To ascertain pregnancy-related deaths, the surveillance system requests copies of death certificates for deaths during or within 1 year of pregnancy from health departments in the 50 states, New York City, and the District of Columbia. Where applicable, copies of live birth or fetal death certificates or reports that were linked to the mother's death also are requested. A death is considered pregnancy-related if it occurred during or within 1 year of pregnancy and resulted from complications of pregnancy itself, a chain of events initiated by pregnancy, or aggravation of an unrelated event by the physiological effects of pregnancy.9 A death was classified as pregnancy-related after a review of all information on the death certificate and, when available, matched live birth and fetal death certificates or reports, maternal mortality review committee reports, and autopsy reports. Deaths in the surveillance system are not assigned ICD codes, and the assigned cause of death is not limited to the underlying cause listed in Part I or II of the death certificate. Deaths are classified in one of 10 cause-of-death groups (ie, hypertensive disorders of pregnancy, hemorrhage, infection, thrombotic pulmonary embolism, amniotic fluid embolism, cardiomyopathy, cardiovascular conditions, noncardiovascular medical conditions, cerebrovascular accident, and anesthesia complications) by clinically trained epidemiologists. The temporal relationship between pregnancy and death was established for deaths reported to the surveillance system from information in the pregnancy checkbox, the interval indicated for the underlying cause of death on the death certificate, and from matched live birth or fetal death certificates or reports.
We divided the current study into two time periods, 1999–2002 and 2003–2005, to separate the potential effect of the change from ICD-9 to ICD-10 from the implementation of the 2003 revision of the death certificate. We used published data for 1995–1997, coded under ICD-9, for comparison.8
After merging the surveillance system data with the vital statistics' mortality files by death certificate number, we classified pregnancy deaths into three groups: “both systems”—pregnancy-related deaths that matched a maternal or late maternal death; “surveillance only”—pregnancy-related deaths that matched a death in the vital statistics system with an ICD-10 code outside O00–O96 and O98–O99 (and therefore not classified as maternal or late maternal); and “vital statistics only”—deaths that were classified as maternal or late maternal but were not linked to any pregnancy-related deaths reported to the surveillance system. We excluded from the final analysis some maternal and late maternal deaths matched to a death that the surveillance system considered not pregnancy-related based on additional information that the death was either not causally or not temporally related to pregnancy.
For 1999–2002 and 2003–2005, we compared the distribution by cause of death for all deaths occurring during or within 1 year of pregnancy for deaths in “both systems” and “surveillance only” using the surveillance system assigned cause of death. Chi-square tests of the change in distribution of deaths in each cause group over time were performed for statistical significance (P<.05). Because we were not able to map some ICD-10 codes within Chapter O to a specific surveillance system cause-of-death group, it was not possible to analyze the maternal deaths in “vital statistics only” by the surveillance system cause-of-death groups.
For each of the two study periods, we calculated revised maternal mortality ratios and pregnancy-related mortality ratios (deaths per 100,000 live births) based on the combined counts from both systems. We compared mortality ratios for the two current study periods, coded under ICD-10, with those previously published for deaths occurring in 1995–1997 and coded under ICD-9.8 We further stratified all states into three groups: states whose 2005 death certificates included the standard checkbox, states that used a nonstandard checkbox in 2005, and states that used no checkbox in 2005. We compared single-year mortality ratios for deaths based on combined counts from both systems for 2002 and 2005 among the three checkbox status groups to assess changes associated with checkbox use. We also compared separately the checkbox status for all maternal and late maternal deaths reported in the vital statistics system to evaluate the effect of the checkbox in that system. We assessed the statistical significance of differences using a z-test (statistically significant at 0.05). The statistical software package SAS was used for all data analyses.
In the United States, 1,582 maternal and late maternal deaths were reported in the National Vital Statistics System and 2,270 pregnancy-related deaths in the Pregnancy Mortality Surveillance System in the 4-year period 1999–2002, resulting in a total of 2,420 unduplicated deaths; during the 3-year period 2003–2005, there were 1,997 maternal and late maternal deaths and 1,947 pregnancy-related deaths, resulting in a total of 2,568 reported deaths. However, in 1999–2002, 60 deaths (3.8%) reported by vital statistics were not considered by the surveillance system to be pregnancy-related based on additional information and were excluded from the analysis. During 2003–2005, the number of maternal and late maternal deaths reported by vital statistics but excluded from the analysis increased over fivefold to 337 deaths (16.9%). Among the deaths excluded as not causally related, the most frequent causes of death were cancer and injury during both time periods. After excluding misclassified deaths, 4,591 deaths resulting from pregnancy were available for analysis: 2,360 in 1999–2002 (average annual 590) and 2,231 in 2003–2005 (average annual 744). Most deaths (61% and 62%, respectively, by time period) were reported by both systems; 36% and 26%, respectively, were reported only by the surveillance system and 4.4% and 13%, respectively, were reported only by the vital statistics system.
Between 1999–2002 and 2003–2005, increases in the proportion of deaths identified by the vital statistics system, and thus in “both systems,” were noted for infection, cerebrovascular accidents, cardiovascular conditions, and noncardiovascular medical conditions (P<.05) (Table 2). Although pregnancy-induced hypertension and hemorrhage were the leading causes of pregnancy death in 1995–19978 and 1999–2002, by 2003–2005, two indirect cause-of-death groups, cardiovascular conditions and noncardiovascular medical conditions, became the leading causes of death.
The causes of death for the 89 and 206 maternal deaths reported in “vital statistics only” during the two study periods were distributed across the O Chapter with a larger proportion in 2003–2005 having ICD-10 codes that indicated an indirect cause (eg, O95–O99) (data not shown).
Using combined data from both the vital statistics and surveillance systems, the average annual maternal mortality ratio increased significantly from 11.6 deaths per 100,000 live births in 1995–1997 to 13.1 in 1999–2002 and to 15.3 in 2003–2005 (Table 3). The average annual pregnancy-related mortality ratio, based on all deaths within 1 year of pregnancy, increased significantly from 12.6 per 100,000 live births in 1995–1997 to14.7 in 1999–2002 and to 18.1 in 2003–2005. Late maternal deaths were not identified as such under ICD-9. The change to ICD-10 was associated with a 12.9% increase in the maternal mortality ratio between 1995–1997 and 1999–2002 and a 16.7% increase between 1999–2002 and 2003–2005; the pregnancy-related mortality ratio increased 16.6% and 23.1%, respectively, for the same periods.
Among the 2,231 pregnancy deaths in 2003–2005, 25.2% came from states that used the standard checkbox, 43.1% used a nonstandard format, and 31.7% had no checkbox. Of the 337 deaths excluded as nonpregnancy-related in 2003–2005, 43.0% came from states with the standard checkbox, 30.3% were from states with a nonstandard format, and 26.7% had no pregnancy checkbox.
Between 2002 and 2005, mortality ratios increased more among states using a pregnancy checkbox in 2005 than among states with no checkbox in 2005 (Table 4). Although 2002 mortality ratios for deaths reported by the vital statistics system were similar for all three checkbox groups, the 2005 maternal mortality ratio increased 135% for states using the standard checkbox in 2005 and 40% for states with a nonstandard checkbox; there was no significant increase for states without a checkbox. In 2005, maternal and pregnancy-related mortality ratios for combined counts from both systems also increased significantly among states using the standard checkbox in 2005 (56% and 45%, respectively); the ratios did not change significantly among states with a nonstandard or no checkbox in 2005.
Our analysis found that pregnancy mortality ratios increased significantly after the implementation of ICD-10; however, the introduction of the 2003 revision of the U.S. Standard Certificate of Death, with the recommended standard pregnancy checkbox, was associated with an even greater increase in mortality ratios. Notably, ascertainment of deaths from indirect causes was enhanced by the combined changes in ICD-10 and the pregnancy checkbox. The new classification under ICD-10 for late maternal deaths was essentially unused by the vital statistics system until the 2003–2005 period; the number of these deaths increased from one to 77. This may be a result of the additional temporal information available from the pregnancy checkbox.
The apparent increase in case ascertainment associated with a pregnancy checkbox confirms previous studies reporting a similar effect.7,10 Although we cannot assume that the 19 states using the standard checkbox in 2005 are representative of all states in 2005, universal adoption of the 2003 revision of the U.S. Standard Certificate of Death and use of the standard checkbox may lead to a continuing increase in reported pregnancy mortality ratios in the United States. However, changes to ICD-10 and the 2003 revision of the death certificate with the recommended checkbox may have more direct effects on the classification of maternal and late maternal deaths in the vital statistics system than on the ascertainment of deaths in the surveillance system because the surveillance system relies on additional methods to identify deaths.
We found significant changes in the distribution of causes of death between the two time periods of this analysis. Although changes under ICD-10 enhanced the ascertainment of pregnancy deaths from indirect causes, temporal information available from the pregnancy checkbox further facilitated the identification of these deaths as maternal. In particular, there were increases across both systems in the ascertainment of deaths resulting from indirect causes such as cerebrovascular accidents, cardiovascular conditions, and noncardiovascular medical conditions.
Our analysis of the pregnancy checkbox suggests an area for improvement in coding maternal deaths. In 2003–2005, 15% of the maternal and late maternal deaths reported in the vital statistics system were excluded based on surveillance system cause-of-death data that indicated that the deaths were not causally related to pregnancy. Forty-three percent of the excluded deaths occurred in states with a standard format pregnancy checkbox on the death certificate, and another 30% occurred in states with a nonstandard checkbox. These findings suggest that the presence of a marked checkbox indicating a temporal relationship to pregnancy may be considered an a priori determination of maternal death under current coding practices. Additional review of cause of death in these cases may be helpful because a temporal association with pregnancy is insufficient to ascertain a causal relationship.
A small percent of deaths (1.9%) classified as maternal or late maternal by the vital statistics system were also excluded because they occurred more than 1 year after pregnancy. Although there may have been a causal relationship between pregnancy and death, the death did not meet the temporal definition. Although the vital statistics system relies solely on the death certificate for interval between pregnancy and death, the surveillance system receives the matching pregnancy outcome certificates for 95% of deaths after a live or stillbirth.2 As more states adopt the 2003 revision with the standard format checkbox, information on the temporal relationship between pregnancy and death in the vital statistics system may improve as well.
Our 2003–2005 combined U.S. mortality ratios are significantly higher than in the preceding periods.8 They are also higher than estimates reported elsewhere. The World Health Organization estimated a 42-day maternal mortality ratio of 11.0 for the United States in 200511 compared with our estimate of 15.3 for 2003–2005. Hogan et al used complex statistical modeling to estimate a mortality ratio of 17 for 2008, including maternal and late maternal deaths5 in comparison to our pregnancy-related mortality ratio of 18.1 for 2003–2005.
Several limitations of this analysis need to be considered. We lacked data on why some deaths classified in the surveillance system were not classified as maternal or late maternal by the vital statistics system and, conversely, why some deaths classified in that system were not reported to the surveillance system. This analysis cannot definitively determine if the increases in mortality ratios are the result of improved ascertainment, increasing numbers of pregnancy deaths, or a combination of both factors. However, we have compared mortality ratios before and after implementation of ICD-10 coding guidelines and the 2003 revised death certificate and find a corresponding increase in mortality ratios consistent with the explanation that these changes have generally improved ascertainment.
Underreporting of pregnancy deaths has been well documented in the United States and elsewhere.12–14 Because no single source of information captures all deaths resulting from complications of pregnancy, combining data from both the vital statistics and surveillance systems provides more comprehensive pregnancy mortality ratios and a more informed assessment of maternal mortality in the United States.
Deaths from pregnancy complications remain an important measure of maternal health in the United States and a sentinel public health indicator.15 Quality data from death certificates are a fundamental tool in monitoring pregnancy mortality and our progress as a nation in reducing it.16 With the two recent improvements in the way pregnancy deaths are defined and identified in the United States, and using combined data from both vital statistics and a national surveillance system, we identified more pregnancy deaths, resulting in higher maternal and pregnancy-related mortality ratios. When multiple changes occur over a period of time, it can be difficult to untangle the contribution each has made to the increase and to determine if there has been a change in the true risk of death. This analysis offers insight into the effects of the two changes on the numbers and types of pregnancy deaths found by two complementary systems and a more refined estimate of the pregnancy mortality ratios and the causes of pregnancy death in the 21st century.
1. Centers for Disease Control and Prevention. Achievements in public health, 1900–1999: healthier mothers and babies. MMWR Morb Mortal Wkly Rep 1999;48:4849–58.
2. Berg CJ, Callaghan WM, Syverson C, Henderson Z. Pregnancy-related mortality in the United States, 1998 to 2005. Obstet Gynecol 2010;116:1302–9.
3. Kung HC, Hoyert DL, Xu JQ, Murphy SL. Deaths: final data for 2005. National vital statistics reports. Vol. 56, No. 10. Hyattsville (MD): National Center for Health Statistics; 2008.
4. National Center for Health Statistics. Health, United States, 2009: with special feature on medical technology. Table 39. Hyattsville (MD): National Center for Health Statistics; 2010.
5. Hogan MC, Foreman KJ, Naghavi M, Ahn SY, Wang M, Makela SM, et al. Maternal mortality for 181 countries, 1980–2008: a systematic analysis of progress towards the Millenium Development Goal 5. Lancet 2010;375:1609–23.
6. World Health Organization. ICD-10: International Statistical Classification of Diseases and Related Health Problems: Tenth Revision. Geneva (Switzerland): World Health Organization; 1992.
7. Hoyert DL. Maternal mortality and related concepts. Vital Health Stat 3 2007:1–13.
8. MacKay AP, Berg CJ, Duran C, Chang J, Rosenberg H. An assessment of pregnancy related mortality in the United States. Paediatr Perinat Epidemiol 2005;19:206–14.
9. Atrash HK, Ellerbrock TV, Hogue CJ, Smith JC. The need for national pregnancy mortality surveillance. Fam Plann Perspect 1989;21:25–6.
10. Mackay AP, Rochat R, Smith JC, Berg CJ. The check box: determining pregnancy status to improve maternal mortality surveillance. Am J Prev Med 2000:19(suppl):35–9.
11. World Health Organization. Maternal mortality in 2005: estimates developed by WHO, UNICEF, UNPHA, and the World Bank. Geneva (Switzerland): World Health Organization; 2007.
12. Horon IL. Underreporting of maternal deaths on death certificates and the magnitude of the problem of maternal mortality. Am J Public Health 2005;95:478–82.
13. Deneux-Tharaux C, Berg C, Bouvier-Colle MH, Harper M, Nannini A, Alexander S, et al. Underreporting of pregnancy-related mortality in the United States and Europe. Obstet Gynecol 2005;106:684–92.
14. Schuitemaker N, Van Roosmalen J, Dekker G, Van Dongen P, Van Geijn H, Gravenhorst JB. Underreporting of maternal mortality in the Netherlands. Obstet Gynecol 1997;90:78–82.
© 2011 The American College of Obstetricians and Gynecologists
16. Atrash HK, Alexander S, Berg CJ. Maternal mortality in developed countries: not just a concern of the past. Obstet Gynecol 1995;86:700–5.