Data were shared for aggregate analysis by jurisdictional MMRCs through the Maternal Mortality Review Information Application (MMRIA). MMRIA supports standardized record abstraction, case summary development, documentation of committee decisions, and analysis. Data analyzed included information on pregnancy-related deaths that occurred during 2017–2019 among residents of 36 states: Alabama (2017–2018), Alaska (2019), Arizona (2017–2019), Arkansas (2018–2019), California (2019), Colorado (2017–2019), Connecticut (2017–2019), Delaware (2017–2019), Florida (2017–2019), Georgia (2017–2018), Hawaii (2017–2018), Illinois (2017–2019), Indiana (2017–2019), Kansas (2017–2019), Louisiana (2017–2019), Massachusetts (2017), Minnesota (2017–2018), Mississippi (2017–2019), Missouri (2017–2019), North Carolina (2018–2019), Nebraska (2017–2019), New Hampshire (2017–2019), New Jersey (2017–2019), New Mexico (2017–2019), New York (2018–2019; 2019 excludes NYC), Ohio (2017–2018), Oklahoma (2017–2019), Oregon (2018–2019), Pennsylvania (2018), Tennessee (2017–2019), Texas (2019), Virginia (2018), Washington (2017–2019), West Virginia (2017–2019), Wisconsin (2017), and Wyoming (2018–2019). In some states, only partial years of data were shared. Some states group review of deaths by cause of death and may have only reviewed some causes before sharing data with CDC. Sensitivity analysis did not indicate any major differences in underlying causes of death when data for those states were excluded.
We used race and ethnicity data from the birth or fetal death records, when available, and from death records when a birth record or fetal death record was unavailable. Race and Hispanic origin are reported separately on the birth, fetal, and death records; more than one race can be selected. All deaths with a notation of Hispanic origin are classified as Hispanic. For deaths with missing notation of Hispanic origin, race and ethnicity of the decedent was classified as missing. For deaths with a notation of non-Hispanic origin but a notation of race was missing, the race or ethnicity of the decedent was classified as missing. For non-Hispanic persons, race was classified as: non-Hispanic single-race White, non-Hispanic single-race Black, non-Hispanic single-race American Indian or Alaska Native (AIAN), non-Hispanic single-race Asian, non-Hispanic single-race Native Hawaiian or Other Pacific Islander (NHOPI). For brevity, text and tables omit the term single-race. Text from the Other race field was not recoded. When Other race or more than one race were noted, race was classified as non-Hispanic other/multiple races. Age at death was based on information from the death record. We used education level from the birth or fetal death record, when available, and from death records when a birth or fetal death record was unavailable. If the death occurred during the postpartum period, and the death record indicated an education one level higher than the birth record, we relied on the higher education level from the death record. Geographic information is based on county of last residence from the death record, when available, and from the birth or fetal death record when missing or undetermined on the death record. Geographic classifications are those produced by Texas A&M Geoservices.
Timing of death in relation to pregnancy was assigned by using the number of days between the date of death and the end of pregnancy, as documented by the MMRC abstractor, or as calculated by using the number of days between the date of death on the death record and the date of birth or fetal death on the linked birth or fetal death record by CDC. MMRIA instructs MMRC abstractors to enter 0 number of days if the death occurs on the day of delivery. Deaths classified as occurring on the day of delivery occurred within 24 hours of the end of pregnancy. If the specific number of days was missing, deaths that the MMRC abstractor classified as pregnant at the time of death, or with the standard pregnancy checkbox on the death certificate marked as pregnant at the time of death, were classified as during pregnancy. We completed a manual review of narratives and abstracted records in MMRIA to confirm the classification of timing of deaths for all decedents, and recoded inaccurate timing of deaths. When available, the timing of death documented in the narrative was used to classify timing of deaths when deaths were missing a timing classification based on the number of days, abstractor-assigned category, and pregnancy checkbox. Deaths that did not have enough information to determine the exact timing of death were classified as unknown. Deaths without a narrative were classified as missing. Pregnancy-related deaths determined by the MMRCs to be suicides were assigned an underlying cause of death of mental health conditions during analysis, if not already assigned this cause of death by an MMRC. Deaths where the MMRCs determined the means of fatal injury to be overdose/poisoning, and where the MMRCs determined that substance use disorder contributed to the death, were assigned an underlying cause of death of mental health conditions during analysis, if not already assigned this cause of death by an MMRC.