Accounting for unobservable preference heterogeneity and evaluating alternative anchoring approaches to estimate country-specific EQ-5D-Y value sets: a case study using Spanish preference data

Objectives: The EuroQol Group published the EQ-5D-Y valuation protocol. This protocol suggested the use of two valuation techniques to elicit preferences: Composite Time-Trade-Off (C-TTO) and Discrete Choice Experiments (DCE). The protocol left the decision of what modelling approach to use open for researchers. Our aims were to explore modelling strategies allowing generation of EQ-5D-Y value sets and to produce an EQ-5D-Y Spanish value set. Methods: We used EQ-5D-Y DCE and C-TTO data collected in Spain following the protocol and adopted a staged approach for our modelling exercise. First, we selected the best performing DCE latent class model. We tested several of models, including 2-10 classes. We selected the preferred model based on best goodness of fit in terms of the Bayesian information criterion (BIC). We considered 2 anchoring approaches to estimate utility values: 1) pits state anchoring and 2) hybrid models (using all available C-TTO responses). All analysis were weighted to be representative of the Spanish population. Results: We collected 1,005 DCE and 200 C-TTO interviews. We selected a DCE model including 4 classes. Hybrid models using all available C-TTO observations produced a narrower range of values than the pits state anchoring approach. Conclusions: In this manuscript we have presented an EQ-5D-Y value set that can be used for cost-utility analysis in Spain. The international EQ-5D-Y valuation protocol should be updated to include a different set of health states for the C-TTO experiment if researchers wish to use alternative anchoring approaches to the “pits state”.