Empirical antibiotic treatment with broad-spectrum antibiotics provides a high probability of covering treatment, but is associated with unnecessary costs as high drug prices, side-effects, and facilitated development of antibiotic resistance. A decision support system (DSS) based upon a causal probabilistic network (CPN) was constructed from a database with 491 cases (1992–94) of urosepticaemia and validated on 426 cases (1995–96). The CPN uses decision theory to balance the gain in life-years due to therapy against the costs of the antibiotic therapy, i.e. price, side-effects and ecological cost. The DSS selected antibiotics of an overall lower price, higher coverage and less ecological cost than the antibiotics actually chosen for empirical treatment. Thus, a DSS incorporating the CPN could achieve a desirable antibiotic policy, and it holds promise for improving empirical antibiotic therapy.