Purpose: The purpose of this study is to introduce an important element from descriptive decision making into health preference measures in order to begin to understand how patients react to uncertainty in medical decisions.

Background: Uncertainty is an essential element in real-life decisions, and is known to change responses in Subjective Expected Utility Theory and Prospect Theory (PT) situations. As health utilities are being modeled with PT, it seems important to try modeling health preferences using an earlier modification in economic utility, namely the inclusion of uncertainty in decisions. This study introduces choice-based and ratings-based health preference measures that incorporate uncertainty. It examines differences in health preferences between response modes, and the presence or absence of uncertainty.

Method: This study was a controlled experimental trial with 55 male veterans suffering from prostate cancer, cared for in the urinary and cancer care clinics at Edwards Hines Jr. VAMC. Two questionnaires and a brief survey were given to all participants, featuring hypothetical health treatments. Each questionnaire featured a different preference measure; either ratings or choices. About two-thirds of the questions contained uncertainty in either the number of benefits, or the probability that the treatment will have the indicated side effects and benefits. Total time for completion was usually about 1 hour.

Results: The majority of patients preferred the choice measure, even when uncertainty was incorporated. Choices were easier to use and represented their treatment opinions better. Both types of uncertainty in the choice measure reduced differences in preferences between treatments. In the rating measure, uncertainty in health benefits lowered preferences for nearly all treatments, while uncertainty in the probability of the treatment working as indicated had the surprising result of increasing preference for all treatments.

Conclusions: Uncertainty in treatment information changes health preferences. Here, uncertain probabilities made a larger difference in both ratings and choices. Choices show the salience of benefits and side effects (outcomes) are greater than that of probabilities, when the response mode does not highlight outcomes or probabilities. However, a rating in the form of a probability may lead patients to artificially raise valuations of uncertain probabilities. These results are further supported by recent economic-based work with Prospect Theory and Ellsberg's paradox.