Development of an automated antibiotic consultant

Abstract

Physicians frequently need to order antibiotic therapy before the results of bacterial cultures and antibiotic susceptibility tests are available. Therapy selected empirically should be chosen on the basis of the most recent information about probable pathogens and their susceptibility to various drugs. We have developed a series of computer programs to help physicians identify the antibiotic regimens with the highest probability of acting against the pathogens that are most likely to be present at suspected sites of infection in individual patients. This automated antibiotic consultant is now available on our hospital information system. A comparison of the antibiotic regimens suggested by the computer with those indicated by cultures and susceptibility tests showed that the computer suggested an appropriate regimen for 717 of 761 culture events (94%).