Feasibility of Using Modified Adaptive Conjoint Analysis Importance Questions

Background: Prior practical experience in conducting adaptive conjoint analysis (ACA) surveys has demonstrated that many subjects have difficulty understanding the 'importance' questions. Objective: To develop a modified version of ACA importance questions. Methods: Modified ACA importance questions composed of two tasks were developed and tested in a pilot study of patients with knee pain. In the first, respondents were presented with the list of attributes and asked to choose which they considered the most important. In the second, they were asked to rate the importance of the remaining attributes relative to the most important one on an 11-point numeric rating scale. Consecutive patients with knee pain, followed at a hospital-based Bone and Joint Clinic, were randomized to complete the original or modified version of the ACA survey. The two versions were identical except for the importance questions. The ACA survey included six attributes: pain, energy, route of administration, stomach upset, bleeding ulcer and cost. Each attribute contained three levels, all of which had a natural order except for route of administration. As this was a pilot study, we present descriptive statistics only. Results: A total of 49 patients were recruited; 24 completed the original version and 25 completed the modified version. Subjects felt that bar graphs illustrating the relative importance were more accurate for the modified version of ACA. The proportion of subjects for which the most important attribute chosen on a card-sorting task matched that generated by ACA was greater for the modified than for the original version (48% vs 29%). The proportion of subjects for which the treatment option chosen on a card-sorting task matched that predicted by ACA was also greater for the modified than for the original version (80% vs 75%). Subjects used a greater number of points to rate the importance of attributes on the modified version of ACA (mean ± SD - 3.4 ± 0.9) than on the original version (mean ± SD - 2.7 ± 1.0). Conclusions: The modified version of the ACA importance questions appears to perform as well as or better than the original version. Use of a simplified set of ACA importance questions is a reasonable alternative for investigators interested in using ACA as a decision-support tool in clinical practice.