The US is no stranger to maternal mortality. The country has gained notoriety for having the highest maternal death rate of similarly wealthy nations; on a global ranking, the US falls 55th, just behind Russia and just ahead of Ukraine. Around 800 pregnant Americans die each year—though those numbers have spiked since the COVID pandemic—and the Centers for Disease Control and Prevention (CDC) reports that two-thirds of these pregnancy-related deaths are preventable. In other words, more than 500 mothers die each year unnecessarily.
These vital statistics are paramount to informing maternal healthcare policy, but the data on death rates is far more complicated than one might expect. Three different systems track maternal mortality, all with varying constraints on how information is analyzed. And as the country enters an era with increasingly strict restrictions on abortion and maternal care, properly measuring and understanding these numbers is crucial for creating support systems and saving lives.
What does it mean to die?
When a person dies, their cause of death is assigned a code. “W67,” for example, refers specifically to a death from accidentally drowning in a swimming pool. “J14” is death from pneumonia, but just from the Hemophilus influenzae bacteria. And “O72.2” is when a mother dies from excessive blood loss more than 24 hours after giving birth. Physicians, medical examiners, and coroners document this information on death certificates, as well as any other contributing causes. It’s a clinical process at the end of an emotional experience; experts collect and analyze the codes of deaths across the country to learn what people are dying from. Then, we can create policies to respond to those causes.
“Births and deaths—this information is the foundation of our public health system,” says Marie Thoma, a reproductive and perinatal epidemiologist and professor at the University of Maryland.
In their research on maternal mortality, Thoma and her colleague Eugene Declercq, an Obstetrics and Gynecology professor at Boston University’s School of Medicine, have studied how the structure of these death certificates and informational processing systems has led to inconsistent and inaccurate data. “People think maternal mortality is kind of straightforward because they assume it’s what they’ve seen on TV, which is a tragedy at the time of birth,” Declercq says. But in reality, it’s more nuanced. Experts must get this data right, he says, because “with maternal mortality, we’re talking 700 or so deaths per year and the mistakes [in analysis] are magnified, as opposed to analyzing overall death rates.”
How do we track maternal mortality in the US?
Determining a precise maternal mortality rate for the entire country is hard because there are three systems tracking this data—and they operate in different ways.
The variation between these different systems starts, most notably, with how maternal deaths are defined. The National Vital Statistics System (NVSS), run by the CDC, is the country’s official source of maternal mortality data and is used for international comparisons. Maternal mortality, as defined by the World Health Organization and so adopted by the NVSS, is considered as the death of a person while pregnant or within 42 days of birth or termination of pregnancy. It also requires that the cause of death be related to or aggravated by the pregnancy. (Not, for example, from a suicide or an accident.) The Pregnancy Mortality Surveillance System (PMSS) is a broader measurement. It counts pregnancy-related deaths – people who died during or within one year of pregnancy and for reasons that are “from a pregnancy complication, a chain of events initiated by pregnancy, or the aggravation of an unrelated condition by the physiologic effects of pregnancy.” Maternal Mortality Review Committees (MMRCs) are multidisciplinary groups in states and cities that perform comprehensive reviews of deaths among women within a year of the end of a pregnancy. They collect the most detailed data, though they are the newest systems and, at the moment, aren’t as comprehensive as the other systems.
These definitions are important because they shape how data is tracked. Remember the codes that indicate a person’s cause of death? Well, there’s a specific set of codes for anything to do with obstetrics: the “O“ codes. Until 2003, identifying maternal deaths was tricky: States had their own systems and there was no way to consistently track whether the death of a pregnant person happened during pregnancy, a week after, or within the past year—the important distinctions that impact how the NVSS and PMSS systems define it. In 2003, the NVSS system introduced a checkbox on death certificates that offered consistent options for people to denote when someone died and their pregnancy status. But not all states adopted the policy until 2017 and it was still being filled out inconsistently—so the NVSS paused its reporting on maternal mortality rates from 2007 to 2018, when officials felt the data would be cleaner and more consistent.
The NVSS system has made two improvements to the system, one of which is limiting the age range of maternal mortalities to people under age 44, to have less potential data interference. But this distorts the data for older women. “It’s possible we may be underestimating [rates for] older women to some extent and perhaps still overestimating [rates] the younger ages,” says Bob Anderson, chief of the Mortality Statistics Branch at National Center for Health Statistics, which monitors the NVSS system. “But we think that overall we’re getting better information.”
The NVSS system double-checks records if they have been marked for pregnancy, but it’s not as accurate as the PMSS system. In the PMSS system, medically trained epidemiologists review background information from births or fetal death certificates to more accurately determine the specific cause and time of death, regardless of age. It’s more detailed, but it does take a long time to review.
“The NVSS system is really more for situational awareness,” Anderson says. “We can capture pretty well whether deaths are going up or down even if the actual [exact] level of maternal mortality isn’t quite right. But the PMSS system can provide a little bit more explanatory information and more information about the true burden.”
Together, these systems provide complementary sets of information. However, the two programs running the systems don’t communicate with each other around the data. The logistics of doing that, Anderson says, would require resources that they don’t have. This is frustrating to researchers like Declercq, who say a streamlined and comprehensive dataset would provide a more accurate understanding of the problem and better inform the corresponding solutions.
What do the statistics show?
Inaccurate maternal mortality data has deep implications. In fact, maternal mortality focuses on such a precise portion of the population that even a slight amount of miscoding can produce problematically misleading statistics. Marion MacDorman, a research professor at the University of Maryland’s Population Research Center who passed away earlier this year, conducted a sensitivity analysis in 2017 to study how overreporting maternal deaths (which can happen when death certificates are filled out improperly) might influence mortality rates across different age groups. With just 1 percent of overreporting, maternal death rates among women older than age 40 more than tripled their mortality rate. It doesn’t take much for a small change to have a big impact with this dataset.
Anderson says that the NVSS numbers are effective. At the same time, he recognizes the urgent need to ensure health professionals and policymakers get the most accurate data. “If you’re producing numbers that are sort of wildly incorrect, that could lead you to develop interventions that aren’t needed. Or it could give you sort of a false sense of security, that everything’s fine, when in fact, it’s not,” Anderson says.
And the disparity in numbers is already showing curious results. For example, 2018 is the most recent year that both systems have reported their rates: the PMSS reported 17.4 deaths per 100,000 people and the NVSS system reported 17.4 deaths per 100,000. It’s curious, Declercq says, because the PMSS system includes deaths that occur over a longer time span (one year after birth, compared to the NVSS’ 42 days) and so, should typically be higher than the NVSS number. The PMSS rates have stayed relatively flat over the past few years, near 17 per 100,000, while the last time the NVSS reported numbers was in 2007: 12.7 deaths per 100,000. It’s difficult to discern how much of the NVSS numbers rose dramatically because of an improvement in how the pregnancy-related deaths were counted, or if there has been a dramatic increase.
Anderson says that it’s likely because the PMSS system is able to correct deaths that shouldn’t have been coded as maternal. He also thinks that the PMSS system will likely reflect a similar increase in deaths because of COVID when they release reports on numbers from 2019 to 2021. There is no further explanation for why the numbers are so similar.
The different definitions of pregnancy-related deaths and maternal mortality also lead to different data analyses for the cause and timing of deaths. According to the PMSS numbers, one-third of all maternal mortalities occur after the 42-day period. The NVSS doesn’t include deaths after that time frame in its statistics.
The causes of maternal death vary considerably and depend on when mothers die. These data are based on a report from maternal mortality review committees. During pregnancy, hemorrhage and cardiovascular conditions are the leading causes of death. At birth and shortly after, infection is the leading cause. In the postpartum period, often during the time when new parents are out of the hospital and beyond the traditional six- or eight-week post-pregnancy visit, cardiomyopathy (weakened heart muscle) and mental health conditions (including substance use and suicide) are identified as leading causes. Understanding these variations is key to identifying these conditions early and introducing clinical interventions that can save lives.
But what’s challenging is when certificates are reported in a way that skips the nuance of when mothers die. This shows up in the codes for late maternal deaths (42 days to a year after birth), which do not provide information about the actual cause of death (they just indicate “late maternal death”). Thoma and her colleagues conducted a study of 2016 to 2017 NVSS data and examined nearly 4,000 records for maternal mortality. Among the 1,691 records originally coded as maternal deaths, 43.5 percent were originally coded to non-specific causes. On her end, Thoma was able to recode 94.4 percent of these cases to more specific causes of death with more information from the death certificates, leaving only about 5 percent unspecified.
Breaking down the numbers this way, Declercq says, allows healthcare providers to allocate resources to where they’re needed most. Over time, the number of deaths that happen from causes related to the time of birth have gone down, while deaths related to issues after birth have gone up. This, Declercq says, should be helpful information that determines where professionals turn their attention.
“Asking if deaths happened during pregnancy, at birth, or after was a game changer for all of this,” he explains. “People started thinking about [maternal mortality] as not just a clinical problem where we have to reform hospital procedures at the time of birth, which they have already done a lot of efforts around. If women are dying after their pregnancy, then making sure you have [medical supplies like] a crash cart for a hemorrhage is not going to save that life. And so you have to start thinking about this at the community level.”
Who will fix the system?
These days, Thoma is studying how suicide and opioid use impact maternal death. According to a 2021 CDC study, homicide, suicide, and drug overdoses are the leading cause of pregnancy-associated deaths, which include individuals who have died up to one year after birth. This information is often missed, Thoma says, because of the constraints in how we define maternal mortality. But identifying this information can create pathways to direct more relevant community support and medical care directed to new mothers in this time period and drastically alleviate strain.
This became clear during the COVID pandemic, when the American Rescue Act created pathways for states to extend Medicaid coverage for pregnant people from 60 days to one year postpartum. State governments bear the responsibility for adopting the change, and health advocates have argued for an extension of Medicaid policy for new mothers from 60 days to a full year, which some say would help decrease the number of deaths. New plans from the White House earlier this year advocate for a similar change, and just last month, Hawaii, Ohio, and Maryland all approved this policy.
Not all states are responding in the same way. More researchers are studying how the recent reversal of Roe v. Wade might impact maternal mortality. Thoma says that places with fewer resources for maternal health and training on how to fill out death certificates properly often have more inaccurate data on mortality rates—which means they have a worse understanding of how pregnant people are dying. This fuels concerns about the health outcomes for mothers and children, especially given that the six states with the highest maternal mortality rates in the nation quickly banned abortions.
This is “geeky” work, Declercq says, but it’s important to do because these policies are a measure of how the US approaches maternal care. “We have a lot we can learn from studying maternal deaths that can be applied to women’s health in general,” he notes. “It’s not just about pregnancy; it’s not about keeping women healthy during their pregnancy and then abandoning them after they’ve had their baby. It’s about how we make a system that allows women to be as healthy as possible.”