Where Next for Discrete Choice Health Valuation Exercises?

This paper explores five under-recognised conceptual issues in the use of discrete choice tasks in eliciting a tariff for health and/or quality of life instruments. Its purpose is not to explore econometric and statistical techniques but to offer advice on design and practical matters gleaned from a series of valuation exercises the author has been intimately involved with. These were largely Best-Worst Scaling (BWS) studies, but the implications hold true in many cases for traditional discrete choice experiments (DCEs) and ranking studies. The first two issues deal largely with the construction of the descriptive system (items). The third issue uses a within-subject DCE led by the author to introduce the worrying possibility that the quest for high statistical efficiency in discrete choice studies may have severe, unintended consequences. The need for quota, rather than representative, sampling is discussed as the fourth issue. Finally there is a discussion of the often used Case 2 type of BWS and how it relates to traditional “multiprofile” (choice between states) DCEs and Case 3 BWS. It hopes to correct misconceptions that continue to be propagated concerning the differences between the methods. The author offers advice on future studies and is encouraged by work funded by the EuroQoL Group