A repository of U.S. maternal mortality data from the Centers for Disease Control and Prevention’s National Center for Health Statistics, from January 2018 through December 2022, encompassing 6,193 pregnancy-related deaths; and a selection of state data, from Michigan, Minnesota and North Carolina, about maternal deaths that occurred in calendar years 2020, 2021 and provisional data from 2022, comprising 607 pregnancy-associated and pregnancy-related deaths.
Since the COVID-19 pandemic began in 2020, MuckRock and Columbia University's Brown Institue for Media Innovation have collected death certificate data from a variety of sources, including the CDC and a selection of states and large counties, through data-use agreements with government agencies and open-records requests. Given the spike in maternal deaths in 2020 and 2021, we began seeking out additional data sources to help explain the reasons behind the increase.
What we found: COVID-19 severely worsened the health of new mothers in U.S., which already had the worst maternal mortality rate of any Western country. And post-COVID, things are only getting worse. Despite the availability of vaccines and a nationwide push to bolster the health of new and recent mothers, the rate of maternal death in 2022 is already higher than pre-pandemic 2019, based on early, provisional data.
It bears repeating: Multiple studies from both the CDC and state maternal mortality review committees have found that anywhere between 80% and all pregnancy-related deaths are preventable.
Below, you'll find a methodology for the datasets in this repository, including their utility, limitations and caveats. We'll be refreshing this repository on a monthly basis. All updates to the data will be noted and appended to the bottom of this readme.
The data in this repository was requested, collected and analyzed by the following data scientists and journalists: Karen Wang, Columbia University's Brown Institute for Media Innovation and MuckRock; Dillon Bergin, MuckRock; and Derek Kravitz, MuckRock.
Each time we refresh the data with new provisional or final numbers, from the CDC or state vital records offices, we will update it here.
March 8, 2023: We have updated the CDC maternal mortality rate by state .csv file to include final live birth data from 2021 from the CDC. This update is also reflected in the 42-day and 1-year maternal mortality rates, which went unchanged from provisional to final.
March 10, 2023: We have updated the Michigan, Minnesota and North Carolina pregnancy-related and pregnancy-associated death counts, to include a subset of deaths that occurred between the ages of 15 through 44. These slightly smaller counts increase the likelihood that each death is related or associated with complications due to pregnancy and decreases the likelihood of a coding error on the part of state or federal agencies.
March 10, 2023: We have included a statement from the CDC's National Center for Health Statistics that addresses discrepancies in the number of pregnancy-related deaths in North Carolina, between the CDC and state vital records counts.
March 14, 2023: We added columns showing raw death counts and live births in cdc_maternal_mortality_rate_by_race.csv, cdc_maternal_mortality_rate_by_state.csv, and cdc_maternal_mortality_rate_ethnicity.csv files to show how rates were calculated. We also updated the multiracial rate for 2021 since it was previously suppressed by CDC due to low count.
March 15, 2023: The CDC has finalized its 2021 maternal mortality rates and counts and rates were unchanged from this repository's original release. All references to 2021 data being provisional have now been changed to final.
| term | description |
|---|---|
| pregnancy-associated death | a death that occurred during pregnancy or within one year postpartum but not directly tied to the pregnancy, such as a postpartum death in a car accident |
| pregnancy-related death | a death found to be related to health problems from the pregnancy, such as obstetric hemorrhage or pre-eclampsia, among other causes. |
| underlying cause of death | the one disease or injury that initiated the chain of events leading directly to death, sometimes referred to as the lead cause of death |
| immediate or multiple cause of death | other factors that contributed to someone's death, with up to 20 being included on a death certificate |
| ICD-10 | the International Classification of Diseases, 10th revision, used by both the Centers for Disease Control and Prevention and the World Health Organization to define causes of death |
| pregnancy checkbox | widely adopted by states for death certificates starting in 2017, there are five checkboxes to indicate pregnancy or a recent pregnancy, or if it's unknown by the person certifying the document |
The CDC data relies on the 27 underlying cause of death, or ICD-10, codes defined as pregnancy-related. The latest refresh of the CDC encompasses deaths occurring through January 21, 2023, as of February 5, 2023. But there is a lag in reporting deaths, which can take as long as six months depending on the jurisdiction.
MuckRock signed a data-use agreement with the CDC's National Center for Health Statistics in 2021 to access mortality data. Since then, the CDC has updated its WONDER query tool to allow the public and researchers to access new mortality data on a monthly basis. Together, with the data-use agreement data used as part of our Uncounted death certificate project, and the WONDER query tool, we routinely pull four sets of data about maternal mortality in the U.S.
The CDC maternal mortality death codes are the following:
• A34 (Obstetrical tetanus)
• O00 through O99 (Pregnancy, childbirth and the puerperium)
Relying on these 27 cause of death codes, we have compiled four sets of maternal mortality data. They are described below:
• Maternal mortality count and rate, by quarter and year: Through the third quarter of 2022, this maternal mortality count and rate is based on nationwide totals and either an actual or estimated number and rate of live births by the CDC's National Center for Health Statistics. All 2022 data is still provisional and subject to change, as jurisdictions continue to process the previoys years death certificates into 2023. This dataset only includes women aged 15 to 44 years old, to remain consistent with 2022 birth estimates.
• Maternal mortality count and rate, by race: Through the fourth quarter of 2021, this maternal mortality count and rate is based on nationwide totals of non-Hispanic white, Black, Asian and Native deaths, as defined by the CDC. This dataset includes women of all ages.
• Maternal mortality count and rate, by ethnicity: Through the fourth quarter of 2021, this maternal mortality count and rate is based on nationwide totals of non-Hispanic and Hispanic deaths, as defined by the CDC. This dataset includes women of all ages.
• Maternal mortality count and rate, by state: Through the fourth quarter of 2021, this maternal mortality count and rate is based on nationwide totals of deaths in 24 states where there have been at least nine pregnancy-related deaths each year between 2018 and 2021. This dataset includes women of all ages.
Below is the suggested citation for the CDC data:
Centers for Disease Control and Prevention, National Center for Health Statistics. National Vital Statistics System, Provisional Mortality on CDC WONDER Online Database. Data are from the final Multiple Cause of Death Files, 2018-2021, and from provisional data for years 2022, as compiled from data provided by the 57 vital statistics jurisdictions through the Vital Statistics Cooperative Program.
The CDC's maternal mortality data relies exclusively on death certificates. If those death certificates are lacking or omit key information, as we have found as part of our Uncounted project, it skews the overall picture, especially for causes that are comparatively few in number like a pregnancy-related death.
The CDC has told us this data is reviewed and audited on a monthly basis, and checked against what is described in the pregnancy checkbox. But because this data only looks at pregnancy-related, and not pregnancy-associated deaths, cases where a pregnancy played a key role in a death, such as an accidental drug overdose or a suicide, might not be counted in the CDC data.
There are also discrepancies when comparing the CDC death certificate data and state data from Michigan, Minnesota and North Carolina, even when looking at an apples-to-apples comparison such as a pregnancy-related death. For example, the CDC reports 67 pregnancy-related deaths in North Carolina in 2020 and 83 for 2021. But North Carolina reports slightly lower counts, of 62 in 2020 and 82 for 2021. The reasons for the differences differ but, in North Carolina, are tied to whether the deaths were North Carolina residents or residents of other states that happened to occur in North Carolina, according to the National Center for Health Statistics.
State-coordinated maternal mortality review committees, or MMRCs, have found that the true number of pregnancy-related deaths is roughly 30% to 40% higher than the CDC figures.
MuckRock chose to analyze death certificate data from these states for two reasons:
- Michigan, Minnesota and North Carolina provided exhaustive death certificate data encompassing all deaths in a given time period, regardless of whether the death had been reviewed by a medical examiner or coroner or occurred in an inpatient or outpatient setting.
- Data from these states included either pregnancy checkbox information or underlying and multiple, or immediate, cause of death codes, allowing for further analysis on the number of pregnancy-related and pregnancy-associated deaths.
As a result, similar death certificate data from a state like New Mexico, which is exhaustive but does not include pregnancy checkbox information nor underlying and multiple cause of death codes, proved insufficient for our analyses.
Other datasets from major metro areas, including Chicago’s Cook County and California’s Los Angeles, San Diego and San Francisco counties, only include deaths reviewed or certified by medical examiners or affiliated entities, making a complete analysis of maternal deaths in a given jurisdiction impossible.
The original sources of the state data are as follows:
• Michigan Department of Health and Human Services, Division for Vital Records and Health Statistics
• Minnesota Department of Health, Office of Vital Records
• North Carolina Department of Health and Human Services, Division of Public Health, State Center for Health Statistics
We chose to remove most personally-identifying information, or PII, from these records in order to protect the privacy of decedents, next of kin, certifiers and funeral homes while still preserving the fields that help explain the demographic profiles of those that have died, including race, ethnicity, age, resident county, death year, manner of death and specific causes of death.
Despite these redactions, it is possible to search for public obituaries using the provided data and, in some cases, find the identity of those that have died. This is true with any death certificate data available for public inspection and release and, in its raw form, this data is a public record, according to state statutes.
For more information about MuckRock’s editorial and ethics policies, you can read it here: https://www.muckrock.com/news/editorial-policy/
If you have questions or concerns about our release of these public records or data or questions about our methodology for excluding most personally-identifying information from these records, you can email us at news@muckrock.com.
To interpret underlying or multiple cause of death codes found in this data, there are several different lookup tools available. Here are three of the most frequently-used options:
The American Medical Association provides an easy-to-navigate codes lookup tool called Codify, available here: https://www.aapc.com/codes/icd-10-codes-range/
The World Health Organization has a similar tool, which provides detail on the conditions and causes of certain codes, available here: https://icd.who.int/browse10/2019/en#/
The U.S. Centers for Medicaid & Medicare Services has .xls files for billable and non-billable ICD-10 codes on its website. Please note that there are differences in the formatting of codes in these files and to properly "join" the data, you would need to combine the billable and non-billable files: https://www.cms.gov/medicare/coordination-benefits-recovery-overview/icd-code-lists
2022 data for Minnesota and North Carolina is provisional, as of March 2023. It will be updated and made final in mid-2023, and file names will be changed to reflect the shift from provisional to final. As a result, provisional data should not be used to analyze year-over-year trends in these states. Rather, it can be used to evaluate "point-in-time" counts ahead of the data becoming final, which is typically at least six months after the end of the calendar year.
The Minnesota and North Carolina death certificate data includes pregnancy checkbox details while the Michigan data does not. That means the Minnesota and North Carolina data allows us to capture deaths during and shortly after pregnancy that are either “pregnancy-associated” or “pregnancy-related."
The Michigan data does not allow us to look at the broader number of “pregnancy-associated” deaths and, as a result, is more limited.
Below are other state-level details, caveats and data dictionary information for Michigan, Minnesota and North Carolina.
The Michigan dataset includes 27 fields, all of which are found in the original file provided by the Michigan Department of Health and Human Services through a data-use agreement we signed for de-identified data in June 2022. The two provided years are final 2020 and 2021 data.
A data dictionary of all of the Michigan death certificate data is available in this repository, with the file, "mi_maternal_deaths_data_dictionary.xlsx."
The Michigan data does not include pregnancy checkbox information but it does include both underlying cause of death and multiple cause of death codes.
The following counts of pregnancy-related deaths in Michigan by year, are as follows, with a separate figure provided for the ages of 15 through 44:
• 2020: 40 *(39 between the ages of 15 through 44)
• 2021: 35 *(34 between the ages of 15 through 44)
Depending on the year, the Minnesota dataset includes either 113 or 114 fields, all of which are found in the original file provided by the Minnesota Department of Health's Office of Vital Records. The 2020 Minnesota dataset does not include a pregnancy checkbox code but does include the pregnancy checkbox text. The 2021 and 2022 Minnesota datasets include both the pregnancy checkbox code and text.
A data dictionary for Minnesota data in its original form is available in this repository — mn_maternal_mortality_data_dictionary.xls — and fields highlighted in yellow have been excluded from this public release in order to preserve identification of the decedent or next of kin.
The data is sorted into final year 2020, final year 2021 and provisional year 2022, with a filter on deaths that have been indicated as recently (within 1 year) or currently pregnant at the time of death. The following counts of pregnancy checkbox deaths in Minnesota, oftentimes referred to by researchers as pregnancy-associated deaths, are as follows, with a separate figure provided for the ages of 15 through 44:
• 2020: 33 *(31 between the ages of 15 through 44)
• 2021: 38 *(38 between the ages of 15 through 44)
• 2022 (as of March 2023, provisional): 18 *(18 between the ages of 15 through 44)
Within the Minnesota dataset are pregnancy checkbox details — a one-digit code and full text terms used to describe the type of pregnancy at the time of decedent’s death. They are as follows:
• “2” = “Pregnant at Time of Death”
• “3” = “Not Pregnant, But Pregnant Within 42 Days of Death”
• “4” = “Not Pregnant, But Pregnant 43 Days to 1 Year Before Death”
In two cases, specific place of death information in the “place_of_death_type” field was edited to remove the named roadways and were replaced with the term: “roadway.”
In two cases from 2020, “unknown” pregnancy checkbox deaths were included in our analysis because their underlying and immediate cause of death codes were one of the 27 maternal mortality-related codes defined by the CDC. They are indicated as “unknown” in the field “decedent_pregnancy_description.”
The North Carolina dataset includes 52 fields, 42 of which were provided in the original file by the North Carolina Department of Health and Human Services, Division of Public Health, State Center for Health Statistics. The data also includes 10 fields not found in the original file that have been created by MuckRock for codes related to multiple, or immediate, causes of death.
The 10 new codes are labeled rac1 through rac10.
A data dictionary for North Carolina data in its original form is available in this repository — nc_maternal_mortality_data_dictionary.xls — and fields highlighted in yellow have been excluded from this public release in order to preserve identification of the decedent or next of kin.
The data is sorted into final year 2020, final year 2021 and provisional year 2022, with a filter on deaths that have been indicated as recently (within 1 year) or currently pregnant at the time of death. The following counts of pregnancy checkbox deaths in North Carolina, oftentimes referred to by researchers as pregnancy-associated deaths, are as follows:
• 2020: 147 *(138 between the ages of 15 through 44)
• 2021: 174 *(138 between the ages of 15 through 44)
• 2022 (as of March 2023, provisional): 122 *(110 between the ages of 15 through 44)
Within the pregnancy checkbox is the following one-digit code and what it refers to:
• “2” = “Pregnant at Time of Death”
• “3” = “Not Pregnant, But Pregnant Within 42 Days of Death”
• “4” = “Not Pregnant, But Pregnant 43 Days to 1 Year Before Death”
We work to have our reporting and analysis available to a wide audience, often through distribution partnerships and syndication. If you are interested in republishing or adapting our work and it does not explicitly note that it is allowable for republishing, or if you would like clarifications about restrictions or to learn more about the underlying information, please email us at news@muckrock.com.
Nina Martin of the Center for Investigative Reporting/Reveal worked with MuckRock throughout its data collection and reporting process. For the Minnesota data, MaryJo Webster of the Minneapolis Star-Tribune provided both data and guidance. For the North Carolina data, Tyler Dukes of the Raleigh News & Observer provided code that helped us process the data in a readable format.
For the CDC data, the following experts provided subject-matter expertise, data and context for our analyses:
• Sungsik Hwang, Graduate Student, Center for Demography and Ecology, University of Wisconsin, Madison
• Dr. Jenna Nobles, Professor, Department of Sociology, Center for Demography and Ecology, University of Wisconsin, Madison
• Dr. Andrew C. Stokes, Assistant Professor, Departments of Sociology and Global Health, Boston University School of Public Health
• Dr. Marie Thoma, Associate Professor, Department of Family Science, University of Maryland School of Public Health
We ask that all republication and citation of our materials include a note that the reporting was originally published by MuckRock and include a link back to the original version. When republishing, this credit should be included at the top of the piece.
Sample language: This story was originally published by MuckRock, a nonprofit journalism organization.
We publish certain stories, data and other assets under a Creative Commons license and most code and data analysis scripts under an open source license. You are welcome and encouraged to reuse these under the terms of the given license, but please ensure that you understand any restrictions. For example, certain Creative Commons licenses allow reuse but only if you do not edit the underlying material.
Sample language: This story was originally published by MuckRock, a nonprofit journalism organization. It is republished under a Creative Commons (BY-ND 4.0) license.
