Uncertainty in Decision Models Analyzing Cost-Effectiveness

Abstract

Purpose. To illustrate the use of a nonparametric bootstrap method in the evaluation of uncertainty in decision models analyzing cost-effectiveness. Methods. The authors reevaluated a previously published cost-effectiveness analysis that used a Markov model comparing initial percutaneous transluminal angioplasty with bypass surgery for femoropopliteal lesions. Each probability in the model was simulated with a first-order Monte Carlo simulation to represent sampling uncertainty. Superimposed on this, a second-order Monte Carlo simulation was performed to represent parameter uncer tainty, drawing the probability values from nonparametric distributions based on published data or from primary collected data as available. After simulation of a mixed (i.e., non-idenfical) cohort of 30,000 patients, 3,000 bootstrap samples of 1,000 pa tients each were drawn and the joint distribution of mean incremental costs and mean effectiveness gained was evaluated. Results. Using a bootstrap sample size of 1,000 patients, 92.7% of the joint distribution of mean incremental costs and mean effective ness gained fell in the quadrant where angioplasty dominated bypass surgery. Another 6.9% of samples demonstrated either greater effectiveness with an incremental cost- effectiveness ratio of at most $20,000/QALY gained, or cost savings with a ratio of at least $20,000 saved/QALY lost. Conclusion. A nonparametric bootstrap method can be used to estimate the joint distribution of mean incremental costs and mean effec tiveness gained, and the results can provide an understanding of the uncertainty in a cost-effectiveness analysis based on a decision model. Key words: cost-effective ness analysis; sensitivity analysis; Monte Carlo simulation; Markov models; bootstrap ping. (Med Decis Making 1998;18:337-346)