objective. To determine whether antimicrobial (AM) courses ordered with an antimicrobial computer decision support system (CDSS) were more likely to be appropriate than courses ordered without the CDSS. design. Retrospective cohort study. Blinded expert reviewers judged whether AM courses were appropriate, considering drug selection, route, dose, and duration. setting. A 279-bed university-affiliated Department of Veterans Affairs (VA) hospital. patients. A 500-patient random sample of inpatients who received a therapeutic AM course between October 2007 and September 2008. intervention. An optional CDSS, available at the point of order entry in the VA computerized patient record system. results. CDSS courses were significantly more likely to be appropriate (111/254, 44%) compared with non-CDSS courses (81/246, 33%, P p.013). Courses were more likely to be appropriate when the initial provider diagnosis of the condition being treated was correct (168/273, 62%) than when it was incorrect, uncertain, or a sign or symptom rather than a disease (24/227, 11%, P!.001). In multivariable analysis, CDSS-ordered courses were more likely to be appropriate than non-CDSS-ordered courses (odds ratio [OR], 1.83; 95% confidence interval [CI], 1.13-2.98). Courses were also more likely to be judged appropriate when the initial provider diagnosis of the condition being treated was correct than when it was incorrect, uncertain, or a sign or symptom rather than a disease (OR, 3.56; 95% CI, 1.4-9.0). conclusions. Use of the CDSS was associated with more appropriate AM use. To achieve greater improvements, strategies are needed to improve provider diagnoses of syndromes that are infectious or possibly infectious.