Using discrete choice experiments to value the benefits of health care
The aim of this thesis is to broaden work in the area of discrete choice experiments (DCEs) in health economics by focusing on the development of some key areas which have to date received relatively little attention. By firstly outlining the background, theory and context of DCEs in health economics, areas are identified that deserve further exploration due to their particular relevance to health economics. Specific contributions of the thesis are in three main areas: modelling the participation decision in DCEs in health care; the use of strength of preference choice modelling approaches; and the applicability of a new approach, best attribute scaling (BSc), to health economics. The contributions of the thesis are to the design, analysis and interpretation of DCE studies. The seven key recommendations for the conduct of DCE surveys arising from this thesis have been made on the basis of the best available knowledge gleaned from empirical analysis carried out. They are as follows: (1) Researchers conducting DCE studies in health care should carefully consider the participation decision when designing and analysing a DCE and where appropriate pay particular attention to describing the opt-out or status quo option as realistically as possible (2) When incorporating status quo or opt-out scenarios within a DCE design researchers should check that the statistical properties of the resulting design are still valid (3) Strength of preference models appear to improve the statistical efficiency of models and produce more accurate estimates of welfare however this may come at the expense of the predictive ability of the model and consistency rates (4) Where possible DCE surveys should provide respondents with the most realistic range of preference ‘capture’ mechanisms as possible, this includes providing opting out, status quo and indifference options where appropriate. (5) As shown in the environmental economics literature strength of preference models facilitates the elicitation of preferences in health economics (6) BSc methods can be used to predict choices in health care preference surveys (7) BSc has the ability to separate scale and weight in choice experiments and hence place all attribute levels on a common scale, unlike traditional choice experiments however future work with larger samples should clarify this result.