This paper focuses on the elicitation of patients' preferences for cardiac rehabilitation activities from a discrete choice experiment using a mixed model. We observed a high level of preference heterogeneity among patients for all the five cardiac rehabilitation activities--even when age and smoking status were taken into account. The random parameter model provided additional policy relevant information as well as a better fit to the data than did the standard logit model. The paper focuses on one of the potential problems with the standard logit specification which in the worst case can lead to wrong policy conclusions by assuming homogeneity in preferences across individuals. The generalised RPL specification may be a more appropriate specification that can provide additional information on the heterogeneity preferences.