Response to Letter Regarding the Impact of U.S. Preventive Services Task Force Recommendations in Breast Cancer Screening Trends

We thank Dr. Levy for his insightful, helpful comments regarding this methodologic change with Behavioral Risk Factor Surveillance Survey weighting from post-stratification to raking. We are grateful for the opportunity to consider how it may have impacted our findings and their interpretation,^1,2 and we agree even a minor adjustment to sampling and survey weights can result in differences in estimates, even for small effects. In this case, it may have had little impact on our primary conclusions, given that the shape of trend lines in screening mammography adherence and initiation were relatively stable over time and within-year trends were consistent across 2010 (before the changes) and 2012 (after the changes) (Appendix Figure 1 and Appendix Table 2 in the manuscript present these results in detail). Further, the pattern of mammography utilization found in our study is consistent with findings from other data sets, including Medicare data analyses.³ Thus, it seems unlikely that our findings would be simply attributed to minor changes in sampling weights applied in the Behavioral Risk Factor Surveillance Survey.

Nonetheless, we fully understand the importance of accurately representing the data. Future analyses with the Behavioral Risk Factor Surveillance Survey and similar population-based data sets should carefully account for evolving methodologic issues and complex sampling designs. We also acknowledge that limitations with surveillance data exist more generally and are also necessary to consider in interpreting population trends in health behaviors and outcomes, particularly as we monitor such a major policy impact such as the Affordable Care Act.