Persuasive Interventions for Controversial Cancer Screening Recommendations: Testing a Novel Approach to Help Patients Make Evidence-Based Decisions

We appreciate the thoughtful comments from Scherr. As we stated, we expected this approach would arouse controversy, since, like IPDAS, we all like to think that we should be able to present patients with balanced, unbiased information, following which they should be able to make an informed choice consistent with their values. Alas, research and experience suggest that this ideal is rarely achieved. Unfortunately, given the extensive advocacy messages to which patients are frequently exposed, the weight people place on powerful anecdotes (e.g., stories of family and friends being treated for advanced, unrelated cancers), and well-documented flaws in human decision making, it is very difficult for patients to make informed, rational decisions consistent with their values. The literature on effects of decision aids for prevention confirms this - typically, knowledge increases modestly (with scores still usually <50%; should we forbid choices until patients score at least, say, 80%?) and there is often little or no change in preferences. We hypothesized, and learned, that getting patients to seriously consider an evidence-based recommendation contrary to what they currently believed required initial, careful, substantial effort to "de-bias" them enough that many were willing to reconsider their preferences. We also learned that, even though patients rarely ask for evidence when cancer screening is recommended, they wanted to know evidence was there if they were going to consider not being screened. We did our best to present unbiased information, incorporating recommendations of other groups that conflict with those of the USPSTF into our videos, though we portrayed a physician endorsing the USPSTF recommendations because they were developed with an evidence-based process that sought to minimize bias. It is impossible to completely avoid bias - even choices about whether to present information about benefits or harms first introduces bias. But, as Scherr notes, we think that our outcomes suggest that we probably did a reasonable job in presenting both sides so that viewers could make better-informed choices and, we think, far less biased than what they would otherwise make given pervasive social messaging.

We set out to develop decision aids that would counter pre-existing biases and support what we believed to be unbiased recommendations from the USPSTF - though some might argue that the USPSTF's typical insistence on taking a very conservative view of the evidence is, itself, a type of bias. We did not intend or have resources to do an extensive literature review/evidence analysis on our own, so we necessarily focused on the benefits and harms addressed by the USPSTF. Patients might weight some factors differently than the USPSTF - e.g., many women in our focus groups indicated they did not weight the harms of false positive mammograms as heavily as the USPSTF did. Some might value earlier diagnosis and treatment of a cancer that does not affect mortality, particularly if it might lower lifetime treatment burden. After presenting different recommendations, we encouraged patients to think about their values and what was most important to them. Our team did discuss issues not well- addressed by the USPSTF, including harms of treatment for advanced, but not ultimately fatal, cancer that might have been avoided by screening. However, without quantitative estimates of such possible outcomes, we felt that raising them could introduce bias and therefore did not include them; viewers would have no information for weighing them against the quantitative benefit and harm information we provided for other outcomes.

We also appreciate Harper's kind compliments. In response to her question, alas, we have no means to follow our participants; the funding is gone and all identifiers have been removed to make our data anonymous. We are hoping to conduct trials of these and/or similar decision aids in health care systems that could provide long-term follow-up and allow recruitment of larger groups of patients - and would welcome inquiries from anyone interested in collaborating on such studies in a setting that could support such assessment.

Competing interests: None declared