Sunday, October 18, 2009
Grand Ballroom, Salons 1 & 2 (Renaissance Hollywood Hotel)
Janine A. van Til, MSc, Universiteit Twente, Enschede, Netherlands, James G. Dolan, MD, University of Rochester, Rochester, NY, Anne M. Stiggelbout, PhD, Leiden University Medical Center, Leiden, Netherlands and Maarten J. IJzerman, PhD, University of Twente, Enschede, Netherlands

Purpose: Conjoint analysis has methodological advantages compared to other multi-criteria decision analyses techniques, like the Analytic Hierarchy Process, because of it’s foundation in utility theory and the more realistic approach to imitating consumer decisions. However, an empirical comparison has suggested superiority of AHP to CA in complex decisions. In a previous study we compared conjoint analysis and AHP by analyzing the preference for 2 treatment options in high level spinal cord injury.  In the present study, we are interested to see how each of these methods behaves in decisions involving more than 2 treatments.

Methods: The present study was carried out in 142 patients with a mean age of 61 years. Of all patients, about 50% suffered from a drop-foot due to stroke and about 20% has had a HNP. All patients suffer from ankle-foot impairments and used walking aids (e.g. ankle-foot orthoses). The attributes in the conjoint analysis study were based on a decision tree that was created previously in collaboration with an interest group of physiatrists. Subsequently, patients were given a survey asking for AHP weights of the main attributes, determined by an expert panel.

Results: The first preliminary findings were that the AHP and CA studies gave different ordering of the attributes. According to AHP, “treatment results”, “risks” and “comfort of additional devices” were the most important attributes. The CA study, however, showed that the “impact of treatment“ (i.e. requiring surgery or not) was most important.

Conclusions: The relative difference between attribute weights was much smaller using CA compared to AHP. The rank reversals and the smaller difference in weights between first and last attribute, are consistent with the findings in a previous study where we compared four multi-attribute weighting techniques. This results in differences in predicted preferences for treatment.

Candidate for the Lee B. Lusted Student Prize Competition