In a comparative trial, we evaluated two versions of a decision aid that differed only in the inclusion or exclusion of two segments, one of which included an explicit discussion and endorsement of the option of deciding not to be screened. The version with the explicit discussion of the option of no screening ("with" version) was rated subjectively as less strongly in favor of CRC screening, but had lower subjective rating of clarity and a lower overall rating. We found no statistically significant or clinically important differences between versions in terms of interest or intent for screening, or knowledge.
We are unaware of other studies that have directly compared including or not including information describing the option of "no screening" within the same decision aid. Two previous CRC decision aid studies included information about the option of not being screened and compared it against a minimal information control group. Both studies were small, but did not find differences in screening between groups [11
]. The decision aid by Wolf and colleagues included this explicit statement that "another option is to not be tested at all, unless signs of colon cancer develop, though at that point, cancer is less likely to be curable" [14
]. Dolan and colleagues used an analytical hierarchy model for "choosing [the] best approach for colorectal cancer screening" which included a "wait & see" option that described the option of not being screened [11
]. Previous iterations of a CRC decision aid developed by our team have not included the explicit option of "no screening" and have shown increases in interest and intent to be screened, as well as actual increases in screening rates [13
A number of questions arise when determining the content of decision aids including the amount and type of information that they should attempt to communicate. In another lab-based study, we have explored the amount of screening information provided, namely the number of screening tests discussed in a decision aid [24
]. This work also found that the number of tests discussed in a decision aid did not affect interest in screening.
Another question to consider is who communicates the information in a decision aid. The "with" version of the decision aid included an endorsement of the option of no screening by a health services researcher and to balance this, a physician endorsing the option to be screened. Opposing endorsements may have contributed to the perception of balance in this version but also may have reduced clarity. Although we did not test the specific effect of expert versus non-expert endorsements, nor the type of expert providing the endorsement, the subjective differences noted between the two versions may be affected by endorsements and who provides them.
In this case, the endorser of the option of no screening placed considerable value on the time requirements for screening and made a judgment about the chance of a potential benefit of screening relative to this time cost. If the viewer holds a different value for his or her time, or feels differently about the chance of benefit, he or she may be misled by the endorser's conclusion. Neutral representation of the options, both by a narrator or even graphical representations, may produce different effects, and should be compared in a future study [25
]. Testing the effect of vignettes versus other methods is a high priority for research [26
Our study has several limitations. First, it is a small study, which does not allow us to definitively rule out small but meaningful differences between groups. We examined several subjective outcomes, and did not adjust for multiple comparisons, so our findings should be interpreted cautiously. Second, the knowledge questions that we used did not evaluate participant's knowledge of the relative risks and benefits of the various screening options. Hence, they may not be ideal for evaluating whether the decision aid fulfilled the goal of educating patients about the relative risks and benefits of the various screening options available. Third, we did not measure actual screening behavior, so it is possible that the two versions would have different effects on that endpoint. Our previous studies have found strong correlations between interest or intent and actual test completion suggesting that those who indicate interest and intent are more likely to complete screening [13
]. Nevertheless, our findings should be confirmed in a larger randomized trial of unscreened individuals viewing the decision aid that also measures whether screening is completed. In addition, we did not assign participants randomly to the two study groups, instead allocating by gender and participant availability. We did note some differences between groups at baseline, but controlling for them did not affect our findings. We had diverse geographic representation, but our participants were volunteers, the majority had undergone some form of screening, and were highly educated, affecting the finding's generalizability to other clinical populations.
It is also important to note that the amount of the decision aid devoted to the discussion of "no screening" was less than two minutes in a 35 minute program, and that overall both programs were seen as favoring screening. The differences between the decision aids were mainly related to the effect of how the health services researcher and physician interviewed with opposing views on screening valued the magnitude of benefit from screening and did not differ in terms of factual information on risks and benefits. It is possible that our results would differ if a greater proportion of time was spent on discussion of the option of "no screening."
Our study examines the question of colorectal cancer screening, where the evidence of benefit is relatively strong and its magnitude of effect relatively large. These results may not generalize to other decisions where the evidence is more uncertain, such as prostate cancer screening, mammography for women in their forties, or decisions about heart disease prevention.