| To Register SMDM Homepage | |
|
|
|
||||
Purpose: When providing risk information, researchers often use side-by-side presentations of total risk. For example, people see their baseline risk of an outcome (e.g., a 10% chance of an MI) and then the risk experienced when taking a medication (e.g. a 14% risk of an MI). This approach, however, forces people to do mental subtraction to tell how much risk the medication causes (here, a 4% increase). We tested whether presenting this risk difference explicitly would improve comprehension in graphical and textual risk communications.
Methods: We recruited women from a demographically balanced panel to participate in an online survey experiment about tamoxifen. Participants read about the use of tamoxifen to prevent breast cancer and then saw information describing how the risk of four types of side effects would increase with tamoxifen. Risks were shown either in numerical text format or in pictographs, and participants were randomized to either see a side-by-side display of total risk with and without tamoxifen or a sequential presentation of the additional risk generated by taking tamoxifen. We also randomly varied two secondary factors that have been shown to bias risk perceptions: the denominator of all risk statistics (“out of 100” versus “out of 1000”) and the presentation order of the side effects (very unlikely, but severe, side effects first versus common but less severe side effects first). Analyses focused on women’s ratings of how much they would worry about each side effect if they took tamoxifen.
Results: 1789 women completed the online survey. Across all four side effect categories, women’s worry ratings were significantly higher for side-by-side total risk presentations than for presentations that highlighted the additional risk caused by tamoxifen (all p’s < 0.001). This effect was significantly larger with textual risk descriptions than with pictographs in 3 out of 4 cases. Furthermore, while side-by-side total risk presentations repeatedly showed significant unwanted denominator and probability order effects, additional risk presentations never did.
Conclusions: Additional risk descriptions lowered worry about side effect risk by highlighting the fact that most risk existed at baseline. Moreover, the additional risk format was resistant to two cognitive biases: order effects and denominator effects. Presenting side effect risk in a pictograph format, with additional risk specifically highlighted, yields the most consistent pattern of risk perceptions.
See more of Opening Plenary Session (6)
See more of The 26th Annual Meeting of the Society for Medical Decision Making (October 17-20, 2004)