Table 3.

Recommendations and lessons learned

Topic	Recommendation
Design	Splitting the design process into two phases—one to develop a generic template with key sections and information that patients want from the results, and one to populate that template with the specific numbers and information for each type of test—may provide an efficient way to produce large numbers of report templates for medical test results.
	Test with users: recommendations from the literature should not be applied blindly. For example, although there are good reasons to present risk figures in multiple formats as a general rule, in our case including “1 in 25 (4%)” and “1 in 4 (25%)” in close proximity caused confusion. User testing permitted us to address the issue in a way that allowed us to continue following the recommendation but also eliminated the confusion.
	Focus on recruitment of diverse representative end users throughout the process. We benefited from multiple perspectives of different user groups (health-care providers, patients, and members of the public with varying levels of experience of genetic testing), and would have benefited from a more concerted effort to recruit participants who were more diverse in other ways (e.g., education).
Evaluation	Following up on comments from interviews with a larger sample size can be a useful way to determine whether an offhand comment (“I don’t know how I missed that!”) is indicative of a larger issue (27% of participants indicating that they did not see the result summary box).
	Formative and summative evaluation both ought to be applied to important patient-facing materials whenever possible.
Vocabulary and wording	When using vocabulary that implies a change in risk (e.g., reduce/increase), the risks being compared must be clearly described.
	For patient-facing materials, “gene changes” is a poor plain-English alternative to “variant,” as it sometimes led to misinterpretations (e.g., “What does it mean by no cystic fibrosis gene changes detected? Can genes change throughout the life course or something? I thought you’re kind of born with it or you’re not.”) In our study, “alterations” seemed to be reasonably well received and interpreted.
	Prior literature39 has found that a quarter of people incorrectly answer the question “Which of the following numbers represents the biggest risk of getting a disease? 1 in 100, 1 in 1000, or 1 in 10?”, not realizing that a larger number in the denominator corresponds to a smaller probability. A quote from one of our participants suggested she had a similar misapprehension (“less than 1 in 500 sounds less scary, because then you can think, oh, it could be 400 or 200”). When presenting probabilities that are intended to be compared with each other, keep denominators constant to decrease the chances of misinterpretation, i.e., compare 1 in 1000 with 6 in 1000 rather than comparing 1 in 1000 with 1 in 167.