Incomplete information and irrelevant attributes in stated‐preference values for health interventions

Violations of the assumptions of complete information [CI] and independence of irrelevant alternatives (IIA) in discrete‐choice experiment (DCE) data imply sensitivity of preference estimates to the decision context and the alternatives evaluated. There is a paucity of evidence on how these two assumptions affect health‐preference results and whether the usual specifications of random‐parameters logit models are sufficient to address these violations. We assessed the appropriateness of these assumptions in a DCE valuating interventions to prevent long‐term health problems that could be identified through whole genome sequencing. A DCE survey was administered to members of a nationally representative consumer panel to elicit their preferences for options to reduce the risk of health problems. The treatment options presented (surgery, medication, and watchful waiting) and the context for the decisions elicited (severity and likelihood of the health problem) were varied experimentally to evaluate the sensitivity of preference results to such changes. We find evidence of IIA violations as the options presented to prevent health changed. Our results also are consistent with the expectation that additional substitutes decrease the monetized value of alternatives. We also find some evidence that the decision context can moderate such effects, which constitutes a new finding.