An acquired taste: latent class analysis to compare adolescent and adult preferences for EQ-5D-Y-3L health states.

Abstract

Objectives: United States stakeholders advised including adolescents in the valuation study for the EQ-5D-Y-3L, a step towards greater acknowledgement of children in informing societal values. This study aimed to assess the relative contribution of adolescent and adult preferences to a model when combined.

Methods: Discrete choice experiment (DCE) data were collected from an online sample of 1,529 adults and 714 adolescents (ages 11-17). Each respondent completed 15 DCE tasks which were analyzed using latent class models representing varying number of preference classes. Within the best-fitting model, the contribution of each class was determined by the 'scale-adjusted class share' (SACS), combining the class's proportion of respondents (class share) and the magnitude of coefficients (within-class scale). We estimated the contribution of adolescent and adult respondents to SACS for each class, with lower SACS representing less contribution to the combined model.

Results: The best fitting model described 6 classes. Adults had higher contribution to all except one class, accounting for 78.7% of the total contribution. After adjusting for the unequal sample size of adolescent and adult respondents, adults contributed approximately 65.0% and adolescents contributed 35.0% of the weights towards a combined model.

Conclusions: Adolescents showed diminished, disproportionate representation in a combined model, due in part to more indifferent, less informative preferences for EQ-5D-Y-3L health states compared to adults. Latent class analysis showcases one approach to estimate and weight contributions from intentionally sampled subgroups in a combined model.