THE USE OF BEST-WORST SCALING CASE 2 TO ESTIMATE INDIVIDUALIZED PREFERENCES FOR VALUE CLARIFICATION AND DECISION SUPPORT

Purpose:

   This study introduces a structured approach for designing value clarification exercises (VCEs) using Best-Worst Scaling case 2 (BWS). When applying BWS in a VCE, one patient is presented with a series of treatment profiles based on the variation in characteristics of the available treatment options. For each treatment profile, the patient has to indicate the most and least attractive characteristic. Based on this data, the relative importance of attributes and the value of treatment options can be estimated to provide the patient with a treatment recommendation.

Method(s):

   The structured approach for designing the VCEs consists of: (1) selection of attributes and levels, (2) construction of the choice task, (3) statistical analysis, and (4) instrument design/presentation of results to patients. Two case studies are used to illustrate the design process and are applications of VCEs in Parkinson’s disease and localized prostate cancer. Each VCE was pilot-tested in 10 patients and evaluated in terms of feasibility, cognitive difficulty, comprehension, and usefulness of presented results.

Result(s):

   Pilot-tests showed patients were willing to spend the extra time and effort to receive additional information about their preferences. Patients also indicated that the VCE made them more aware of the trade-offs between benefits and harms that were needed to choose treatment. However, patients found it difficult to designate a best option from negative defined options and some did not comprehend the task with only the written explanation. Patients experienced also difficulties with recalling the range of possible outcomes on attributes, because each choice task only displays one outcome per attribute.

Conclusion(s):

   This study showed positive results regarding usefulness and feasibility of constructed VCEs. Yet, much effort needs to be put on the written explanation of the task and the familiarization of patients with the varying attribute-levels. Besides, several assumptions have been made in the design of the constructed VCEs to balance clinical and methodological preferences. For instance, the trade-off between design efficiency and respondent efficiency led to the decision to use inefficient and partial profile designs. However, in individual preference estimation it is unknown whether methodological compromises such as these are affecting validity of preferences. Further research should test whether using inefficient designs, partial profile designs, counts analysis or Hierarchical Bayes modelling are affecting the validity of individual preference estimates.