Computer-assisted bacterial identification utilizing antimicrobial susceptibility profiles generated by autobac 1

Abstract
A computer program was developed to identify bacteria solely on the basis of their relative susceptibility to various antimicrobial agents. A sample of 481 clinical isolates from 9 of the most commonly isolated gram-negative groups was identified by the quadratic discriminant function technique. Various combinations of antimicrobials were tried, and 1 set of 18 resulted in a > 97% correlation with conventional identification procedures. The antimicrobial set could be decreased to 14, while a > 95% correlation with the conventional procedures was maintained.