As part of our integrated hospital information system (the HELP system), we developed computer-assisted decision support programs for antimicrobial prescribing that are available at bedside terminals throughout our 520-bed community hospital. Recently, options have been added to allow direct physician order entry of anti-infective agents in the shock-trauma intensive care unit (STRICU). Physicians prescribed the computer-suggested regimens for 46% but followed the suggested dose and interval for 93% of the orders during a 1-year study period. In comparison to a 2-year pre-intervention period, improved drug selection and reductions in adverse drug events and costs were seen. Antimicrobial resistance patterns for nosocomial gram-negative isolates remained stable or improved in the STRICU over an 11-year period of computer-assisted antibiotic management. We conclude that strategies for optimizing antimicrobial prescribing have the potential to stabilize resistance and reduce costs by encouraging heterogeneous prescribing patterns, use of local antimicrobial susceptibility patterns to inform empiric drug selection, and reduced "tonnage" of antibiotic use.