Background: Accurate measurement of people's risk perceptions is important for numerous bodies of research and in clinical practice, but there is no consensus about the best measure.
Objective: This study evaluated three measures of women's breast cancer risk perception by assessing their psychometric and test characteristics.
Design: A cross-sectional mailed survey to women from a primary care population asked participants to rate their chance of developing breast cancer in their lifetime on a 0% to 100% numerical scale and a verbal scale with five descriptive categories, and to compare their risk to others (seven categories). Six hundred three of 956 women returned the survey (63.1%), and we analyzed surveys from the 566 women without a self-reported personal history of breast or ovarian cancer.
Results: Scores on the numeric, verbal, and comparative measures were correlated with each other (r > 0.50), worry (r > 0.51), the Gail estimate (r > 0.26), and family history (r > 0.25). The numerical scale had the strongest correlation with annual mammogram (r = 0.19), and its correlation with the Gail estimate was unassociated with participants' sociodemographics. The numerical and comparative measures had the highest sensitivity (0.89-0.90) and specificity (0.99) for identifying women with very high risk perception. The numerical and comparative scale also did well in identifying women with very low risk perception, although the numerical scale had the highest specificity (0.96), whereas the comparative scale had the highest sensitivity (0.89).
Conclusion: Different measures of women's perceptions about breast cancer risk have different strengths and weaknesses. Although the numerical measure did best overall, the optimal measure depends on the goals of the measure (i.e., avoidance of false positives or false negatives).