Modeling methods and the degree of parameter uncertainty in probabilistic analyses of economic evaluations

Abstract

There is a framework providing some general guidance to interpret the levels of parameter uncertainty of the probabilistic analysis in economic evaluations. Given that this framework may not fully address underlying causes for the uncertainty, we sought to extend it for two specific scenarios. We provided the mathematical interpretations and conducted simulation studies for two scenarios. The first study examined a case where the intervention and control strategies were associated with different health states (e.g., active surveillance versus surgery for treatment of non-invasive cancer). The second study evaluated the quality-adjusted life-years (QALYs), estimated from reported summary statistics (i.e., mean and standard deviation) of longitudinal post-treatment utility data from a clinical trial. The first simulation study showed that the magnitude of uncertainty of cost-effectiveness results was much greater if a decision model considered different health states for the intervention and control strategies than if the model considered the same health states. The second study showed that variance in the estimates of QALYs and incremental QALYs using the summary statistics was substantially underestimated when the correlations of repeated measures which are generally not available in the literature were omitted. We further discussed the implications of our findings for the economic modeling. In addition to qualitative categorization of uncertainty proposed by the general framework to assist with decision-making, we also need to comprehend methods used to address uncertainty in economic evaluations to enable informed policy making.