AVERSION TO AMBIGUITY ABOUT THE RISK OF TREATMENT HARMS INFLUENCES DECISION MAKING AND VARIES BY LEVEL OF RISK AND AMBIGUITY

Purpose: Most people exhibit at least some aversion to both risk and ambiguity—i.e., uncertainty arising from limitations in the reliability, credibility, or adequacy of risk information. The objective of this study was to determine the relative importance of each in the treatment decision-making context, and to determine whether aversion to ambiguity about the risk of treatment harms varies by level of risk and ambiguity.

Methods: We invited members of an online panel to complete a survey which sought choices between two hypothetical rheumatoid arthritis treatments based on different levels of 4 attributes: probability of benefit (40%,55%,70%), probability of iatrogenic harm requiring treatment withdrawal (10%,20%,30%), ambiguity regarding risk of harm, expressed by risk estimate imprecision (none: point-estimate, low: range+/-5%, high: range+/-10%), and life expectancy (8,9,10 years). Each respondent answered 10 pairwise Discrete Choice Experiment questions indicating strength of preference for each treatment,  described using differing levels for each attribute as generated by a D-efficient experimental design. Conditional and mixed logit models were used to estimate coefficients for each attribute level and allow for estimation marginal willingness to pay (WTP) associated with giving up life to move between each of the attribute levels in the survey.

Results: Of 252 respondents, the mean age was 34 (20-67), 68% were male, and 52% had an education level of high school or less. Respondents placed greatest value on greater probability of benefit (WTP=1.34 years for 60% vs 40%, p<0.001), lower risk of harm (WTP=0.4 years for 10% vs 30%, p<0.001), and lower ambiguity (WTP=0.14 years, high vs none, p<0.001). Preliminary investigation of interactions suggested respondents were ambiguity-seeking at low harm risk (10%), ambiguity-averse at high risk (30%), and ambiguity-neutral at intermediate risk (20%). The degree of ambiguity aversion at high risk (30%) increased with the degree of risk estimate imprecision.

Conclusions: Ambiguity aversion may influence treatment decision making, and varies by level of risk and ambiguity.  People are ambiguity-averse at higher treatment harm risk, but ambiguity-seeking at lower risk, and the degree of ambiguity aversion depends on the degree of ambiguity.  More research is needed to quantify differences in responses to ambiguity at other levels of risk and ambiguity, and for other outcomes, treatments, and disease states.