The number of screening options presented in a decision aid does not appear have a large effect on interest in colorectal cancer screening. Test choice appeared to differ modestly (although the difference did not reach statistical significance) between the 5-option and 2-option version when no out-of-pocket costs were assumed. This difference was not apparent when participants were asked to assume modest out-of-pocket costs.
Those participants viewing the two-test option decision aid were somewhat more sensitive to out-of-pocket costs than those viewing the 5-option version, perhaps because the two tests included, FOBT and colonoscopy, were those that differed most in terms of potential out-of-pocket costs. By focusing the decision on these two options, viewers may have been drawn to weigh the out-of-pocket costs more heavily than if they had been exposed to a wider range of options.
The psychology literature has examined the relationship between the number of choices offered and its effect on decision making. Schwartz' [5
] reviewed much of this work and concluded that providing more choices could lead to poorer decision making processes in both health-related and non-health-related contexts. Work in consumer psychology by Iyengar [6
] and colleagues demonstrated that the provision of extensive choices led to dissatisfaction and decision regret which also supports this premise. Using survey data, Lafata and colleagues found that patients who reported being offered a choice of CRC screening modality were less likely to have complete a CRC screening test in the last 5 years [13
Based on this work, some cancer screening researchers and policymakers have questioned the effect of offering patients more than one option for how to be screened, suggesting that overall screening rates might be higher if the patient is only presented with one method. Although our study did not directly test this question, we did not see major differences in screening interest between offering two or five options. Randomized trials from Italy and Australia have compared the effect of offering one test versus choice of FOBT or sigmoidoscopy and did not find important differences in screening rates [14
Our findings, with respect to test preferences and the effect of out-of-pocket costs, are consistent with some of our previous work as well as work of others [16
]. We had previously shown that out-of-pocket costs affected patient preferences for CRC screening when considering three options: FOBT alone, sigmoidoscopy alone, or the combination of FOBT and sigmoidoscopy [17
]. Leard and colleagues reported that they did not find cost to be an important factor in CRC screening decision making, but they did not compare the effect of providing or not providing such information [16
]. Like Leard and colleagues, we found that nearly all participants expressed a preference for some form of screening.
Our study has several limitations. First, it is a small randomized trial. We did not have sufficient power to confirm or exclude modest, but potentially meaningful, differences in effect. We also did not have a large enough sample size to ensure equal distribution of potential confounders, requiring adjustment in multivariate analysis. However, adjusting for these baseline differences did not change the relationships noted in the bivariate analyses. In addition, we did not assess the amount of information that participants viewed beyond the initial required introduction section of each program. Participants were reporting their hypothetical interests and preferences for screening after exposure to a decision aid. The relationship between interest and preferences and actual test ordering or completion is imperfect. Future studies should examine the effect of the number of test options on actual test completion, including whether the "preferred" test is also the one ordered and completed.
The tests discussed in the decision aid did not include such newer screening options as fecal immunochemical tests, stool DNA, or CT colonography, all of which may have produced different preference patterns. Finally, we did not randomize the order of questions about preferences based on inclusion or exclusion of out-of-pocket costs. It is possible that we would have obtained somewhat different results if we had elicited preferences with such costs before eliciting them without such costs and other co-variates.