OBJECTIVES: DCE preference research can assist in revealing patient preferences for defined treatments. However, DCEs that incorporate challenging decision contexts can create a cognitive burden for respondents, especially those from vulnerable populations. Limited comprehension of the survey contents can lead to incorrect respondent interpretations and/or unreliable and inaccurate responses. Using a case study specific to mCRPC, we present the survey design process we used to improve respondents’ comprehension and engagement.

METHODS: A cross-sectional survey was conducted using the DCE method. Literature review and physician interviews were used to inform attribute selection, which included: Overall survival, Months until patients develop a fracture/bone metastasis, Percentage of patients who need radiation to control bone pain, Worsening of fatigue, Nausea, and Administration. We conducted pretest interviews (n=10), a pilot study (n=10), and a soft launch (n=30) with US patients recruited as DCE respondents via an online panel. In response to the challenges encountered by some respondents during the pretest/pilot/soft launch, we developed comprehension check questions, warm-up tasks, and a tutorial video to include in the final survey. Choice responses were computed using conditional and mixed logit models.

RESULTS: The final dataset included responses from 160 patients (mean age, 71.6 years old; mean duration since prostate cancer diagnosis, 8.96 years). During the pretest and pilot, some patients struggled with the concepts of Bone pain and Fracture/bone metastasis, which led to a high rate of incorrect answers to comprehension check questions and ultimately counterintuitive signs of the logit model results. The different strategies we deployed for the final survey resulted in a 41% increase in the comprehension rate and positive survey experience reported by the patients.

CONCLUSIONS: Instead of focusing on identifying and excluding poor-quality responses after data collection, researchers may consider utilizing thoughtful patient-centered survey designs to maximize the chances of high-quality preference data input.