Bayesian regression models for cost-effectiveness analysis.

Abstract

Recent studies have shown how cost-effectiveness analysis can be undertaken in a regression framework. This contribution explores the use of practical regression models for estimating cost-effectiveness from a Bayesian perspective. Two different Bayesian models are described. The first considers the outcome measure to be a quantitative variable. In the second model the individual outcome measure is a binary variable with value 1 if any objective has been achieved. We describe the implementation of the model using data from a trial that compares two highly active antiretroviral therapies in HIV asymptomatic patients. Data on direct cost and data effectiveness (percentage of patients with undetectable viral load and quality of life) were recorded. If we consider the quality of life as an effectiveness measure, the new treatment is preferred for a willingness to pay more than Euro 142.3 for an increase in the quality of life. For illustrative purposes, if we compare the results with an analogous model that does not include covariates, the critical value becomes Euro 247.4. For the binary measure of effectiveness the control treatment dominates the new treatment.