U.S. Maternal Mortality Rate is Not Really Decreasing—Data Selection Has Changed

I’ve often written in my Substack posts about the inexcusable rate of maternal mortality in the U.S. So it surprised me to find a National Public Radio (NPR) article claiming the data on maternal mortalities reported by the National Center for Health Statistics (NCHS) was too high by a factor of three. This new study claims that the maternal mortality rate in the U.S. is 10.4 deaths per 100,000 births instead of the rates reported by the NCHS below:

Rates increased with maternal age. Rates in 2021 were 20.4 deaths per 100,000 live births for women under age 25, 31.3 for those aged 25–39, and 138.5 for those aged 40 and over (Figure 2 and Table). The rate for women aged 40 and over was 6.8 times higher than the rate for women under age 25.

This team of outside researchers were from Rutgers University, the University of British Columbia, and “other universities,” whatever that means. It should be noted that the above link to the report on the study from Rutgers is a PR piece about the study, not a copy of the study itself where data could be evaluated.

In various studies, the data collected about maternal mortality covers widely different time periods. Many studies define maternal deaths as death during pregnancy or within 42 days after birth from pregnancy-related causes. Other studies extend the period covered to one year after the end of the pregnancy. In this NPR report, we do know the maternal mortality data is limited to pregnancy deaths, deaths during delivery, and postpartum deaths for 42 days after delivery. Clearly, narrowing the timeframe when pregnancy related deaths are counted makes a big difference in results.

The Joint Commission considers behavioral deaths the cause of many maternal mortalities. There are, however, many obstetricians, who don’t feel responsible for any of those deaths, and refuse to count them as pregnancy-related deaths. In their minds, these deaths have nothing to do with obstetrics. In my mind, these deaths do have something to do with obstetricians because obstetricians have the opportunity in the 6, 7, 8 or 9 months of prenatal care to determine depression and to take the opportunity to educate and avoid postpartum behavioral difficulties.

As long as obstetricians don’t see behavioral deaths as their purview, they also don’t see the prenatal course as an opportunity to prevent a large share of maternal mortality. Furthermore, if you don’t count 1/3 of the maternal deaths, naturally, your maternal mortality numbers will look better.

Another problem with the collection of data on maternal mortalities can be found in how death certificates are filled out. What doctors choose to put on a death certificate or not put on a death certificate makes a big difference in what maternal mortality data is collected from death certificates.

When a physician looks at a death certificate that has a choice of a checkbox for death and another checkbox for death during pregnancy, along with a blank to write reason for the pregnancy related death, how the data is regarded by researchers can vary significantly. If a death certificate has the checkbox for “death during pregnancy” marked, some of those researchers collecting maternal mortality data will not count the “death during pregnancy” as a maternal death unless a reason for the death is given.  The person filling out the death certificate may not know the reason for the maternal death. As the NPR article notes:

…when this outside team did their analysis, they only considered deaths where a pregnancy-related cause was mentioned on the death certificate. That's different from the dataset CDC used, which is why the rate came out so different.

I have always considered the maternal mortality rate to be an extremely poor (insensitive) sign of the canary in the coal mine. Nobody knows for sure how many near deaths occur compared to the maternal mortality data, although the near-death number is usually admitted to being greater than the number of maternal mortalities.

According to Simmons-Duffin in her discussion of the Rutgers study with NPR, the CDC initially declined to review the study and to provide comment. However, after the story was published, a spokesperson from the CDC responded that the CDC “disagrees with the findings, asserting that the methods used in the Rutgers study undercounts the number of deaths that should be included.” Well, yes, of course that is a massive understatement. I have indicated how cherry-picking the death certificate data undercounted the maternal deaths to arrive at this new lower maternal mortality rate.

According to Simmons-Duffin, the takeaway here is “That getting this right is complicated.” I would say it is not that complicated. If the box on the death certificate is marked death, and the second box is marked due to pregnancy, then that should be counted as a pregnancy-related maternal death. There is nothing ambiguous about that.

In summary, this Rutgers study and the article have discounted or undercounted maternal deaths in three ways. The first one would be to exclude deaths from day 43 to 365. The second one would be to exclude maternal deaths that didn’t have a written explanation as to the cause of the death. And thirdly, we don’t know whether this study did or did not count behavioral deaths due to suicide, homicide and drug overdose. In her NPR interview, Simmons-Duffin notes:

Well, one takeaway is that getting this right is complicated. It can be really hard to assess when someone dies whether pregnancy or childbirth played a role, let alone capture that on a form like a death certificate.

I believe  the promotion of the notion that maternal mortality is over reported is flawed. I believe the 32.9 maternal mortality rate is more accurate, (closer to the truth) as it does not cut out large groups of maternal deaths which should be included in studies.

I believe the CDC information on maternal mortalities is far more representative of the number of maternal deaths in the U.S. It’s time researchers stopped cherry picking data to try to make the U.S. maternal mortality rate look less than it is. The conclusion of the Rutgers study suggests more work is needed to be sure the data collected is more accurate. The accuracy of the data collected isn’t the problem. Those studying maternal mortality don’t interview physicians with low maternal mortality rates to find out how low maternal mortality rates are achieved in the real world.