Background: Existing methods of preference assessment are limited by the ability of certain populations to comprehend the required techniques, such as time tradeoffs and standard gambles. Conjoint analysis (CA) facilitates preference assessment through the presentation of hypothetical scenarios that vary according to attributes of the issue being studied.

Purpose: (1) To develop the tools to conduct a CA experiment among low literacy and minority primary care patients; and (2) to evaluate the effectiveness of CA as a preference assessment technique in these populations using the example of colorectal cancer screening (CRCS).

Methods: This research followed the 5 stages of CA: 1) attribute identification; 2) assignment of levels to attributes; 3) creation of scenarios; 4) preference assessment using scenarios; 5) analysis. Phase I (stages 1-3) consisted of in-depth interviews with 74 patients (25 white, 27 African American, 22 Hispanic) to identify the leading attributes and associated levels relating to CRCS. These data were used to create and pilot-test an initial CA assessment instrument: section (1) rating of individual scenarios comprised of attribute/level combinations on a 1-10 scale; and section (2) ranking the attributes against one another. Phase II (stages 4-5) involves preference assessment using the final instrument among 225 low literacy, minority patients.

Results: Phase I revealed that test accuracy, preparation, frequency, discomfort, and cost were the top 5 attributes relating to CRCS, and that patients wanted very simple descriptions of attribute levels. Sixty-four scenarios were generated, and fractional factorial design used to reduce the number to 14. Pilot testing found that lower literacy patients were not able to comprehend the attribute/level combinations or to complete the rating and raking exercises. The modified CA instrument includes visual representations, a “story-type” scenario format, and a card-sort method for attribute ranking. Re-piloting found lower literacy patients able to perform rating and ranking and tradeoff attribute/level combinations effectively.

Conclusions: CA is a valuable and innovative method for preference elicitation research. With appropriate tailoring, CA can assess preferences among low literacy and minority populations where other preference assessment methods may fail. CA can be used in special populations to inform interventions designed to increase use of preference-based health services such as CRCS.