The questions are straightforward, with public health implications that would seem impossible to shrug off.

How many American women die each year from causes related to pregnancy or childbirth? How many of these deaths are preventable? How does the nation’s current rate of maternal mortality compare to the rate 10 or 20 or 30 years ago?

The answers are central to any true picture of U.S. maternal health, and an essential tool in limiting such tragedies going forward. Much as an accurate census is vital to a functioning democracy, so reliable information on what goes right or wrong in pregnancy and childbirth is key to saving lives.

Yet because of flaws in the way the U.S. identifies and investigates maternal deaths — a process perennially short on funding and scientific attention — what data exists on this particular set of vital statistics is incomplete and untrustworthy. Indeed, for the last decade, the U.S. hasn’t had an official annual count of pregnancy-related fatalities, or an official maternal mortality rate — a damning reflection of health officials’ lack of confidence in the available numbers.

“Our maternal data is embarrassing,” said Stacie Geller, a professor of obstetrics and gynecology at the University of Illinois College of Medicine and a leading scholar on the subject. “Maternal health in the U.S. is simply still not a priority. It’s not interesting. Preventable maternal deaths are not in the basement of our priorities, they are in the sub-basement.”

It’s generally agreed upon that about 700 to 900 American women die each year for reasons tied to pregnancy and that many of these deaths are preventable. Over the last two decades, other affluent nations have reduced their maternal mortality rates, in some cases dramatically.

The best estimates show U.S. rates rising over that period, but those estimates vary — a lot. A study in The Lancet put the 2015 U.S. rate at 26.4 per 100,000 live births, almost twice the World Health Organization’s estimated rate of 14 per 100,000 live births. (Researchers calculated maternal deaths slightly differently in each model, yet came to similar rates for several other countries.)

Doubts about U.S. data on maternal deaths are so profound that some experts have questioned if the rise in U.S. rates over the last 25 years is a mirage, reflecting noise in the numbers rather than a real increase in fatalities.

Elements of how deaths are counted have changed over that time. Most significantly, since 2003, states have added a checkbox to death certificates, asking if the person who died was pregnant, or had been within the last year of their lives. The checkbox has helped identify maternal deaths that previously would have been missed, but also may be capturing cases not related to pregnancy, studies show.

Dr. William Callaghan, who leads the Center for Disease Control and Prevention’s maternal and infant health branch and is a prominent researcher on maternal mortality, said changing data explains some of the rise in U.S. rates — but not all of it.

Even factoring in some over-counting as a result of the checkbox, he and other experts say, the U.S. would still have the highest maternal mortality rate of any affluent country.

“That’s still not great,” he said. “We should be better.”

When it wants to, the American health care system produces a dizzying array of detailed data on diseases and drugs, treatments and outcomes.

States rigorously track infant mortality, for example, collecting dates of birth and death for all babies, as well as rates of teenage pregnancy, capturing the age of mothers on birth certificates. State cancer registries amass granular data to look for trends and patterns. Using standardized data from doctors and hospitals, government contractors scrupulously tabulate every conceivable measure of the performance of transplant centers, including the number of transplants performed, deaths on transplant waiting lists and death rates after transplant.

Callaghan rated the difficulty of measuring maternal deaths at about a three on a scale of one to 10, so he views the longstanding lack of reliable data as a reflection of the scant importance American society places on expectant and new mothers and the urgency of acting to save them.

There’s an “implicit value decision when we don’t do everything we can,” he said. “What we choose to measure is a statement of what we value in health.”

The United Kingdom, which Callaghan describes as having the “gold standard for maternal health data,” starts with vital statistics information but goes much deeper. As part of a national inquiry, health care practitioners review every maternal death, using medical records and more to identify deaths related to pregnancy or childbirth and to determine why they happened and if better care could have prevented them.

Four million women give birth in the U.S. each year, roughly six times as many as in the U.K. Yet the U.S. has no national review of maternal deaths. Slightly more than half the states have maternal mortality review committees — experts that take on in-depth case analyses — that have been operating for at least a year. Many receive little or no funding and rely on volunteers to take on time-consuming case analyses. They publish reports irregularly and, in some cases, do not address the issue of preventability at all.

Members of Congress have introduced several bills attempting to fund review committees nationwide as well as standardizing and pooling their data, but none have moved forward. Instead, a CDC scientist is leading an effort to establish a standardized national reporting system, but, for funding, it depends on Merck for Mothers, a charitable initiative created by the pharmaceutical giant.

Marian MacDorman, a research professor at the University of Maryland’s population research center, said it’s frustrating that decades of flawed data collection have left the U.S. where it is today — still without maternal mortality numbers it trusts enough to certify as official and still without a roadmap to prevent women from dying.

“It’s a tragedy,” she said. “I really draw a direct link. People are dying because the federal government is not publishing this data.”

---

The problems with U.S. maternal mortality data start at the most basic level. For more than a century, the U.S. has been largely dependent on a single source of information — death certificates — to count maternal deaths and understand their causes.

For everyone who dies there is a death certificate, making it the common currency of an otherwise decentralized system. There are 52 separate reporting jurisdictions for vital statistics — 50 states, plus New York City and Washington, D.C.

Certificates are filled out with information from doctors, collected by state and local vital statistics offices and further assessed by experts at the CDC. Jurisdictions have long relied on death certificates (sometimes matched with birth certificates or certificates that report deaths of fetuses not developed enough to breathe on their own) to identify women who died during pregnancy or within one year after giving birth. They shared these records with the CDC, where epidemiologists with the agency’s Pregnancy Mortality Surveillance System use standardized codes for causes of death to try to determine if a death was related to pregnancy or childbirth.

There have always been gaping holes in this process.

The certificate is supposed to identify the cause of death, but when an expectant or new mother dies, it’s not always easy to determine why. Death certificates ask for the immediate cause of death (i.e., the ailment that directly precedes it), intermediate causes (which lead to the immediate cause) and the underlying cause, which sets off the events that result in death. The CDC’s mortality statistics rely on the underlying cause, though this may not always show a link to pregnancy or childbirth.

Death certificates are notoriously prone to error and, often, are missing critical information, CDC officials say. In the case of maternal deaths, they may not be filled out by an OB-GYN or anyone trained to recognize a link to pregnancy — or even anyone conscious that what he or she is recording has public health implications. For decades, researchers found that many death certificates for women in the right age group to give birth missed that they had been pregnant, especially if they didn’t die in childbirth.

Callaghan offered one example of how this could happen: Expectant and new mothers are at greater risk of pulmonary embolism. A young woman shows up in an emergency room, has an embolism, then dies — but the emergency room doctor doesn’t know or note on the death certificate that she gave birth a week earlier. Thus, the death isn’t counted as related to pregnancy or childbirth.

To this day, the codes used to specify causes of death are quite exact for some underlying factors in maternal deaths, but not for others, Callaghan added. This can make deaths tied to pregnancy hard to discern, even for experts like him. “There are death certificates I stare at all day long and I wouldn’t be able to tell you what the real cause of death was,” he said ruefully.

---

Tabulating maternal deaths this way led to a substantial undercount of maternal deaths through at least the early ’90s and possibly beyond, CDC officials say. Studies suggested the true totals could be 50 to 100 percent higher.

In an effort to catch deaths that were being overlooked, the National Center for Health Statistics revised the standard U.S. death certificate to include a checkbox question asking whether the person who died, if female, was pregnant at the time of death or within a year before death.

In some ways, however, the saga of this well-intentioned change reflects the hurdles to achieving comprehensive, accurate data.

Sixteen states already had pregnancy checkboxes on their death certificates well before the NCHS move, some going back to the early ’90s. Problem was, each state used its own wording — some asked if a woman had been pregnant within 42 days of dying, or six weeks, or three months, or 12 months. They weren’t capturing consistent information; a woman whose death might be linked to pregnancy in one state might be classified differently if she died across the state line.

Even after NCHS added checkbox language to the standard national death certificate in 2003, states and cities weren’t required to use it and didn’t institute the change all at once. Instead, a few states made the switch each year, with a couple finally trickling into line in 2016 and 2017. Within New England alone, states added checkboxes as early as 2005 (Connecticut) and as recently as 2014 (Massachusetts).

The checkbox had an instant effect on numbers of maternal deaths.

“Adding the checkbox almost doubles a state’s reported maternal mortality rate,” said Eugene Declercq, a professor at the Boston University School of Public Health who has studied the effect.

There are persuasive signs the checkbox has solved the problem of missed deaths that had been the surveillance system’s biggest flaw.

A recent report bringing together data from maternal mortality review committees in Colorado, Delaware, Georgia and Ohio showed that without the checkbox, substantial numbers of pregnancy-related fatalities would have gone uncounted; they included about half the deaths that occurred during pregnancy, 11 percent of deaths that occurred within 42 days after childbirth, and 8 percent of deaths that occurred 43 days to one year after childbirth.

But the checkbox has brought its own set of problems.

Researchers are finding the checkbox produces so-called false positives — deaths categorized as related to pregnancy or childbirth that aren’t. The data from Colorado, Delaware, Georgia and Ohio showed no evidence of pregnancy within the prior year for about 15 percent of the women — or 97 out of 650 — whose deaths were identified as related by the checkbox.

A separate study by MacDorman, Declercq and a third researcher found that “implausibly high” numbers of maternal deaths identified via the checkbox involve women 40 or older or attributed to less specific causes, i.e., other obstetric complications, or obstetric death of unspecified cause. Based on past patterns, it’s likely at least some of these deaths were not pregnancy-related, the study concluded.

The study offered several suggestions for shoring up the data, from training for those who complete death certificates to rechecking certain death certificates to make sure older women or women whose deaths were coded to non-specific causes really had been pregnant. So far, however, these steps haven’t been taken.

The federal government stopped publishing official maternal mortality numbers starting in 2007 to wait for all states to adopt the checkbox and produce consistent data.

But even now, with virtually all jurisdictions on board, doubts about the reliability of data based solely on the checkbox have made health officials hesitant to start publishing again, MacDorman said. The longer this continues, the more problems with the data will become entrenched, she warned.

“If you have a data system and you’re not publishing the data, you’re not monitoring it and cleaning it and working with it,” she noted. “Data quality’s going to go down.”

---

MacDorman’s study also offered another suggestion for improving data on maternal deaths: Using information collected by state maternal mortality review committees to validate and improve information gleaned from death certificates.

Review committees draw not only from vital records, but from a variety of sources, from obituaries to social media, to identify and compile facts about maternal deaths. For each case, they typically extract demographic data such as a woman’s age and race, then use medical records, notes from care providers and other resources to look at what happened, determining whether the catastrophic outcome was related to pregnancy or childbirth and whether it was preventable.

The most active committees play crucial roles in uncovering the truth about maternal deaths — both their number and their nature.

California’s committee, the California Pregnancy-Associated Mortality Review, was established in 2006, at the crest of a 15-year increase in the state’s maternal mortality rates. “It gave us data and it gave us stories,” said Dr. Elliott Main, whose group, the California Maternal Quality Care Collaborative, partners with the state Department of Public Health on the review.

Main said the ability to turn insights from data into prevention has helped California engineer a dramatic turnaround. The state has designed hospital toolkits and safety bundles based on preventable causes of death and mistakes in care identified by the review. Since 2006, California’s maternal mortality rate dropped by more than 55 percent.

Stacie Geller has been a member of the Illinois Maternal Mortality Review Committee since it began in 2001. After the panel determined the No. 1 cause of preventable maternal death in Illinois was obstetrical hemorrhage, the state spent $1 million retraining hospitals on how to deal with it. Within a couple of years, Geller said, clinical practices improved.

Despite review committees’ clear benefits, however, a fluctuating, but substantial number of states lack review processes.

Indiana, Oregon and Alabama don’t have them. Pennsylvania, the sixth most populous state, has never had one, though it has a working group studying the idea and the city of Philadelphia has a longstanding review panel. Loren Robinson, Pennsylvania’s deputy secretary for health promotion and disease prevention, conceded it was “hard to know why it had taken so long” for the state to move forward.

Several large states established committees only recently, prodded by poor maternal outcomes.

In Georgia — where the maternal mortality rate routinely ranks among the country’s highest — the review panel began meeting in 2012 and produced its first report in mid-2015. Texas launched its Maternal Mortality and Morbidity Task Force in 2013, as the state experienced an apparent surge in maternal deaths and near-deaths.

Most experts blame review committees’ struggles primarily on a dearth of resources, saying they often lose out to other health care priorities perceived as more pressing.

They are hardly pricey items by government standards. Committee members — typically health practitioners who specialize in maternal care — volunteer their time, sometimes with part-time support from staff at the state health department.

Still, an assessment completed in 2012 by The Association of Maternal & Child Health Programs found that at least one-third of states that had maternal mortality review committees had no budgets for these efforts, making them extremely fragile.

Geller said she and the other members of Illinois’ panel had subsidized its work with “donated time,” and that, by providing the state with much of its most meaningful data, they’ve, in effect, “done the state’s work” for a pittance.

She blames the lack of funding in part on this not being a “sexy” or “innovative” area for research.

“Fixing maternal health is pretty basic,” she said.

There’s a considerably stronger record of sustaining review processes related to child deaths, which all 50 states and the District of Columbia have. While child deaths far outnumber maternal ones, experts see something else behind the differences.

“It’s a reflection that for decades, it’s really the infant child that has been the focus and priority, and much less so the mother,” said David Goodman, who oversees maternal health efforts at the CDC’s maternal and infant health branch.

Other obstacles, too, hinder maternal mortality reviews. Since committees need confidential medical information to do their work, states trying to set them up often have to pass laws to assuage concerns about patient privacy and liability for hospitals and doctors. This can take years.

Goodman is leading a new collaborative effort by the CDC, its foundation and AMCHP to build a less piecemeal review system.

They’ve created a tool, the Maternal Mortality Review Information Application, to collect and analyze standardized data from review panels, the first-ever attempt to pool such information nationally. So far, about a dozen states are feeding into it. Goodman’s crew also has designed a web portal to help states and local jurisdiction start or improve maternal mortality reviews.

Few would argue this isn’t essential public health work, but the project receives not one penny of public money. Instead, it’s paid for by Merck for Mothers, which has poured in $2.3 million so far.

Asked if this was surprising, Goodman allowed that “I’ve been struggling with this a little bit myself.” In the end, though, he said he doesn’t parse the source of the money — he just appreciates the accelerated progress it allows.

“I’m very pragmatic, and in public health, you have priorities that emerge and go, and you have those you can act on and those where the timing isn’t right,” he said. “It doesn’t actually make me crazy because I understand that, and I’ve understood for all the time I’ve been working on this that it was important to keep going.”

There were times when he doubted they’d ever get this far. He sees Merck for Mothers’ support as a sign that the moment for improving U.S. maternal mortality data has arrived.

“Every little step you take is a little step forward and you’re keeping the conversation alive for when that opportunity does come up, and I think it’s now that the timing seems to be right,” he said.

Many in the field are eager to believe that.

Bette Begleiter, a member of Philadelphia’s review panel and an official with the city’s Maternity Care Coalition, said the struggle to get reliable data and sustain committees in the face of a rising death rate was a “national disgrace.” As she sees it, one test of whether things are improving will be whether Pennsylvania goes ahead and actually creates and funds a state panel.

“Everyone we talk to supports paying attention to maternal deaths,” Begleiter said. “But there’s been a disconnect between support and action. We need to see action.”