Reply

Abstract

We appreciate the opportunity to respond to the thoughtful comments on the methodology that we used in our analysis of breastfeeding outcomes and their relationship to the designation as a Baby Friendly Hospital. The authors state that they performed alternative analyses using the 2014 birth cohort and weighted regression, as well as sensitivity analysis of the 2016 births, and obtained results that differ from ours. As we explained in our article, the 2018 CDC Breastfeeding Report card is based on the 2015 birth cohort.1Centers for Disease Control and PreventionNational Center for Chronic Disease Prevention and Health Promotion, Division of Nutrition, Physical Activity and Obesity. United States breastfeeding report card, 2018.https://www.cdc.gov/breastfeeding/pdf/2018breastfeedingreportcard.pdfGoogle Scholar Because there is no published 2015 report card, we used the 2016 birth cohort because it includes all the 2015 Baby Friendly–designated facilities, as well as those that were in the final stages of designation. Our use of the 2016 birth cohort thus provided a greater opportunity for all facilities participating in Baby Friendly designation to show a positive impact on outcomes. This was an important consideration, given the implementation during that same time period of 2 major federal initiatives, including the Best Fed Beginnings Program, in which Louisiana (the authors’ home state) participated. In 2016, 18.6% of US births occurred in Baby Friendly facilities, including 2 states with >85% Baby Friendly penetrance.2Centers for Disease Control and PreventionNational Center for Chronic Disease Prevention and Health Promotion, Division of Nutrition, Physical Activity and Obesity. United States breastfeeding report card, 2016.https://www.cdc.gov/breastfeeding/pdf/2016breastfeedingreportcard.pdfGoogle Scholar In contrast, in 2014, only 7.79 % of births occurred in Baby Friendly facilities, with the highest penetrance of 35.98% in a single state.3Centers for Disease Control and PreventionNational Center for Chronic Disease Prevention and Health Promotion, Division of Nutrition, Physical Activity and Obesity. United States breastfeeding report card, 2014.https://www.cdc.gov/breastfeeding/pdf/2014breastfeedingreportcard.pdfGoogle Scholar Therefore, we feel that using of that 2014 birth cohort for this analysis lacks construct validity. Of note, the breastfeeding initiation rates in 2014 and 2016 were quite similar (79.2% and 81.1%, respectively), and the outcomes that the authors noted may simply reflect the positive results of breastfeeding initiation, consistent with our conclusions. The authors note that we did not use population weighting in our regression analysis but instead treated each state equally. Given the unique and heterogeneous characteristics of the individual states, including their demographics and coexisting programs for support of breastfeeding, weighting by population would erroneously diminish the impact of those important differences. The authors also performed a sensitivity analysis, excluding Delaware and Rhode Island, the 2 states with >85% Baby Friendly penetrance, treating these as outliers. We suggest that this is not appropriate, because these states, which are the least subject to the ecological fallacy and thus have the greatest relevance to the results, should be included for subgroup analysis, as we reported. The authors also dismiss our use of an ecological design to address the relationships that we examined. We disagree, and note that this method is considered particularly applicable to the evaluation of public health strategies when obtaining individual data may be impractical.4Morgenstern H. Uses of ecologic analysis in epidemiologic research.Am J Public Health. 1982; 72: 1336-1344Crossref PubMed Scopus (641) Google Scholar There are many historical examples of important and unanticipated results that have come to light from such studies. We agree that the issue of the ecological fallacy is a limitation; however, there are accepted methods to diminish the impact of that limitation,4Morgenstern H. Uses of ecologic analysis in epidemiologic research.Am J Public Health. 1982; 72: 1336-1344Crossref PubMed Scopus (641) Google Scholar including multiple comparative regression analytics, subgroup analysis of groups with high factor penetrance, and contextual examples of alternate analyses and contemporary approaches, all of which we included to support our findings and confirm that Baby Friendly designation might not be the optimal approach to achieving the US Healthy People 2020 postdischarge breastfeeding objectives. Response to outcomes from the Centers for Disease Control and Prevention 2018 breastfeeding report card: Public policy implicationsThe Journal of PediatricsVol. 227PreviewThank you for publishing content highlighting the importance of evaluating breastfeeding programs and policies. We wish to raise several methodological concerns with the recent analysis1 of the Center for Disease Control and Prevention's (CDC) breastfeeding report card data2 and percent Baby Friendly Hospital Initiative (BFHI) births.3 Full-Text PDF