Program impact implications
There is little debate as to the importance of exclusive breastfeeding ; however, the effectiveness of programs such as the BFHI has been questioned and there has been a reduction in international support for this program (Labbok M, personal communication from global query carried out in 2006, ). Despite the many studies demonstrating the impact of BFHI on breastfeeding rates at the hospital level [6–11], and those that show its impact at the individual country or sub-regional level [8, 12, 13], no previous study has utilized a multi-country construct based on the actual year of national BFHI implementation, with exclusive breastfeeding as the outcome variable.
Results suggest that among the countries under study, there were no significant upward trends in EBF rates in the years prior to BFHI implementation, but that BFHI implementation was associated with a statistically significant annual increase in rates of EBF in the first two, as well as during the first six, months. The two month rate of increase was higher than the six month rate, as might be expected with an immediate postpartum intervention.
According to the models, a country that implemented the BFHI would experience, on average, a 7.7- and 5.5-percentage point increase in the first two, and first six, months of EBF respectively, over a subsequent period of five years. If improvements in EBF practices are sustained over time, such an increase could provide a significant improvement in child health outcomes. One can estimate the impact of such an increase as follows: based on the accepted estimate that a 51% increase in EBF is needed to reduce child mortality by 13% (i.e., from the 2006 estimate of a 39% prevalence to the 90% prevalence used for calculation of the 13% reduction in child mortality ), we estimate that a 5.5% increase in EBF in the first six months has directly reduced annual child mortality by about 1.4%, or prevented about 140,000 deaths. The fact that the slopes of these trend lines did not differ significantly from one another is a call for caution in interpreting these findings. However, the fact that the definitions of EBF became more conservative over time may have blunted the slopes in the later data, reducing the likelihood of achieving significance even if a true increase in positive breastfeeding patterns had occurred.
One strength of our analysis is the use of the "zero" year to re-center all data to the time of country-specific BFHI implementation. This allowed for countries to serve as their own comparisons over time, and for cross-national trends to be considered in relation to the start of BFHI programming, adding strength to the argument that observed trends are derived from BFHI activities. This adds to our understanding of the impact of the BFHI as it was implemented.
The limitations of this study include the reality that the 14 countries analyzed represent a small portion of all developing nations that have implemented the BFHI, and exhibit relatively low rates of hospital certification. In addition, we had no measure of the level of ongoing adherence to the Ten Steps or of the general quality of BFHI implementation over time in this sample. Our results, therefore, are not necessarily reflective of the program's potential to improve breastfeeding rates if implemented on a national or global level.
A serious limitation of any effort to evaluate the effectiveness of the BFHI on cross-national breastfeeding trends is the lack of data collected specifically for this purpose. In this study, we have had to rely on a relatively small number of data points. As such, our study was not powered to detect small differences in trends between pre- and post-BFHI time periods. The limited number of data points also hindered examination of non-linear models that may have provided more insight into the behavior of these trends over time. Our use of overlapping trend lines to compensate for a lack of "zero" year data points, and incomplete information on Baby-Friendly changes that may have been instituted prior to actual certification, further impaired our ability to detect differences between pre- and post-BFHI time periods. We cannot fully predict how access to additional EBF measurements, data from additional countries or information about possible preexisting Baby-Friendly practices may have changed our results.
The use of fixed effects models allowed us to control for the presence of measured and unmeasured confounders that were fixed over the time period studied, but did not control for factors that were variable over the time period, such as demographic changes, shifts in maternal employment patterns, or other breastfeeding promotion programs implemented concurrently with the BFHI. We lacked sufficient information to control for these variables appropriately. With the exception of concurrent public health programming, we would expect most changes over this broad time period, including increased urbanization and women's employment, to have negatively impacted EBF . For this reason, we feel that the observed trends may represent a conservative estimate of the program's potential.
If and when sample size and available data permit, additional analyses are needed to consider trends taking into account the percentages of maternity facilities ever-certified as Baby-Friendly, the percent of all births that occur in these facilities, and continued compliance with and investment in the program. Such analyses would help to determine whether a dose-response relationship exists between the level of BFHI programming and trends in EBF over time. Further research is also needed to investigate the existence and impact of other local and national breastfeeding promotion and support programs implemented concurrently with the BFHI.