Interpretation of Antimicrobial Susceptibility Test of Enterobacteriaceae to beta-lactams with Expert System

Lee, Chae Hoon; Cho, Hee Soon

Abstract

BACKGROUND
At antimicrobial susceptibility testing (AST), clinical isolates may appear susceptible sometimes to the antibiotics that are clinically ineffective or due to technical errors in the testing. So an interpretive reading of AST should be done, but most hospitals do not perform it routinely. Here, we developed and evaluated a computerized expert system to interpret AST of Enterobacteriaceae to beta-lactam antibiotics. METHODS: We made a rule-based expert system according to the natural resistance of the members of Enterobacteriaceae and the common phenotypes of resistance mechanisms for Enterobacteriacae. Antimicrobial suceptibility testings were performed using the disk diffusion method with 12 beta-lactam antibiotics for a total of 1, 016 clinical isolates. Then we compared the raw and expert results of AST. RESULTS: An overall discrepancy rate due to natural resistance was 5.9%; 10.4% for Klebsiella spp. and Citrobacter diversus, 15.0% for Enterobacter spp., 2.6% for Serratia marcescens, 31.6% for Morganella morganii and Providencia stuartti. Accoriding to acquired antimicrobial resistant mechanisms, overall resistant discrepancy was 21.8%; 18.8% for Escherichia coli, Proteus mirabilis, Salmonella and Shigella spp., 25.9% for Citrobacter diversus and Klebsiella spp., 21.6% for Citrobacter freundii and Enterobacter spp., 45.6% for Morganella morganii, 10.0% for Proteus vulgaris, 12.2% for Serratia spp.. CONCLUSIONS: These results indicate that the application of the expert system for interpretation of antimicrobial susceptibility test may provide more reliable data for the treating physician. Additional information should be applied on the software for new resistant mechanisms or some misinterpretive readings.