We are unable to determine whether providing the value

clarification first, as was done in Groups 2 and 3, led to improved decision quality, independent of the order effects. Third, while using Mechanical Turk as a recruitment method enabled us to enrol a fairly large sample in spite of limited study resources, the method raises some BI-6727 concerns about sample representativeness and data quality [23]. Turkers are more likely to be younger than the general population, female, and have a lower income [33]. They therefore do not reflect the characteristics of sleep apnea patients. In terms of quality, we had to exclude 5% of the sample for not reading and understanding the treatment information correctly. Otherwise, we believe our data quality reasonably reflects that of other studies [34]. Fourth, our use learn more of MCDA to ascertain the optimal option for each individual relies on certain assumptions [35]. We chose MCDA because it is a simple approach for individuals to use in deciding between options. While we assume that some treatments are suboptimal, we acknowledge that these options actually may be optimal for some individuals. Finally, we could have increased our ability to identify order effects if we had used a PtDA for a more complex

treatment decision. Over 70% of individuals were able to select the optimal option using a fixed order, which leaves limited room for improvement. Future studies should focus on decisions where individuals tend to make poor judgments. Harnessing the influence of order effects and individualizing the way health information is presented may help patients make better quality decisions. While the effects we observed are relatively small, order effects can be implemented at little cost, particularly as web/computer based PtDAs are becoming indispensable for delivering individualized risk estimates and communicating patient stories [36]. This study

contributes to a growing SB-3CT literature demonstrating that developers of static PtDAs may have unintentional but important influences on which options patients choose. This work represents one example of using behavioural design to help individuals overcome cognitive errors. Other strategies to overcome position effects have included methods to debias health information, such as through use of pictographs or incremental risk information [15]. However, these approaches typically require individuals to view even more information, making them susceptible to other biases such as information overload [37]. One promising approach for improving patient decision-making is through exploiting cognitive biases or by using so called ‘nudges’ – “aspect[s] of the choice architecture that alter people’s behaviour in a predictable way without forbidding any options or significantly changing their economic incentives” [5].