Eliciting preferences for cancer screening tests: Comparison of a discrete choice experiment and the threshold technique

Abstract

Objective: To compare results of three preference elicitation methods for a cancer screening test.

Methods: Participants (undergraduate students) completed a discrete choice experiment (DCE) and a threshold technique (TT) task. Accuracy (false positives, false negatives), benefits (lives saved), and cost for a cancer screening test were used as attributes in the DCE and branching logic for the TT. Participants were also asked a direct elicitation question regarding a hypothetical screening test for breast (women) or prostate (men) cancer without mortality benefit. Correlations assessed the relationship between DCE and TT thresholds. Thresholds were standardized and ranked for both methods to compare. A logistic regression used the thresholds to predict results of the direct elicitation.

Results: DCE and TT estimates were not meaningfully correlated (max ρ = 0.17). Participant rankings of attributes matched only 20% of the time (58/292). Neither method predicted preference for being screened (ps > 0.21).

Conclusions: The DCE and TT yielded different preference estimates (and rank orderings) for the same participant. Neither method predicted patients' desires for a screening test.

Practice implications: Clinicians, patients, policy makers, and researchers should be aware that patient preference results may be sensitive to the method of eliciting preferences.

Keywords: Discrete choice experiment; Patient preference; Preference elicitation; Shared decision making; Threshold technique.