At the 2001 ISPOR Annual Meeting we presented a stochastic decision analysis of treatment of pyelonephritis with ciprofloxacin or trimethoprim sulphamethoxazole (TMP/SMX) based on an RCT. OBJECTIVE: Using the same dataset, the purpose of this study is to analyze the value of funding research to reduce the uncertainty around the decision about which drug to prescribe. METHODS: We used a non-parametric approach employing Monte Carlo simulation to determine the expected value of perfect information (EVPI) for the full model and for a particular model parameter (probability of resistance). We incorporated empiric distributions of the net benefits associated with each treatment for patients found to have sensitive and resistant bacteria (4 distributions for each model) created from 1,000 bootstrap replications of the mean cost and effect. The bootstrap replicates involved resampling from original cost and effect data about 210 patients of whom 47 had bacteria resistant to TMP/SMX. In the base case the net benefits were constructed from the mean cost and effect pairs by re-scaling the effects into a monetary value assuming a societal willingness-to-pay $200 to prevent one failure of treatment for pylonephritis. The model for the population assumed an incidence of 0.4% per annum for pyelonephritis in the US and UK and that the decision would be valid for 10 years. RESULTS: Ciprofloxacin is the a priori choice because it is associated with the maximum net benefit for each diagnostic scenario for each diagnostic setting in the model. Nonetheless, there is still residual uncertainty and the EVPI for the decision is $2.18m for the US and £1.65m in the UK. However, the partial evaluation revealed that the EVPI for the prevalence of resistance is only $3,048 in the US and £1,876 in the UK. CONCLUSIONS: Uncertainty about the prevalence of resistance makes a minor contribution to the overall EVPI for the decision.