Can Data Modeling and Analytics Help Reduce Pregnancy-Related Deaths?

In the past two decades, the U.S. maternal mortality rate has doubled. That makes the United States the only country in the developed world to see an increase in women dying from pregnancy-related complications. And, the number of deaths is even higher for non-Hispanic black women—three to four times higher.

According to the Centers for Disease Control and Prevention (CDC), at least 66 percent of these deaths could have been prevented with timely and proper interventions. That's one reason why MITRE researchers began studying the statistics behind these alarming numbers.

A year ago, we began by concentrating on states that have had success in decreasing maternal mortality and identifying what worked for them. Could their approaches be applied in other states with equal success? How can states use data to decide which approaches would reduce risk for their populations? How can this data be used to influence funding decisions?

"Our team developed a prototype—a microsimulation/policy simulator—that can help predict the outcomes of applying various approaches within each state," says Rachel Mayer, the co-lead of the research project. "We call it the MITRE Maternal Mortality Interactive Dashboard or 3MID."

This tool can help determine quantitative correlations among demographic factors, such as race, age, socioeconomic status, and maternal mortality. The goal is to use 3MID to help determine what interventions may reduce the risk for a particular demographic, which can help states make data-driven funding decisions.

"For example, we've taken the successful approaches that California developed to significantly decrease its maternal mortality rate and built these approaches into 3MID to assess whether they would also help other states," Mayer explains.

California's Progress, the Benchmark

In 2006, the state created a program that reduced the pregnancy mortality rate by 55 percent over seven years. How? They created the California Maternal Quality Care Collaborative (CMQCC). This is a public-private, multi-stakeholder organization dedicated to resolving preventable maternal morbidity (illness), mortality, and racial disparities in California’s maternal care. The CMQCC focuses on three major components: research, Quality Improvement Toolkits, and the Maternal Data Center—an online web application that automatically produces performance metrics and data results on maternal care from participating hospitals.

The CMQCC first conducted an in-depth analysis of the causes and contributing factors of pregnancy-related deaths during pregnancy, or within one year of the end of pregnancy, to identify opportunities to enhance California’s maternal care.

Seven years of research led to the creation of Quality Improvement Toolkits, which are continually developed to address the leading causes of preventable pregnancy-related death. For example, the current toolkits are Early Elective Deliveries, Preeclampsia, Obstetric Hemorrhage, Vaginal Birth, Cardiovascular Disease, and Venous Thromboembolism. The toolkits outline best practice guidelines, literature, and tools for a wide variety of audiences, including clinicians, health educators, and hospital administrators.

By 2015, California had reduced its maternal mortality rate to 4.5 deaths per 100,000 live births, the lowest maternal mortality rate of any state. However, while these interventions reduced the rate overall, they did not reduce the mortality racial discrepancy. Non-Hispanic black women are still three to four times more likely to die from a pregnancy-related death than non-Hispanic white women (in California and across the country).

California continues to work on this issue, coming up with new toolkits to address the problem. One of these focuses on cardiovascular disease, which is the leading cause of pregnancy-related mortality within the United States, according to the CDC. Black women have almost a 10-fold higher risk of cardiovascular-disease-related death and a four-fold overall risk compared to any other racial groups.

Using Data to Make Decisions

Building on what California has learned, the MITRE team applied modeling and simulation to the maternal mortality problem across the country. It created a microsimulation program that policy makers can use to test assumptions and create forecasts that will help them prioritize resource allocations and optimize the impact of scarce resources on maternal mortality and health disparities.  

“Our research team theorized that the application of California’s maternal care programming to all 50 states could reduce the national maternal mortality rate,” Mayer says. “We predict California’s new cardiovascular disease interventions might lower the racial disparity between black and white women, and we will be watching to see if this happens and how similar interventions could lower the racial disparity between black and white women within each state.”

The 3MID is based on theoretical assumptions. The model applies California’s maternal care programming to eight within the United States and analyzes the effect of the programming on the maternal mortality rate within each state. It also shows that the maternal mortality racial disparity for black women still exists today despite the implementation of maternal care programming. (The team chose California, Florida, Georgia, Illinois, Michigan, New Jersey, New York, and Texas because these states had the most available and contiguous annual data from CDC and the most thorough data on maternal mortality for white and non-Hispanic black women.)

“Our model applies novel data analytic techniques to research the potential impact and scalability of evidence-based interventions on maternal mortality,” adds Alison Dingwall, who co-leads the research team.

“The 3MID is easy to use. We select a state and set parameters—for example, we would designate the level of healthcare funding per capita. Then we can adjust levers to see how the deployment of each of the six CMQCC quality toolkits would affect the maternal mortality rate in that state. And we can estimate lives saved and total program costs. We’ve created “what if?” scenarios to help users visualize how the different variables could interact with each other to reduce the rate.”

In testing the 3MID, the team applied the variables to a representative synthetic dataset of women of child-bearing age. Using demographic data and socioeconomic indicators rooted in available data on maternal mortality, they assessed the risk of maternal mortality for agents in the synthetic population. Understanding that morbidities affect women at different rates, the model shows that various interventions will have distinct impacts depending on the person’s demographics. While California is currently used as a proof of concept, factors such as cost of care and toolkit effectiveness are expected to vary in each state.

“The 3MID dashboard inputs provide a traceable methodology. Users can project the level of maternal mortality in future years, given the implementation of certain toolkits, and then compare that to empirical historical data,” Mayer adds.

“In the future, we hope to include data from each state,” she says. “One of the problems with using data to address this issue is that here is no centralized data source for information on maternal mortality. Each state collects different data, which is not always available.”

Other Factors

Our team has also captured information about some of the other contributing factors to maternal mortality, such as chronic diseases, access to care, inherent bias, and delayed diagnosis of treatment.

“For example, California is piloting a new project on improving racial equity through partnership and education, based on research that highlights the role structural racism and inherent bias play in the high maternal mortality rates for non-Hispanic black women,” explains Mayer. “The state wants to build tools and Quality Improvement strategies for hospitals, physicians, staff, and patients. We believe these factors must be considered when creating responsive programs, even if they can’t be easily represented statistically in a tool such as 3MID.”

The research team continues to improve the tool with the goal of getting the 3MID into the hands of policymakers.

“The 3MID will allow state decision-makers to make data-driven decisions to improve maternal care programming and allocation of resources,” adds Mayer. “We hope our work will allow MITRE to have a positive influence on the national quality improvement efforts to reduce maternal mortality.”

—by Beverly Wood

WATCH how MITRE is working to improve maternal mortality rates across the country.