Baby Friendly Increases Breastfeeding Rates— The Problem with the Fancy Graph Study

By Melissa Bartick, MD, MSc, FABM

Did the Baby-Friendly Hospital Initiative meet its intended goal of increasing breastfeeding across the United States? A recent study claims it failed. The study, published in The Journal of Pediatrics by Bass and colleagues, is replete with fancy statistics and graphs, but the obvious answer is that Baby-Friendly is succeeding.  Anyone interested in breastfeeding knows that breastfeeding rates across the US have risen as the percentage of births in Baby-Friendly Hospitals has risen.  Study after study, including a meta-analysis, have shown the efficacy of the Baby-Friendly Hospital Initiative (BFHI) in increasing breastfeeding rates and improving health outcomes. Furthermore, other studies have shown that the Ten Steps have additive effects in increasing breastfeeding rates.

The problem with the study by Bass and colleagues is that it compared breastfeeding rates among all US states and the percentage of births in Baby-Friendly hospitals among all US states at a single recent point in time. And, not surprisingly, there was no correlation between breastfeeding rates in the different states and the percentage of Baby-Friendly births. Why? For two reasons. First, the very best way to look for the effect of an intervention is with a randomized control trial. This was done with Baby-Friendly in the PROBIT trial, which found that 3 month exclusive breastfeeding rates were 6% in the hospitals without the intervention compared to 43% with implementation of a BFHI-type intervention.  If such a trial isn’t possible, the best way to judge the efficacy of an intervention is to look at the outcomes over time as the intervention is implemented. That way, one can infer cause and effect.

The other reason why the study’s conclusions aren’t valid is that states can have markedly different baseline breastfeeding rates due to a variety of socio-demographic factors, like education, income, and culture.  For example, even before there were many Baby-Friendly Hospitals, Louisiana and Vermont had very different breastfeeding rates. So, course, if you compare their breastfeeding rates to each other as a function of Baby-Friendly, the result will look like nonsense. You can only compare a state to itself over time as Baby-Friendly is implemented. Or you can look at the entire US over time as a function of Baby-Friendly implementation. But you can’t compare Louisiana to Vermont at the same time, which is what these authors did.

As far as the Bass study goes, the authors claimed they tried to minimize a type of bias known as ecological fallacy by controlling for each state’s birth rate. But breastfeeding rates are unaffected by birth rates. It is commonly known that breastfeeding rates are highly affected by sociodemographic factors, yet the authors made no attempt to control for these factors. The CDC reports breastfeeding rates by racial and ethnic groups, and their data shows different racial and ethnic groups have markedly different breastfeeding rates, and we know the populations of different racial and ethnic groups are not evenly distributed among the states. Their study reports sophisticated statistical methods and shows complex figures, complete with mathematical formulae, which gives their study the appearance of professorial authority. Unless one has an advanced degree in biostatistics, the way their methods and findings are displayed may be difficult for most readers to understand. Readers may even fail to notice that the basic premise of their methods is so flawed that the study should have never been performed, let alone published. It is the intellectual equivalence of the “The Emperor Has No Clothes.” The efficacy of an intervention simply cannot be ascertained by looking at one point in time and across populations that are highly diverse.

Here are some statistics to dig deeper, taken from the CDC’s Breastfeeding Report Cards:

In 2007, only 1.8% of US births occurred in Baby-Friendly Hospitals and the US breastfeeding initiation rate was 75% with a 3-month exclusive breastfeeding rate of 33%.  Yet in 2016, 18.3% of US births occurred in Baby-Friendly hospitals and the 2015 initiation rate grew to 83.2% and with a 3-month exclusive rate of 46.9%.  As individual hospitals become Baby-Friendly, they see their exclusive breastfeeding rates at discharge increase markedly.  Achieving certain exclusive breastfeeding rates at discharge is even part of the criteria for becoming a Baby-Friendly hospital, so we know on a micro level that Baby-Friendly increases breastfeeding rates.

For example, in 2004, Louisiana had 0 Baby-Friendly hospitals and a 56.6% initiation rate with an exclusive breastfeeding rate 22.0% at 3 months. By 2016, 12.7% of births occurred in Baby-Friendly facilities, which would leap to 41% in 2018, related to work from the CHAMPS initiative. By 2015, Louisiana’s breastfeeding initiation rate grew to 67% and it’s 3 month exclusive rate grew to 39.6%.

Vermont, by contrast, had an 85.2% breastfeeding initiation rate and 47.3% exclusive breastfeeding at 3 months in 2004, much higher than the national average. By 2015, was at 89.3% and 62.8%.  Vermont had 3.8% of births occur in Baby-Friendly facilities in 2004, and this went up to only 10% in 2015. 

Vermont was starting the breastfeeding race way ahead of Louisiana when Baby-Friendly came in, so a recent snapshot comparing their breastfeeding rates and Baby-Friendly births is meaningless.  In Vermont, breastfeeding has been a normal part of the culture, whereas breastfeeding has historically been seen as unusual in Louisiana. Louisiana has also consistently ranked among the lowest states in household income and education, while Vermont has consistently been ranked as one of the highest.

Vermont’s relatively wealthy, well-educated population may be predisposed to breastfeed at increasing rates over time, despite not having formal Baby-Friendly designation at 90% of its hospitals.   If we only look these states from the current snapshot in time, Louisiana has higher Baby-Friendly “penetrance” with lower breastfeeding rates than Vermont, and we miss the likely impact of Baby-Friendly on that state.   In addition, research shows that Baby-Friendly is important for reducing racial disparities in breastfeeding the Southeastern US.

Multiple previous studies have demonstrated the efficacy of the Baby-Friendly Hospital Initiative and the Ten Steps. Those studies still hold more power than one poorly done study with some fancy graphs.

Blog posts reflect the opinions of individual authors, not ABM as a whole.

###