OBJECTIVES:

The SF-6Dv2 was developed to improve wording, floor effects and scoring of its widely used predecessor. Value sets for use in the estimation of quality adjusted life years (QALYs) exist for the UK, China and Australia. This study presents a new value set for the USA.

METHODS:

Valuation was based on a discrete choice experiment including a duration attribute. A core set of 304 pairwise choice tasks was generated by a full factorial design excluding very rare combinations. Core tasks were combined with 2 additional sets of 76 tasks each: 1) pairs of commonly experienced health states; 2) pairs of severe health states with a third option of immediate death (ternary). Training and consistency tasks were included, for a total of 15 tasks/participant. Data (n=3.807) from the USA general population were analyzed using the conditional logit model, with adaptations for monotonicity, inclusion of a parameter for worst level on any dimension (WORST) and interactions for 2-way dimension-duration. Impact of interactions was evaluated as the difference in mean utility in a disease group versus overall mean, using an independent database with 31 diseases.

RESULTS:

The recommended value set was based on all task types (core+common+ternary), had minor non-monotonicity, did not include the WORST parameter (not statistically significant), nor dimension interactions, and resulted in a range 1 to -0.527. Pain, mental health and physical functioning had the largest decrements (-0.611 to -0.229 for worst level), and role functioning the lowest (-0.071 for worst level).

CONCLUSIONS: The study presents the SF-6Dv2 USA value set. Despite having the same dimension impact order, the resulting value set differs from the UK set in various ways: more sensitivity to low levels of pain and mental distress, the exclusion of an interaction term, and less sensitivity to impact on role and physical functioning.