IMPROVING EMPIRIC ANTIBIOTIC SELECTION USING COMPUTER DECISION-SUPPORT

Background: Physicians frequently need to start antibiotic therapy bef ore the results of bacterial cultures and antibiotic susceptibility te sts are available. We developed and evaluated a computerized antibioti c consultant to assist physicians in the selection of appropriate empi ric antibiotics. Methods: We used a two-stage random-selection study t o compare antibiotics suggested by the antibiotic consultant with 482 associated antibiotic susceptibility results and the concurrent antibi otics ordered by physicians. The antibiotics ordered by randomized phy sicians were then compared between crossover periods of antibiotic con sultant use. Results: The antibiotic consultant suggested an antibioti c regimen to which all isolated pathogens were shown to be susceptible for 453 (94%) of 482 culture results, while physicians ordered an ant ibiotic regimen to which all isolated pathogens were susceptible for 3 69 culture results (77%) (P<.001). The physicians who prescribed antib iotics to which all pathogens were susceptible did so a mean of 21 hou rs after the culture specimens were collected. Physicians ordered appr opriate antibiotics within 12 hours of the culture collection signific antly more often when they had use of the antibiotic consultant than d uring the period before use (P<.035). Moreover, 88% of the physicians stated they would recommend the program to other physicians, 85% said the program improved their antibiotic selection, and 81% said they fel t use of the program improved patient care. Conclusions: Information f rom computer-based medical records can be used to help improve physici ans' selection of empiric antibiotics for infections.