Empirical treatment of bacteraemic urinary tract infection. Evaluation of a decision support system

In a Danish county with a low prevalence of antibiotic resistance to most antibiotics, we have constructed and evaluated a decision support system (DSS) for guidance of empirical antibiotic therapy in patients with bacteraemia originating from the urinary tract.

The DSS was based upon a causal probabilistic network, and a decision theoretic approach was used to balance the costs of antibiotic therapy against the therapeutic benefit. The costs included direct cost of purchasing antibiotics, side effects, and the risk of development of antibiotic resistance. The therapeutic benefit was defined as the increase in life-expectancy caused by antibiotic therapy. Life-years were chosen as the common currency unit. Four hundred and ninety-one bacteraemias seen during 1992-1994 were used to construct the DSS (derivation set), and 426 bacteraemias during 1995-1996 were used for evaluation (validation set). The cases were identified in a regional register of bacteraemias. The study was non-interventional.

The DSS suggested antibiotics which would provide coverage in 377 of the 426 episodes (88.5%) compared to 259 episodes (60.8%) for which empirical therapy actually provided coverage (p < 0.01, McNemar-test). The regimens suggested by the DSS included mecillinam as monotherapy in 240 episodes (56.3%), gentamicin as monotherapy in 81 (19.0%), and a combination of gentamicin and ampicillin in 80 (18.8%).

A decision theoretic approach shows promise of improving empirical antibiotic treatment, and may be a measure to support an antibiotic policy.