I wrote a piece that appeared on Slate on Friday. Will the Tide Ever Turn on Breastfeeding? highlights the fact that most of the promised benefits of breastfeeding have never appeared. Worse, aggressive breastfeeding promotion is harmful.
There has been an increase in babies falling from their mothers’ hospital beds or suffocating [due to forced rooming in]. There has been a rise in serious harms to babies including dehydration, starvation, brain injuries, and even deaths. Indeed, exclusive breastfeeding on discharge is now the leading risk factor for hospital re-admission.
Nearly all the pain, suffering and death can be traced to the Ten Steps of the Baby Friendly Hospital Iniative. How could lactation professionals, good people with good intentions, turn out to be so wrong? As I explained in the Slate piece, health recommendations were issued on the basis of small studies without waiting for confirmation by larger studies. Most of these early studies have been debunked. Moreover, small studies, by their very nature, cannot reveal the risks that become serious problems in large population.
[pullquote align=”right” cite=”” link=”” color=”” class=”” size=””]When a measure becomes a target, it ceases to be a good measure.[/pullquote]
The tragedy of the Baby Friendly Hospital Initiative is that a program designed to improve babies’ health has harmed them instead; tens of thousands of newborns are readmitted to the hospital each year for dehydration and jaundice. Equally tragic, a program designed to support mothers has ended up undermining their psychological health.
Why? The answer can be found in data science: Goodhart’s Law.
When a measure becomes a target, it ceases to be a good measure.
Campbell’s Law, a corollary of Goodhart’s Law, is equally instructive:
The more any quantitative social indicator is used for social decision-making, the more subject it will be to corruption pressures and the more apt it will be to distort and corrupt the social processes it is intended to monitor.
What does that mean?
Incentive structures work,” as Steve Jobs put it. “So you have to be very careful of what you incent people to do, because various incentive structures create all sorts of consequences that you can’t anticipate.” Sam Altman, president of Y Combinator, echoes Jobs’s words of caution: “It really is true that the company will build whatever the CEO decides to measure.”
This sketch from the fantastic website Sketchplanations illustrates the problem:
Imagine you are the CEO of a company that manufactures nails and you want to incentivize your workforce to increase profits by rewarding them for meeting production targets.
If you tell them you will offer a bonus to workers who meet a target number of nails produced each week, workers will start making tiny nails so they can produce more of them. If instead you offer a bonus to workers who meet a target for weight of nails produced each week, workers will switch to producing a few massive nails. When a measure becomes a target, it ceases to be a good measure because setting a target distorts and corrupts the process it is designed to monitor.
What should the CEO of the nail factory have chosen to measure?
Data scientist Roman Shraga offers this answer:
You need to ask questions that ensure the measure relate [sic] to the ultimate goal. Additionally, think about whether it would be possible to get a perfect score on the measure, and if it would be possible, to do so without adding any value. This line of reasoning will allow you dissect a measure until you understand whether or not it is doing a good job of indicating performance.
In other words, the CEO should have chosen a measure, or a combination of measures that increases productivity without compromising quality.
Now let’s look at the Baby Friendly Hospital Initiative. Its leaders chose to incentivize hospitals, nurses and lactation consultants on exclusive breastfeeding rates at discharge. That seemed like a good target to choose because the goal was to increase long term exclusive breastfeeding rates and the mothers who breastfeed exclusively for the long term are likely to be exclusively breastfeeding at discharge. In addition, it is much easier to measure exclusive breastfeeding rates at discharge than to track down mothers and babies to see if they are breastfeeding 3, 6 or 12 months later.
How has this target distorted and corrupted the provision of breastfeeding support to new mothers? Just look at the Ten Steps:
Since hospitals, nurses and lactation consultants want to maximize the rate of exclusive breastfeeding at discharge, the Ten Steps make it nearly impossible to avoid breastfeeding. Mothers are hectored to breastfeed, forced to breastfeed within the first hour, denied formula, forced to endure the baby’s cries of hunger by rooming in, and denied pacifiers that might soothe the baby.
When lactation consultants are incentivized to maximize rates of exclusive breastfeeding at discharge, they ignore dehydration, hypoglycemia (low blood sugar) and jaundice because treating them would involve formula and that would reduce the rate of exclusive breastfeeding.
When lactation consultants are incentivized to maximize rates of exclusive breastfeeding at discharge, they make formula hard to get: they restrict access to it, lock it up and force women to sign odious, shaming “consent forms” to get it.
When lactation consultants are incentivized to maximize rates of exclusive breastfeeding at discharge they make sure that hospital personnel will not have to endure the anguished cries of starving infants by closing well baby nurses and leaving babies in mother’s rooms around the clock. And should nurses break down because of simple human compassion and offer formula, they are excoriated by official policy.
Is it any wonder then that exclusive breastfeeding has become the leading risk factor for hospital readmission? By setting the wrong target, the BFHI incentivizes poor, even deadly, care.
What should breastfeeding promotion incentivize?
Since our goal ought to be providing breastfeeding support for anyone who wishes to breastfeed:
1. Mothers’ desires must be accommodated, instead of ignored as they are now.
2. The quality, availability and accessibility of SUPPORT should be measured not the absolute number of infants breastfeeding exclusively.
3. Hospital readmissions must be measured since any effort to promote breastfeeding that leads to an increase in dehydration, hypoglycemia and jaundice is a failure regardless of how high the rate of exclusive breastfeeding at discharge might be.
4. Formula should be easily available; women should be taught how to use it; and judicious formula supplementation should be freely recommended in the early days when babies are most likely to suffer breastfeeding complications.
5. The only breastfeeding rates that are clinically relevant are rates beyond two months. That’s much harder to measure but that’s what actually matters. Measuring exclusive breastfeeding rates on discharge reflects the streetlight effect, also known as the drunkard’s search principle: searching for keys lost in a unlit park under a streetlight because that where it is easiest to look.
The tragedy of the Baby Friendly Hospital Initiative is that bad outcomes were nearly guaranteed by focusing on the wrong target … and failing to understand Goodhart’s Law: when a measure becomes a target, it ceases to be a good measure.
