four systems, respectively. The errors were random in all systems, with the exception of two atypical Salmonella enteritidis strains, each of which was misidentified by three systems. After the initial recommended incubation period, both API and Cobas were significantly less accurate than Vitek (Yates' corrected P < 0.05). No significant differences were noted between the results of Vitek and W/A or between the results of API and W/A. After additional tests were completed, Cobas was significantly less accurate than W/A (P < 0.05) but was equal in accuracy to Vitek and API. API, Vitek, and W/A were equal in accuracy after these same additional tests. All four systems were significantly more accurate after additional biochemical testing than after the initial reporting period (194 of 252 versus 241 of 252 for API, 213 of 252 versus 234 of 252 for Vitek, 198 of 252 versus 243 or 252 for W/A, and 192 of 252 versus 230 of 252 for Cobas [P<0.05]).