PURPOSE: To provide “guidebook” assistance for practitioners conducting health technology assessments (HTAs) using the Generalized Risk-Adjusted Cost-Effectiveness (GRACE) method for valuing medical interventions.

DESCRIPTION:

GRACE generalizes traditional Cost-Effectiveness Analysis (CEA) by introducing decreasing returns to health-related quality of life (QoL). Three important changes emerge for the proper conduct of HTAs: (1) In GRACE (compared with current CEA practice), WTP is lower for low-severity illnesses and rises exponentially with illness severity. In stark contrast, conventional CEA assumes that WTP does not vary with disease severity. Similarly, GRACE shows why WTP for QoL improvements is greater for disabled- than for otherwise-similar non-disabled persons, an issue creating major concern regarding current CEA methods, which impute lower value to QoL improvement for disabled persons. (2) Mean improvements in QoL must be adjusted to account for uncertainty in health outcomes; lower uncertainty in QoL outcomes increases value, independent of mean improvements. (3) Willingness to trade life expectancy (LE) for QoL improvements varies with untreated illness severity. We will review GRACE-related methods to conduct all of these analyses, combining information on population risk preferences and information regarding treatment-specific distributions of health outcomes.

GRACE requires new “done-once” estimates of population preferences regarding how health creates utility and uncertainty in health outcomes. We will discuss how to estimate relevant parameters using “economics of happiness” methods, which provide all necessary parameters for GRACE analyses.

GRACE requires more-detailed evidence on distributions of health outcomes than does traditional CEA, all collected at illness/treatment levels, including variance and skewness of outcome distributions in both untreated and treated states. We will show how to obtain these data from current HTA methods (Randomized Controlled Trials, Comparative Effectiveness studies, or Monte-Carlo simulations). We will conclude by discussing how to combine the population-level risk-preference data with illness/treatment specific data to complete GRACE evaluations.