Modelling Heterogeneity in Patients' Preferences for the Attributes of a General Practitioner Appointment

Rationale: Data from discrete choice experiments are usually analysed using probit or logit models or random effects extensions of these. These approaches produce estimates of the mean taste weights attributed to the attributes in the experiment by the sampled individuals. Further, if a cost attribute or a proxy for cost is included in the experiment the taste weights can be used to derive estimates of mean willingness to pay for the attributes. It is likely, however, that individuals have different preferences, and that some of the preference heterogeneity is unrelated to observable personal characteristics. This issue cannot be investigated using the traditional modelling tools. Objectives: The objective of this paper is to study the distribution of preferences in a sample of patients who responded to a discrete choice experiment where they were asked to choose between different hypothetical general practitioner appointments. Particular attention is paid to the distribution of willingness to pay for the attributes of the appointment. Methodology: In addition to standard logit models, mixed and latent class logit models are used to analyse the data from the choice experiment. Both mixed and latent class logit models make it possible, given certain assumptions, to estimate the distribution of preferences for the attributes in the experiment. These modelling techniques also take into account the fact that each individual makes several choices which cannot be assumed to be independent. Results: It is found that there is significant preference heterogeneity for all the attributes in the experiment and that both the mixed and latent class logit models lead to significant improvements in fit compared to the standard logit model. Moreover, the distribution of preferences implied by the preferred mixed and latent class models is similar for many attributes. Conclusions: The results underline the importance of accounting for preference heterogeneity when analysing data from discrete choice experiments. While preference heterogeneity is often partially accounted for by interacting attributes with socio-demographic characteristics, the distribution of preferences is likely to also depend on unobserved characteristics of the decision makers. Mixed and latent class logit models provide tools for uncovering the distribution of preferences even when the characteristics driving the preference heterogeneity are unknown