A multicenter model for examining in vitro susceptibilities of bacteria to antibiotics

In vitro bacterial susceptibility data are often considered in both hospital formulary and antibiotic prescribing decisions. In this multicenter study, susceptibility data from 105 United States hospitals resulted in a database containing 211,142 isolates for 31 different bacteria and 38 antibiotics. To identify susceptibility patterns of bacteria to antibiotic alternatives, this computerized database was subjected to confidence interval analysis, while controlling for potential sources of random error. This paper describes the multicenter study design and provides several examples of the type of information the analysis can provide to P & Committee members and to practitioners who ultimately must make decisions regarding antibiotic use, sometimes without the benefit of complete hospital-specific historical antibiogram data. Limitations on the interpretation of aggregate in vitro susceptibility data gained from multiple institutions are also discussed.