In economic analyses, and especially in cost-effectiveness analyses, the incremental cost-effectiveness ratio is used to measured the economic efficiency of a new treatment, program or intervention, relative to an existing treatment (or control). Costs and effectiveness are seldom known with certainty. There may be incomplete knowledge about parameter values and additional uncertainty may result from sampling variability. Traditionally, sensitivity analysis (the process by which parameter values are varied, and the cost-effectiveness ratio is recalculated) has been used to tackle this uncertainty in parameter values, while more recently, statistical analysts have used confidence interval estimation techniques to handle sampling variability. Existing methods for estimating confidence intervals around the cost-effectiveness ratio do not take into account parameter uncertainty and do not obviate the need for sensitivity analysis. What is needed is a method that simultaneously addresses both forms of uncertainty (parameter uncertainty and sampling variability). Recently, the Bayesian technique has been used to overcome some difficulties of interpretation of confidence intervals for cost-effectiveness ratios. The few studies that use this technique consider directly measured (as opposed to modeled) costs and effectiveness.
We aim to develop a method using the Bayesian method for computing confidence intervals around the cost-effectiveness is modeled; we will apply this Bayesian method to a cost-effectiveness analysis of a randomized trial comparing three HIV prevention interventions for seriously mentally ill men and women. This analysis will allow us to compare the Bayesian method with existing confidence interval estimation techniques, such as the bootstrap and the Fieller methods. This proposed, more inclusive approach to estimating the cost-effectiveness of HIV prevention should produce more accurate estimates for use in policy analysis and resource allocation decision-making. Consequently, decision makers will be able to known how much confidence they can place in estimates of the cost- effectiveness ratio.
| Johnson-Masotti, Ana P; Laud, Purushottam W; Hoffmann, Raymond G et al. (2004) A Bayesian approach to net health benefits: an illustration and application to modeling HIV prevention. Med Decis Making 24:634-53 |
