Order of Presentation of Dimensions Does Not Systematically Bias Utility Weights from a Discrete Choice Experiment.

Abstract

Discrete choice experiments (DCEs) are increasingly used to value aspects of health. An issue with their adoption is that results may be sensitive to the order in which dimensions of health are presented in the valuation task. Findings in the literature regarding order effects are discordant at present. To quantify the magnitude of order effect of quality-of-life (QOL) dimensions within the context of a DCE designed to produce country-specific value sets for the EORTC Quality of Life Utility Measure-Core 10 dimensions (QLU-C10D), a new utility instrument derived from the widely used cancer-specific QOL questionnaire, the European Organisation for Research and Treatment of Cancer Quality of Life Questionnaire-Core 30. The DCE comprised 960 choice sets, divided into 60 versions of 16 choice sets, with each respondent assigned to a version. Within each version, the order of QLU-C10D QOL dimensions was randomized, followed by life duration in the last position. The DCE was completed online by 2053 individuals in France and Germany. We analyzed the data with a series of conditional logit models, adjusted for repeated choices within respondent. We used F tests to assess order effects, correcting for multiple hypothesis testing. Each F test failed to reject the null hypothesis of no position effect: 1) all QOL order positions considered jointly; 2) last QOL position only; 3) first QOL position only. Furthermore, the order coefficients were small relative to those of the QLU-C10D QOL dimension levels. The order of presentation of QOL dimensions within a DCE designed to provide utility weights for the QLU-C10D had little effect on level coefficients of those QOL dimensions.