A NOVEL ACCOUNT OF DEPENDENCIES IN HEALTH UTILITY ASSESSMENT

Extensive work in health utility assessment has revealed robust and systematic dependencies on the elicitation method (e.g., choice,
matching, ranking). Based on traditional analyses, this phenomenon requires task-dependent changes in the utility function and is problematic for deriving an accurate preference ordering over health states. We present a novel approach that predicts the response-dependent differences while still retaining a single underlying utility representation. It extends a computational process model successful in accounting for choice-pricing preference reversals to the medical domain. We introduce the model and apply it to multiple data sets, illustrating both its qualitative and quantitative success.