Identifying Deaths During and After Pregnancy: New Approaches to a Perennial Challenge

Maternal mortality is increasingly recognized as a public health crisis in the United States, with growing attention on the 3 major systems of national surveillance. National maternal mortality statistics are reported by the Centers for Disease Control and Prevention’s (CDC’s) National Vital Statistics System (NVSS) ¹ within the National Center for Health Statistics (NCHS) and the Pregnancy Mortality Surveillance System (PMSS) ² within the Division of Reproductive Health (DRH). Comprehensive surveillance data at the state and local level are also available from maternal mortality review committees (MMRCs) via the Maternal Mortality Review Information Application (MMRIA). ³ Each surveillance system has inherent strengths and limitations, but all rely on valid, accurate, and timely case identification via vital statistics data. We outline recent innovations to improve identification and ensure robust and accurate surveillance of maternal mortality in the United States.

For all 3 systems, maternal mortality statistics rely on death certificates, commonly referred to as death records, which are filed in states for all US deaths. ⁴ Ascertaining if a death occurred as a result of pregnancy complications requires information from multiple fields in the medical section of death records, including the pregnancy checkbox and the descriptive causes of death. This section is completed by death certifiers—typically physicians, medical examiners, or coroners—and varies in accuracy and completeness. The pregnancy checkbox was introduced with the 2003 US Standard Certificate of Death and allows the certifier to indicate whether the decedent was pregnant at the time of death; not pregnant, but pregnant within 42 days of death; not pregnant, but pregnant 43 days to 1 year before death; not pregnant within past year; or had an unknown pregnancy status. ⁵

While the pregnancy checkbox improved identification of deaths during or within 1 year of pregnancy (pregnancy-associated deaths), challenges persist. Several analyses concluded that the pregnancy checkbox played a role in explaining temporal increases in maternal mortality rates, especially for women aged ≥40 years,^6-8 and variation across states in implementation of the 2003 US Standard Certificate of Death led to gaps in national reporting. ⁶ In addition, a 2016 multistate quality assurance pilot of the pregnancy checkbox found that 21% of identified pregnancy-associated deaths were false positives. ⁹

To supplement the pregnancy checkbox information, PMSS and MMRCs incorporate a linkage between death records and birth and fetal death records into their identification. While DRH provides guidance on pregnancy-associated death identification, linkage and identification have historically been implemented by the jurisdiction’s vital records office or maternal mortality review program, and variation exists across states. In addition, a 2- to 3-year delay occurs in reporting maternal mortality statistics using PMSS data because of the time it takes for states to conduct linkages and then voluntarily share identified deaths with DRH.

States and CDC recognize the need to improve maternal mortality surveillance through enhanced case identification. DRH has collaborated with jurisdictional partners on 2 initiatives to improve identification of pregnancy-associated deaths. In 2018, DRH established the Pregnancy-Associated Death Identification Workgroup (hereinafter, Workgroup) to develop enhanced identification methods and best practices to standardize and expand identification methods across jurisdictions. Separate from the Workgroup, DRH initiated the State Vitals & PMSS Data Integration pilot in 2019 in collaboration with the National Association for Public Health Statistics and Information Systems (NAPHSIS). Through this pilot, DRH receives vital records data from states in real time to perform pregnancy-associated death identification. These 2 initiatives aim to improve the timeliness, validity, and completeness of identification for robust and accurate surveillance of maternal mortality in the United States (Table 1).

Table 1.

Maternal mortality surveillance systems in the United States

System	Organization	Strengths	Limitations	Recent innovations
National Vital Statistics System 1	National Center for Health Statistics, CDC	Strongest source of data for international comparisons and historical data, dating to 1900	• Excludes deaths 43-365 days after pregnancy from the maternal mortality rate • Constrained to data from death certificates • Lacks detail to inform prevention efforts • Most susceptible to death certificate data quality issues, including false-positive and false-negative cases and incomplete or inaccurate reporting by death certifiers	Rigorous evaluation of pregnancy checkbox, data quality, and misclassification errors led to new coding procedures implemented in 2018 reporting
Pregnancy Mortality Surveillance System (PMSS) 2	Division of Reproductive Health, CDC	• Includes deaths during pregnancy or within 1 year of pregnancy from every state and the District of Columbia • Most clinically relevant national measure of maternal mortality since 1987 • Vital records linkage and medical epidemiologist review reduce susceptibility to death certificate data quality issues including false-positive and false-negative cases and incomplete reporting by death certifiers	• Constrained to data from death certificates and linked birth or fetal death certificates • Lack detail to inform prevention efforts • Historical 2- to 3-year delay in reporting because of time requirements for linkages and voluntary reporting by jurisdictions • Historical variation in identification methods across jurisdictions	The State Vital Records and PMSS Integration Pilot has decreased time between date of death and date of identification and standardized identification methods across participating states
Maternal Mortality Review Information Application (MMRIA) 3	Division of Reproductive Health, CDC (aggregate national data), maternal mortality review committees (jurisdiction-specific data)	• Includes deaths during pregnancy or within 1 year of pregnancy • Incorporates medical records, nonmedical records, autopsies, informant interviews, community-level health indicators, and other sources of information to provide much more detail than vital records alone • Details contributing factors and recommendations for prevention, informing actions to prevent future deaths	• Most resource-intensive and least timely • As of February 2022, several states do not yet have active maternal mortality review committees • Historical variation in identification methods across jurisdictions	• The Pregnancy-Associated Death Identification Workgroup promotes standardized and timely pregnancy-associated death identification across jurisdictions • Automatic upload of vital records to MMRIA decreases data entry time in pilot jurisdictions

Overview of Maternal Mortality Surveillance Systems in the United States

National Vital Statistics System

NVSS reports on maternal deaths using death certificates submitted from jurisdictions’ vital registrars. ¹⁰ Maternal deaths are defined as deaths “while pregnant or within 42 days of termination of pregnancy, irrespective of the duration and site of the pregnancy, from any cause related to or aggravated by the pregnancy or its management but not from accidental or incidental causes.” ⁶

NVSS uses International Classification of Diseases, Tenth Revision (ICD-10) codes to assign cause of death codes to all death records. ¹¹ If the pregnancy checkbox, descriptive cause of death, or medical conditions sections of the death record indicate pregnancy, a maternal cause of death code (A34, O00-O99) is applied. ¹⁰ To minimize potential misclassification and limit false positives, use of the pregnancy checkbox for cause of death coding is limited to decedents aged 10-44 years. ¹⁰ Records with cause of death codes indicating pregnancy at the time of death or within 42 days of death are defined as maternal deaths and included in the maternal mortality rate, defined as the number of maternal deaths per 100,000 live births. ⁶ Because it relies solely on death records, NVSS can produce the maternal mortality rate relatively quickly; as of July 2022, the most recent summarized data available through NVSS covered deaths that occurred in 2020. However, its accuracy depends on the accuracy of the death certificates submitted.

Pregnancy Mortality Surveillance System

DRH initiated national surveillance of pregnancy-related deaths through PMSS in 1986 to fill clinical gaps in understanding that cannot be gleaned through NVSS’s statistical accounting system. ² PMSS captures pregnancy-associated deaths and, through clinical review by medical epidemiologists, classifies a subset of these deaths as pregnancy-related, defined as “deaths occurring during pregnancy or within one year of pregnancy, regardless of the duration and site of pregnancy, from a pregnancy complication, a chain of events initiated by pregnancy, or the aggravation of an unrelated condition by the physiologic effects of pregnancy.” ² PMSS then reports the pregnancy-related mortality ratio, the number of pregnancy-related deaths per 100 000 live births.

Annually, jurisdictions voluntarily submit death records of women who died during or up to 1 year after the end of a pregnancy, along with linked birth and fetal death records where applicable. Because PMSS incorporates both the linkage of death records to birth and fetal death records and review by medical epidemiologists, PMSS relies less than NVSS on the validity of the pregnancy checkbox. ² While this clinical review enhances the classification of pregnancy-related deaths, PMSS still depends on the quality of information in vital records data, and manual processes of linkage and data submission limit its timeliness. As of July 2022, the most recent data available from PMSS covered deaths that occurred in 2018.

Maternal Mortality Review Committees

MMRCs are multidisciplinary bodies that convene at the state or local level to comprehensively review deaths of people during or within 1 year of pregnancy. ¹² CDC directly funds 30 jurisdictions, which support review programs in 31 states, and supports all 50 MMRCs through the Enhancing Reviews and Surveillance to Eliminate Maternal Mortality initiative. ³

Like PMSS, MMRC case identification begins with death records linked to birth and fetal death records, where applicable, and identification of death records with a pregnancy checkbox indicating pregnancy at the time of death or within 1 year of death and/or a maternal cause of death code. Many states also incorporate a search of the literal cause of death fields for terms indicating pregnancy. Several MMRC reports document the benefit of vital records linkages to identify pregnancy-associated deaths that would be missed if the pregnancy checkbox were used alone.^13-16

In addition to death records linked to birth and fetal death records, MMRCs have access to medical, social service, and autopsy records, and in some jurisdictions even informant interviews, to understand the clinical and nonclinical contributors to deaths. MMRCs use abstractors, usually individuals with obstetric nursing experience, to review and abstract all available information into MMRIA, a standardized data system developed by DRH in partnership with MMRCs. MMRCs review pregnancy-associated deaths to determine pregnancy-relatedness, the underlying cause of death, and preventability; identify contributing factors to the death; and make detailed recommendations to prevent similar deaths in the future.

With expanded access to medical and nonmedical records, most MMRCs review injury deaths, namely suicides and overdoses, which are excluded from the deaths reported by NVSS and PMSS.^1,2 A report using MMRIA data from 14 MMRCs found that mental health conditions, including unintentional overdoses/poisonings and suicides, are a leading cause of pregnancy-related death. ¹⁷ MMRCs’ enhanced data sources and expert review processes enable MMRIA to provide more detailed and actionable information than NVSS or PMSS, although the process is resource- and time-intensive. As of February 2022, forty-two jurisdictions were using MMRIA for state-specific analyses and sharing their MMRIA data with DRH for forthcoming aggregate analyses. MMRCs typically review deaths within 2 years of the date of death, and analysis and reporting follow. Resource constraints and process variations create substantial differences in the timeliness of jurisdictions’ reporting of MMRC data.

Because MMRCs are now well-established nationally, it is essential to standardize pregnancy-associated death identification across MMRCs to enable national understanding and catalyze actions to prevent future deaths.

Initiatives to Improve Identification of Pregnancy-Associated Deaths

Pregnancy-Associated Death Identification Workgroup

The Pregnancy-Associated Death Identification Workgroup consists of members from DRH and volunteers from 8 state health departments. Relying on their experience, Workgroup members developed best practices for pregnancy-associated death identification (Table 2), including details on linkage methods and inclusion criteria. ¹⁸ Deaths that occur during pregnancy or have an outcome other than live birth or fetal death, such as ectopic pregnancies or miscarriages, will not have birth or fetal death registrations to link. Knowing that vital records linkages alone may not identify such deaths, the Workgroup also recommended additional identification methods.

Table 2.

Methods for pregnancy-associated death identification recommended by the Centers for Disease Control and Prevention’s (CDC’s) Pregnancy-Associated Death Identification Workgroup, 2019 ^a

Recommended identification method	Known challenges with implementation
• Deterministic linkage between death records and birth and fetal death records using mother’s social security number (SSN)	• Some jurisdictions do not allow for the use of SSN in linkages. When an adoption is finalized, the original birth record is sealed, and the biological mother’s information can no longer be linked with death records.
• Probabilistic linkage between death records and birth and fetal death records	• Probabilistic linkage often requires a time-intensive manual review of potential matches, limiting how frequently linkages can reasonably be performed. When an adoption is finalized, the original birth record is sealed, and the biological mother’s information can no longer be linked with death records.
• Identification of death records with a pregnancy-related term in the descriptive cause of death fields	• False-positive deaths may be identified if timing information is not provided in the description. For example, the record may mention “postpartum cardiomyopathy” but not specify the timing of the pregnancy.
• Identification of death records with an ICD-10 cause of death code related to pregnancy (A34, O00-O99.9, excluding O97) 11	• Death records identified using only the ICD-10 cause of death codes and/or pregnancy checkbox require confirmation of pregnancy status because of possible checkbox error. Certifier confirmation is not always feasible, especially as time between death and confirmation increases.
• Identification of death records where the pregnancy checkbox field indicates the decedent was pregnant at the time of death, pregnant within 42 days, or pregnant 43 days to 1 year before death	• Death records identified using only the ICD-10 cause of death codes and/or pregnancy checkbox require confirmation of pregnancy status because of possible checkbox error. Certifier confirmation is not always feasible, especially as time between death and confirmation increases.

Abbreviation: ICD-10, International Classification of Diseases, Tenth Revision.

^aData source: Pregnancy-Associated Death Identification Workgroup. ¹⁸

Despite previously described limitations, the death record pregnancy checkbox and ICD-10 cause of death codes are important components of a comprehensive identification process. To reduce false positives, the Workgroup recommends confirming pregnancy status with the death certifier for all deaths identified based only on the checkbox and/or ICD-10 codes.^18,19 If certifier confirmation is not feasible, the Workgroup recommends searching obituaries, hospital and emergency department data, and media reports to confirm whether the decedent was in fact pregnant within 1 year of death. The best practice recommendations also include searching the death record literal cause of death fields for additional details and to capture deaths that may have been missed by other methods. The immediate, intermediate, and underlying causes, and the significant conditions contributing to the death, can all include terms related to pregnancy. The Workgroup identified a list of specific search words. ¹⁸

MMRCs are now implementing the Workgroup’s best practices, with technical assistance from DRH. These conversations resulted in increasing standardization across jurisdictions, although challenges with implementation still exist (Table 2). In addition, pregnancy-associated deaths can occur among transgender and nonbinary people, and identification methods may need to be adapted to remove the gender restriction to ensure these deaths are included in maternal mortality surveillance.

The Workgroup plans to expand best practices to include additional data sources beyond vital records. Hospital discharge data are used to identify severe maternal morbidity, ²⁰ and 42% of US births in 2018 were covered by Medicaid, ²¹ making both data sources potentially important for identifying deaths. Multiple states already incorporate hospital discharge data into their identification processes,^22-25 and some states’ abstractors incorporate Medicaid data into their medical record review. Timely access to hospital discharge and Medicaid data has been challenging for Workgroup states thus far, so it will be important to establish methods that ensure timely identification and reporting if these data sources are used.

State Vital Records and PMSS Data Integration Pilot

Currently, DRH does not ask jurisdictions to submit vital records for PMSS until they have a final annual death dataset from NCHS. To improve the timeliness of PMSS data and standardize identification across jurisdictions, CDC initiated a pilot in partnership with NAPHSIS and 6 states (Colorado, Kansas, Louisiana, Mississippi, Utah, Wyoming) to receive vital records data directly through the State and Territorial Exchange of Vital Events system operated by NAPHSIS. Through the pilot, DRH receives vital records data from states as they are transmitted to NCHS, typically daily. DRH then completes all identification steps involving vital records data, following the Workgroup’s best practices. This process allows for identification in almost real time; DRH had already identified deaths that occurred in 2021 for the 6 pilot states before requesting 2020 PMSS data from all other jurisdictions. In addition, DRH can use an automatic upload function in MMRIA to import the death, birth, and fetal death records for identified deaths, saving hours of data entry time and costs.

Timely identification enables CDC to support pilot states in verifying identified pregnancy-associated deaths. In a prior quality assurance initiative, among 212 deaths identified with only the pregnancy checkbox, 45.8% had no confirmed pregnancy and 16.3% had a pregnancy status that could not be determined. ⁹ For the NAPHSIS pilot, DRH is sharing any deaths identified by only the pregnancy checkbox, ICD-10–coded causes of death, and/or descriptive causes of death with the states. Each state then confirms the pregnancy status for these deaths through certifier confirmation or alternative methods suggested by the Workgroup (eg, obituaries, hospital discharge data). Shortening the time between identification and confirmation may improve response rates from certifiers and reduce recall bias. If errors in the pregnancy checkbox are identified, it may be easier to make corrections to the death records closer to the date of death. Promptly identifying and resolving false positives will improve the quality of maternal mortality surveillance by NVSS, PMSS, and MMRCs.

This pilot also ensures consistent identification methods across all participating states. Among 3 states, we identified 34 additional pregnancy-associated deaths for inclusion in the 2018 PMSS review. These deaths were not included in the voluntary state submission for PMSS, resulting in a 49% increase in identified deaths across these 3 states.

With a successful pilot of 6 states complete, CDC and NAPHSIS are now scaling the State Vital Records & PMSS Data Integration project. As of February 2022, twelve jurisdictions are participating and 11 jurisdictions are initiating data-sharing agreements, and the project aims to expand to all 50 states, Puerto Rico, and the District of Columbia. Once the project is fully scaled, the responsibility of identifying deaths through vital records can shift to DRH, and individual jurisdictions will no longer need to identify deaths for PMSS. This shift to DRH will also enable MMRCs to focus on pregnancy verification and applying the additional recommended methods of pregnancy-associated death identification. Ultimately, both identification and surveillance data will become more timely, comprehensive, and accurate within and across jurisdictions.

Conclusions

Awareness is growing that too many people die each year in the United States because of pregnancy complications, yet our understanding of these deaths has been limited by identification methods that rely on manual processes at the jurisdiction level and are not standardized across jurisdictions. Using innovative strategies and approaches, CDC, NAPHSIS, and jurisdiction-based vital records offices and MMRCs have improved the accuracy and speed of identifying deaths. With valid, accurate, and timely identification, robust and accurate surveillance is possible. The United States is entering a new stage in its efforts to determine leading causes of and contributors to pregnancy-related deaths and take action to prevent future deaths.

Disclaimer

The findings and conclusions in this article are those of the authors and do not necessarily represent the official position of the Centers for Disease Control and Prevention.