Developing attributes and levels for discrete choice experiments using qualitative methods

Objectives: The rigour with which the first two stages of discrete choice experiments (attribute development and the choice of levels of these attributes) are generally conducted is questionable. This paper provides a case study describing how attributes and their levels were developed for a study of access to dermatology specialist services for non-urgent skin conditions. Methods: Semi-structured interviews were conducted with 19 dermatology patients with non-urgent skin conditions. Informants were purposively sampled for maximum variation and interviews continued until all attributes were fully and clearly defined. An iterative approach was used with data collection and analysis proceeding concurrently. Results: The interviews and parallel analysis generated three iterations. The first iteration comprised early exploratory work with expertise and waiting time emerging as important to informants. The second iteration continued to emphasize these attributes, but individualized care and convenience were added. By the end of the third iteration all attributes were fully elaborated. Conclusions: Qualitative methods enabled attributes to be defined. There was clear tension between the aim in qualitative work to explore and describe, and the reductiveness needed to encapsulate the different aspects of the service within a minimum number of attributes for use in the discrete choice modelling. Improved reporting of the methods of attribute development in all discrete choice experiments is required. © The Royal Society of Medicine Press Ltd 2007