The automated Cobasbact system was adapted for direct antimicrobial susceptibility testing of positive blood culture broths and its performance compared with that of the conventional Kirby-Bauer agar disc diffusion method using 278 positive blood samples. Overall, 1746 antibiotic-organism combinations were tested. Full agreement was 86.9% and essential agreement (i.e. including minor discrepancies) was 91.8%. The system would seem to produce acceptable susceptibility results within five hours after detection of a positive blood culture broth.

The Cobas-Bact (Roche Diagnostics, Basel, Switzerland) new rotor for the identification (ID) in 4 h 20 min of 33 members of the family Enterobacteriaceae to genus and species level was evaluated by testing 444 strains of which 398 belonged to common species and 46 belonged to rare species of Enterobacteriaceae. Each strain was identified by the API 20E system (Analytab Products, Plainview, N.Y.), and additional discriminating tests were set up if necessary. Only first-choice ID were considered in this study and were classified either as high-confidence ID (normalized likelihood, .80%) or as low-confidence ID (normalized likelihood, <80%) requiring additional tests for confirmation. The data were analyzed by two versions of Cobas-Bact software. With the first version of the software (SW8446), the overall accuracy of Cobas-Bact was 95.5% (424 correct ID of 444). When restricted to high-confidence ID it rose to 99.4% (350 of 352) for the common species and 96.9% (31 of 32) for the rare species. Only three strains of the high-confidence group were misidentified. Sixty ID were considered unacceptable because of their low confidence. Using the first software version (SW8446) they represented 12% (46 of 398) of the common species (17 typical strains, 10 Shigella species, 10 inactive Escherichia coli strains, and 9 rare biotypes) and 30% (14 of 46) of the rare species. The same data analyzed by the new version (SW8524) of the Cobas-Bact software resulted in an overall accuracy of 93.9% (417 correct ID). The number of high-confidence ID rose to 401, of which 392 (97.7%) were accurate. The decrease in low-confidence ID (43 versus 60) was mainly due to the Shigella species. In conclusion the accuracy of Cobas-8act identification system was very good when restricted to high-confidence ID. The Cobas-Bact performance for rare species ID was poorer, but the small number of strains tested does not allow definitive conclusions.

No abstract

A direct antimicrobial susceptibility test and a direct identification of positive blood culture broths for gram-negative rods confirmed with Gram stain by using a new instrument, Cobas-Bact, were compared with the conventional Kirby-Bauer agar diffusion disk method and with the in-house set of identification or API 20E, respectively. The bacterial pellet of centrifuged positive blood culture broth was used to inoculate a Cobas-Bact susceptibility and identification rotor. Bacteria from 206 cases of monomicrobial septicemia due to members of the family Enterobacteriaceae were tested. In 198 episodes (96%), direct identification and antimicrobial susceptibility testing results were obtained for the same bacterial pathogen within 5 h of detection. Of 204 direct identifications obtained, 177 (86.6%) were "high-confidence" correct identifications (percentage of likelihood [P] greater than or equal to 80%) and 25 (12.5%) "low-confidence" correct identifications (P less than 80%), whereas only 2 misidentifications occurred (1 Escherichia coli and 1 Proteus mirabilis). Direct susceptibility testing was performed in 199 episodes (96%), providing 1,885 antibiotic-microorganism combinations. Full agreement reached 86.3%, and essential agreement reached 92.8%. Minor discrepancies were found in 120 (6.5%) of the tests, major discrepancies were found in 127 (6.8%) tests, and very major discrepancies were found in only 7 (0.4%) tests. Subsequent MIC determinations in cases of major or very major discrepancies reduced the number of major discrepancies involving cephalosporins from 60 to 16, whereas all those involving aminoglycosides remained. Overall, this direct and rapid Cobas-Bact identification and susceptibility testing procedure offered accurate information with 5 to 6 h after the laboratory detection of bacteremia and septicemia due to members of the Enterobacteriacease.