Purpose: Determining what makes any given decision aid effective is challenging because each aid includes multiple components. One possible solution to this problem is the use of fractional factorial designs (FFDs). We describe our use of a FFD to identify which of 5 components of a tamoxifen prophylaxis decision aid will have the greatest impact on patient decision making, risk perceptions, and knowledge.

Methods: We are currently testing 5 different factors in our decision aid: 1) information order (risks of tamoxifen before or after benefits), 2) denominator for all risk statistics (100 vs. 1000), 3) risk information presented in graphs vs. tables, 4) describing risk reductions in terms of total risk amounts or risk change amounts, and 5) inclusion of other risks for comparison or not. A fully factorial design of this size requires 64 cells. However, the underlying philosophy of FFD is that only a small subset of components and their interactions will be important, thus reducing the number of cells needed. A FFD study has 3 phases. Stage 1 (screening): Identify the components hypothesized as having the greatest impact, using scientific theory, clinical experience or past experimental evidence. Based on this information, reduce the number of cells by confounding main effects with higher order interactions. Stage 2 (refining): Investigate observed interactions and untangle important effects that were aliased because of the incomplete nature of the FFD. Stage 3 (optimizing): Experimentally compare and validate the optimized components.

Results: Fractional factorial design is allowing us to test main effects and 2 way interactions using 16, rather than 64, cells. This reduces the burden of designing and testing 64 different decision aids. The results will provide valuable information regarding what factors have the greatest impact on risk perception, knowledge, and decision making and whether these factors can reduce commonly found biases. We can also determine whether the factors have a greater or lesser impact for different types of people (e.g., low numeracy).

Conclusions: Fractional factorial designs are very useful for breaking down the many components of a decision aid. To further the field of shared decision making, more studies need to use FFD to determine which factors are critical (or detrimental) to a decision aid's effectiveness.