Conducting Discrete Choice Experiments to Inform Healthcare Decision Making

Abstract

Discrete choice experiments (DCEs) are regularly used in health economics to elicit preferences for healthcare products and programmes. There is growing recognition that DCEs can provide more than information on preferences and, in particular, they have the potential to contribute more directly to outcome measurement for use in economic evaluation. Almost uniquely, DCEs could potentially contribute to outcome measurement for use in both cost-benefit and cost-utility analysis. Within this expanding remit, our intention is to provide a resource for current practitioners as well as those considering undertaking a DCE, using DCE results in a policy/commercial context, or reviewing a DCE. We present the fundamental principles and theory underlying DCEs. To aid in undertaking and assessing the quality of DCEs, we discuss the process of carrying out a choice study and have developed a checklist covering conceptualizing the choice process, selecting attributes and levels, experimental design, questionnaire design, pilot testing, sampling and sample size, data collection, coding of data, econometric analysis, validity, interpretation and welfare and policy analysis. In this fast-moving area, a number of issues remain on the research frontier. We therefore outline potentially fruitful areas for future research associated both with DCEs in general, and with health applications specifically, paying attention to how the results of DCEs can be used in economic evaluation. We also discuss emerging research trends. We conclude that if appropriately designed, implemented, analysed and interpreted, DCEs offer several advantages in the health sector, the most important of which is that they provide rich data sources for economic evaluation and decision making, allowing investigation of many types of questions, some of which otherwise would be intractable analytically. Thus, they offer viable alternatives and complements to existing methods of valuation and preference elicitation.

Preferences for prenatal tests for Down syndrome: an international comparison of the views of pregnant women and health professionals

“…Each attribute must be measured on a common scale to enable comparison, this can be done in a number of ways including calculating the marginal rates of substitution and predicted probability analysis. 17 The marginal rates of substitution were calculated as a ratio of the coefficients of two attributes to allow direct assessment of how much of one attribute participants were willing to trade-off for more of another attribute. We also determined the predicted probability that two tests from the choice set representative of NIPT and invasive testing would be selected based on the model coefficients.…”

“…We also determined the predicted probability that two tests from the choice set representative of NIPT and invasive testing would be selected based on the model coefficients. 17 The software package Stata 12.0 (StataCorp, College Station, TX, USA) was used to perform all analyses.…”

“…16,17 Nine countries participated: Canada, Denmark, Iceland, Israel, Italy, the Netherlands, Portugal, Singapore, and the United Kingdom (Supplementary Table S1). Local ethics approval was gained by each participating research team (Supplementary Table S2).…”

A discrete choice experiment investigating preferences for funding drugs used to treat orphan diseases: an exploratory study

“…The values of the attributes vary across choice scenarios, and by observing the choices people make it is possible to infer their preferences over the attributes of the goods under study. DCEs are commonly used to identify people's preferences in a variety of non-market situations/services/commodities (Bateman et al 2002;Hensher, Rose, and Greene 2005;Louviere, Hensher, and Swait 2000) and have been widely utilized in health economics (Gerard et al, 2003;King et al, 2007;Ryan et al, 2008;Lancsar and Louviere, 2008).…”

“…Following recommendations on sample size (Hensher et al, 2005;Lancsar and Louviere, 2008), 20 to 50 participants per block was deemed adequate for robust estimation, suggesting a target a sample of 200 individuals. The vast majority of subjects were students.…”