Published online by Cambridge University Press: 05 October 2015
This chapter uses Case 2 (the profile case) best-worst scaling to elicit utilities associated with attributes and levels that might result from the extraction of third molar teeth or conservative management. First reported by McIntosh (2003) in her PhD, it was the first Case 2 study conducted in health in Europe and only the second globally (after that of Szeinbach et al., 1999). We use it in this chapter as an empirical illustration of the simpler methods of analysis introduced in Chapter 3; accordingly, we focus on the summary (sample-level) statistics, including the weighted least squares of the natural logs of the choice frequencies. Chapter 12 develops the approach further in the context of the ICECAP-O instrument (Coast, Peters et al., 2008a).The empirical study reported in this chapter was part of a larger study to assess the cost-effectiveness of alternative implementation strategies for third molar clinical guidelines. For the purposes of this chapter only, the methodological components of the benefit assessment of the economic evaluation are reported as per the original research (McIntosh, 2003; McIntosh and Louviere, 2002); for a more detailed guide on the use of discrete choice methods in health care economic evaluation, see McIntosh et al. (2010). The aim of the methodological work was to use Case 2 BWS to identify attributes and levels of third molar management (extraction and conservative management) so that preferences could be elicited and attributes of importance identified and valued.
The study was conducted as part of the first author's PhD in 2003. A number of methodological issues and analyses were reported when the data were presented at the Health Economists' Study Group conference (McIntosh and Louviere, 2002). Some of the proposed data analyses, such as those involving the decomposition of attribute weight and scale, have developed since those early days (see Marley, Flynn and Louviere, 2008). However, the study remains innovative for multiple reasons, including its comparison with conventional discrete choice experiment estimates. These were relatively simple comparisons, reflecting the state of the art at the time; however, such comparisons remain a key issue on the research frontier, and the understanding gained in this chapter is valuable for the reader in comprehending more recent technical papers.