BACKGROUND: Death certificates are the primary source of case identification for maternal mortality surveillance, but can fail to identify pregnancy in cases of death due to non-obstetric causes. The 2003 National Center for Health Statistics (NCHS) revision of the standard death certificate added a pregnancy status checkbox to improve reporting of deaths occurring during pregnancy, or within one year of the end of pregnancy. The Massachusetts Department of Public Health (MDPH) has conducted enhanced maternal mortality surveillance using the linkage of vital records since 1997, but did not collect pregnancy status on death certificates until implementing the 2003 NCHS revision in September 2014. We compared deaths identified using record linkage with deaths identified using the new pregnancy status checkbox to evaluate its potential role in improving maternal mortality surveillance.
METHODS: We linked death certificates of Massachusetts resident women who died in 2015 with birth and fetal death files from 2014 and 2015 using a multistep deterministic linkage algorithm. All female decedents aged 5–75 years are assigned a pregnancy status: not pregnant within past year, pregnant at time of death, pregnant within 42 days of death, pregnant 43 days to 1 year before death, and unknown if pregnant within the past year. Agreement between pregnancy-associated deaths identified by the pregnancy checkbox and those identified by record linkage was calculated using kappa coefficients. Fisher’s exact test was used to identify differences in the characteristics of death certificates identified by each method.
RESULTS: A total of 38 death certificates were identified where the decedent was pregnant at the time of death or within the year prior to death: 19 identified by the pregnancy checkbox alone, 14 identified by record linkage alone, and 5 identified by both methods (κ=-0.452). The majority of deaths identified by the pregnancy checkbox occurred during or within 42 days of pregnancy (71%), while the majority of deaths identified by the linkage occurred at least 43 days after pregnancy (95%). There were no other statistically significant differences in death certificate variables found between the deaths identified by each method.
CONCLUSIONS: The pregnancy checkbox identified 19 potential pregnancy-associated deaths that were not identified through the linkage of vital records, which would have a large impact on the Massachusetts pregnancy-associated mortality ratio. Investigation of these deaths will confirm whether they are pregnancy-associated and reveal amongst which populations of women the pregnancy checkbox can improve maternal mortality surveillance in Massachusetts.