Is The Maternal Death Rate Crisis In The US Based On Data Errors?
The finicky issue of maternal mortality and death reporting
For the last decade, it has been fairly well-acknowledged that the United States is having a crisis with maternal mortality, which is broadly defined as deaths that are caused or contributed to by childbirth and pregnancy. Most estimates suggest that maternal mortality rates in the US are not just bad - they’re getting worse every year. Indeed, based recent CDC reports, maternal mortality seems to have nearly doubled in the last 5 years alone.
But a recent scientific study appears to cast doubt on this idea. As headlines have been saying, a new paper casts doubt on the claim that maternal mortality has increased sharply in the last two decades in the US. In fact, the data may show that the rate hasn’t shifted since way back in 2001. The original reporting from the Washington Post even argues that the US maternal mortality crisis may simply be a case of mistakes in the data.
Except, that’s not quite true. What the new study seems to show is that using different ways of analyzing maternal mortality statistics gives us different answers as to how bad the situation in the US really is. It’s possible that the crisis really is real, but it’s also possible that some of the massive increase is down to problems with how deaths are reported.
Let’s dig in.
The Science
The new study is a report in the American Journal of Obstetrics and Gynecology that looks at death reporting data from the US between the years of 2001 and 2021. Specifically, they look at two separate parts of death certificates - the death coding data, and the pregnancy/childbirth check box.
This next bit is very finicky, but bear with me for 3 paragraphs. It’s important.
In the US, as with most countries, death certificates include information for causes of death. This gives whoever is filling out the certificate - usually a doctor, mortician, or similar - a chance to write down what they believe caused the person’s death as well as underlying causes and contributing causes. Underlying causes are things that caused the primary cause - for example, a primary cause of death might be internal bleeding and the underlying cause might be trauma due to car accident - and contributing causes are health issues that the person had which may have made their death more likely. These are sometimes called comorbidities rather than contributing causes, because they tend to be stuff like diabetes or heart disease which often contribute to deaths even if they are not the primary underlying cause.
From these cause of death notes - number 32 parts I and II in the above form - we derive clinical codes which are associated with an individual’s death. These codes are based on the International Classification of Diseases 10 (ICD-10), and are used in death reporting systems both across the US and internationally.
In addition to these codes, however, the US has a fairly unique method of capturing death data on people who died from maternity-related causes. If you look at the above picture, and drag your eyes to box 36 (about 1/3 down, in the middle), you’ll see the pregnancy checkbox. This asks the person who is filling in the form to report whether the deceased person was pregnant at the time of death, in the last year/42 days, or not pregnant at the time of death if they are female. This is a quality improvement measure that the CDC introduced in the mid-00s over concerns that the US was undercounting pregnancy-related deaths that were not caused directly by the pregnancy itself. This is the primary way in which maternal mortality is tracked - if a woman dies, and the checkbox for pregnant at the time of or shortly before death is ticked, it’s considered a maternal death.
Ok, we’re back. The new study looked at two different ways of counting maternal deaths. Firstly, they used the standard measure of the pregnancy checkbox, and found the same results everyone has seen for decades. Then, they looked at something more specific - they ignored the checkbox, and only coded a death as a maternal fatality if it included at least one pregnancy-related disease code in the underlying causes of death. Other deaths in women with the pregnancy checkbox ticked were considered “pregnancy-related”, rather than caused by pregnancy. In this definition, they also excluded suicides, homicides, poisonings, and traffic accidents.
What they found is fairly simple. If you use this alternative measure of death reporting, there has been minimal change in maternal mortality in the US since 2001. Between 2001 and 2021, the rate went from 10.2 maternal deaths to 10.4 maternal deaths per 100k live births. This is in contrast to the standard checkbox-based method of reporting, which shows that such deaths increased from 9.65 to 23.6 per 100k in the same time period.
There were also some odd things about the checkbox-related data. The authors found that some causes of death had increased hugely in the last two decades, particularly cancer-related death. This had leapt up from 0.03 to 1.42 per 100k in the time period studied, which is a remarkable leap considering that cancer treatments have substantially improved in the last 20 years.
The authors used this and the other data they gathered to argue that the differences between the two ways of counting maternal mortality were not just problems, they were mistakes. This is what has caused all of the headlines, because if the authors are right much of the increase in maternal mortality in the US is just people ticking the wrong box on the death certificate form.
The Maternal Mortality Crisis
If you take the authors’ analysis at face value, this means that the maternal mortality crisis in the US is more of a mortality meh. The rate hasn’t improved, which is still bad, and the US is doing very badly regardless when you compare it to other similar nations, but it’s not the overwhelming tsunami that’s been talked about for the last few years.
The issue here is that the authors haven’t really proven that their argument is correct. It’s certainly possible that people are misusing the pregnancy checkbox, but we don’t really know that’s true. You could just as easily argue that in 2001 we just weren’t very good at identifying maternal deaths which were related to things like cancer, which if you’ll remember is the entire reason the checkbox was created in the first place.
This is the key thing that much of the reporting missed. Perhaps the checkbox is working exactly as designed, and helping doctors to capture deaths that they think are related to pregnancy even though the underlying causes might not immediately suggest this. This actually seems to be somewhat supported by the data - while the pregnancy checkbox was introduced in 2003, maternal mortality rates have continued to rise even more than a decade later. This is not what we’d expect to see if this increase was related directly to errors on death certificates, because those would usually be a fairly constant rate of mistakes.
Regardless, it seems that maternal mortality remains a big problem for the US. It’s possible that some portion of the ‘crisis’ that we’ve all been hearing about in the news is overstated, but it’s also possible that this increase is very real.