Objectives: TREAT, a decision support system for antimicrobial therapy, was implemented in an acute medical ward. Methods: Patients admitted on suspicion of infection were included in the study. The evaluation of TREAT was done both retrospectively and prospectively. Coverage of empirical antimicrobial treatments was compared to recommendations from TREAT and the optimal use of local guidelines. Results: Five hundred and eleven patients were included, of whom 162 had a microbiologically documented infection. In the retrospective part of the study, TREAT, physician, and guideline antimicrobial coverage rates were 65%, 51%, and 79%, respectively, and in the prospective part, 68%, 62%, and 77%, respectively. TREAT provided lower coverage than local guidelines (p < 0.001), but was similar to the performance of physicians in a university hospital (p = 0.069). No differences were found in length of hospital stay, or hospital or 30-day mortality. Direct costs were significantly higher for TREAT advice than for local guidelines or the physician prescriptions (p < 0.001), but the ecological costs were lower for TREAT advice than for both local guidelines (p < 0.001) and physician prescriptions (p = 0.247). The coverage of TREAT advice for the bacteraemia patients was non-inferior to the physicians (p = 1.00). Conclusions: TREAT can potentially improve the ecological costs of empirical antimicrobial therapy for patients in acute medical wards, but provided lower coverage than local guidelines.