A comparison of methods for converting DCE values onto the full health-dead QALY scale

J Brazier, BV Hout, D Rowen

Abstract

Cardinal preference elicitation techniques such as time trade-off (TTO) and Standard Gamble (SG) receive criticism for their complexity and difficulties in using them in more vulnerable populations. Ordinal techniques such as discrete choice experiment (DCE) and Best Worst Scaling (BWS) are easier, but values generated by them are not anchored onto the full health-dead 1-0 QALY scale required for use in economic evaluation. This paper explores new methods for converting modelled DCE latent values onto the full health-dead QALY scale: (1) anchoring assuming worst state is equal to being dead; (2) anchoring DCE values using dead as valued in the DCE; (3) anchoring DCE values using TTO value for worst state; (4) mapping DCE values onto TTO; (5) combining DCE and TTO data in a hybrid model. We use postal DCE data (n=263) and TTO data (n=307) collected by interview in a general population valuation study of an asthma condition-specific measure (AQL-5D). Methods (4) and (5) using mapping and hybrid models perform best; the anchor-based methods perform relatively poorly. These new methods have a useful role for producing values on the QALY scale from ordinal techniques such as DCE and BWS for use in cost utility analyses.