In September 2001, the first National Summit on Safe Motherhood was held in Atlanta, focusing attention on the importance of the health and safety of women before, during, and after pregnancy.1 According to official US vital statistics, the risk of death from complications of pregnancy decreased approximately 99% during the 20th century.2 However, this progress halted in 1982, and since then, there has been no improvement in the maternal mortality ratio for the United States.3 In the most recent global figures from the World Health Organization, the United States ranked 20th in maternal mortality, behind most countries of Western Europe as well as Canada, Australia, Israel, and Singapore.4 Many consider a maternal death to be a sentinel event, reflecting a breakdown in the health care system in its broadest sense.5 Mortality caused by pregnancy and its complications remains an important issue for clinical medicine, for the health care system, and as a public health indicator.
In 1987, the Division of Reproductive Health at the Centers for Disease Control and Prevention, in collaboration with state health departments and the American College of Obstetricians and Gynecologists, established the Pregnancy Mortality Surveillance System.6 Although death certificates are its core data source, the Pregnancy Mortality Surveillance System uses additional methods to attempt to identify all deaths caused by pregnancy and its complications and to provide relevant information on each death. Thus, this surveillance system permits greater precision in measuring the magnitude of pregnancy-related mortality and describing the groups at increased risk of death than do systems relying on death certificate data alone. This report presents results of the analysis of pregnancy-related deaths from the Pregnancy Mortality Surveillance System for the years 1991–1997.
MATERIALS AND METHODS
The Pregnancy Mortality Surveillance System was designed to collect data on all deaths causally related to pregnancy. Health departments in the 50 states and the District of Columbia are asked to provide deidentified copies of death certificates for all pregnancy-related deaths. For those deaths after a live birth or stillbirth, the matching birth or fetal death certificates are also requested. State maternal mortality review committees, the media, and individuals report some cases not otherwise identified. Beginning with deaths occurring in 1991, states were asked to send certificates of all deaths that occurred during or within 1 year of the end of pregnancy regardless of the cause of death or causal relationship between pregnancy and the death.
The Pregnancy Mortality Surveillance System includes data from 1979 though 1997. Data from 1979 to 1986 are considered to have been collected retrospectively.7 Data from 1987 through 1990 represent the first four years of what is considered prospectively collected data in the Pregnancy Mortality Surveillance System data.8 Data from 1991 through 1997 are the focus of this analysis.
Following the system developed by the American College of Obstetricians and Gynecologists/Centers for Disease Control and Prevention Maternal Mortality Study Group, information on all deaths was reviewed and coded by clinically experienced epidemiologists regarding the cause of death, associated obstetric conditions, and the outcome of pregnancy. Data were coded after review of all available information (including cause of death codes, notes, and other information written on the certificate, linked birth and fetal death certificates, and any other available information).
Criteria used to establish that a temporal relationship existed between pregnancy and a death included: A linked birth or fetal death certificate indicated pregnancy within the previous year; the death certificate indicated that the woman was pregnant at the time of death or had been pregnant within the previous year; or a pregnancy check box on the death certificate had been marked. Deaths were considered pregnancy related if they had a temporal and causal relationship to pregnancy, ie, the death occurred during pregnancy or within 1 year of pregnancy termination and resulted from 1) complications of pregnancy itself, 2) a chain of events initiated by pregnancy, or 3) aggravation of an unrelated event or condition by the physiologic effects of pregnancy.
The pregnancy-related mortality ratio was defined as the number of pregnancy-related deaths per 100,000 live births. Numerator data, the number of deaths occurring in the 50 states and the District of Columbia, were obtained from the Pregnancy Mortality Surveillance System. Denominator data, the number of live births that occurred in the 50 states and the District of Columbia, were obtained from public use natality tapes from the Centers for Disease Control and Prevention's National Center for Health Statistics.9 Cause-specific proportionate mortality was defined as the percent of all pregnancy-related deaths in a given time period attributed to a specified cause of death.
For both the numerator and denominator, race was defined as the race of the mother and categorized as white, black, and other (nonwhite, nonblack). For seven deaths for which maternal race was unknown, race was assigned on the basis of distribution of pregnancy-related deaths in the decedents' state for the study period. In the live-birth denominator data, the mother's race, if not stated, was imputed by the National Center for Health Statistics.9 Live-birth order, defined as the number of live births including the index pregnancy that the woman had delivered, was used as a proxy for parity, as it is included on the natality tapes whereas parity is not. Time of onset of prenatal care was categorized as the first, second, or third trimester, or no prenatal care. Because information on live-birth order and prenatal care was available only on the live-birth certificates, analyses of these variables were limited to pregnancy-related deaths occurring after a live birth. Birth certificates were available for 91.7% of cases in which the death was associated with a live birth. Marital status was categorized as married (currently married) or unmarried (never married, divorced, separated, or widowed).
For the years 1991 through 1997, the Division of Reproductive Health received a total of 5346 death certificates from state health departments. Of these, 3201 were pregnancy-related deaths (ie, they occurred during or within 1 year of pregnancy and were causally related to pregnancy). There were 1959 reported deaths that occurred during or within 1 year of pregnancy but were not caused by pregnancy complications (ie, not pregnancy related), 76 reported deaths that occurred more than 1 year after the end of pregnancy, and 110 reported deaths for which either the temporal or causal relationship to pregnancy could not be determined. All the following analyses will be limited to the 3201 reported pregnancy-related deaths.
The outcome of the pregnancy during or after which the woman died was known for 2827 (88.3%) of the women. Of women for whom the outcome of pregnancy was known, 68.2% died after a live birth; 7.8% died after a stillbirth; 11.7% died undelivered (ie, died in the second half of pregnancy but before actual delivery); 7% died from an ectopic pregnancy; 4.8% from a spontaneous or induced abortion; and 0.5% from gestational trophoblastic neoplasm. In most cases for which the outcome of pregnancy was unknown, it could not be determined whether the pregnancy had ended in a live birth or a stillbirth.
The pregnancy-related mortality ratio varied from 10.3 per 100,000 live births in 1991 to 12.9 in 1997, with an overall rate for the 7-year period of 11.5 per 100,000 live births. (Table 1). For the 7-year period, the pregnancy-related mortality ratio for white women was 7.9, for black women 29.6, and for women of other races 11.1. The pregnancy-related mortality ratio for black women was 3.8 times greater than that for white women.
Pregnancy-related mortality was lowest for women 15–19 years of age (Table 1). The risk of pregnancy-related death was higher for teens under 15 years of age and also increased progressively with maternal age. The increase in the pregnancy-related mortality ratio among older women was particularly pronounced between the ages of 30–34 and 35–39, and between the ages of 35–39 and greater than 39, when the pregnancy-related mortality ratio approximately doubled between each of the two sets of intervals. Overall, the relative ratio of pregnancy-related mortality for women over 39 years of age compared with that of women 15–19 years was 5.2 for white women but 10.2 for black women. For each age category, black women were at a higher risk of death than were white women. The ratio of pregnancy-related mortality for black teens under 15 compared with white teens under 15 was 1.5. This excess risk increased with increasing maternal age, until for women 40 and older black women were 5.6 times more likely to die than white women of a similar age.
Limiting the analysis to women whose deaths were associated with a live birth and for whom the live-birth certificate contained data on previous pregnancies, the risks of pregnancy-related death after first and second deliveries were very similar. In general, the pregnancy-related mortality ratio increased with increasing live-birth order, with the risk of pregnancy-related death approximately twice as high for live-birth order 5 or greater as at live-birth order 1 or 2. At each birth order, black women were approximately three to four times more likely to die than were white women of the same birth order.
Overall, unmarried women were at higher risk of a pregnancy-related death—16.5 deaths per 100,000 live births for unmarried women compared with 9.1 deaths for married women. However, the pattern of this risk differed for black women and white women. The pregnancy-related mortality ratio for married black women was slightly higher than that for unmarried black women (ie, 32.4 versus 28.0). Married white women had a lower pregnancy-related mortality ratio than unmarried white women, with ratios of 7.1 and 10.2. Among women of other races, there was no difference in the pregnancy-related mortality ratio by marital status. These relationships did not change when we limited the analysis to deaths associated with pregnancies of 20 weeks' gestation or greater.
Limiting the analysis to the risk of death after a live birth, we found essentially no association between the trimester of onset of prenatal care for women who had at least some prenatal care and the pregnancy-related mortality ratio (Table 2). White women who received no prenatal care had a pregnancy-related mortality ratio four times greater than did white women with any care, whereas black women with no care were 2.2 times more likely to experience a pregnancy-related death than were black women who had any care.
The interval between the end of pregnancy and death was available for 80.3% of deaths. The percent of deaths for which this information was known ranged from 96.4% of deaths after a live birth or stillbirth, to 71.1% of deaths associated with an abortion, to 19.3% of deaths after an ectopic pregnancy. Among women for whom the interval between the end of pregnancy and the date of death were known, 12.9% died while they were still pregnant; 32.4% died within 24 hours after pregnancy ended; 18.0% died between 24 hours and 7 days after the end of pregnancy; 26.1% died 8 days to 6 weeks after pregnancy had ended; and 10.7% died between 42 days and 365 days. The percent of deaths that occurred during or within 42 days of the end of pregnancy (the temporal definition of a maternal death in the International Classification of Diseases, 9th Revision10 and that used to calculate the maternal mortality ratio) varied by the cause of death. Over 95% of deaths from hemorrhage, embolism, and hypertensive disorders of pregnancy occurred within the 42-day time frame; however, this cutoff missed 14% of deaths caused by infection, 9% caused by complications of anesthesia, 9% caused by cerebrovascular accident, 16% caused by other medical conditions, and 46% caused by cardiomyopathy.
Overall, the leading causes of pregnancy-related mortality were embolism, hemorrhage, and other medical causes, the latter comprised mainly of preexisting conditions such as cardiovascular disease, diabetes, and hemoglobinopathies that were exacerbated by pregnancy (Table 3). However, the most common causes varied somewhat by the outcome of pregnancy. Almost two-thirds of deaths after a live birth were caused by either embolism, hypertensive disorders of pregnancy, or other medical causes; for deaths after a stillbirth, hemorrhage, embolism, and hypertensive disorders of pregnancy were the leading three causes. More than 90% of ectopic deaths were caused by hemorrhage. Infection and hemorrhage caused more than half the deaths associated with abortion.
Cause-specific proportionate mortality was calculated for the current study period (1991–1997) and compared with the proportionate mortality for the time period 1979–1986 and the time period 1987–19907,8 (Figure 1). Between time periods 1979–1986 and 1991–1997, the percent of deaths caused by hemorrhage decreased by more than one-third, and the percent of deaths caused by embolism decreased by 20%, although these conditions remained leading causes of pregnancy-related death. The percent of deaths caused by cardiomyopathy and other medical conditions both more than doubled between the time periods 1979–1986 and 1991–1997.
From a low of 7.2 deaths per 100,000 live births in 1987,8 the pregnancy-related mortality ratio reported by the Pregnancy Mortality Surveillance System increased to 12.9 in 1997. In the time period 1991–1997, over two-thirds of pregnancy-related deaths (and probably close to 80%) occurred after a live birth. Black women continued to have an increased risk of pregnancy-related death, almost four times greater than that of white women, as did women 35 years of age and greater and women who received no prenatal care. More than 10% of reported deaths occurred more than 42 days after the end of pregnancy, including 16% of deaths classified as stemming from other medical causes and almost half of deaths from cardiomyopathy.
Embolism and hemorrhage remained leading causes of death; however, although the latter was responsible for 28% of pregnancy-related deaths in 1979–19867 and 25.8% in 1987–1990,8 for the years 1991–1997 it was responsible for only 18.2% of reported deaths. The percent of deaths caused by other medical conditions increased from 11.7% in 1987–19908 to 18.2% in 1991–1997. Although numerically still small, the percent of pregnancy-related deaths from cardiomyopathy increased from 3.1% of all reported deaths in 1979–19867 to 7.7% in 1991–1997.
The increase in the reported pregnancy-related mortality ratio from 1991 through 1997 almost certainly reflects improved ascertainment of pregnancy-related deaths. Beginning in 1991, the Division of Reproductive Health at the Centers for Disease Control and Prevention requested that states send to the Pregnancy Mortality Surveillance System all certificates for deaths occurring during or within 1 year of pregnancy, not just those which were coded as having been caused by pregnancy complications. At that time, the death certificates for 16 states and New York City contained check boxes or questions asking if the decedent had been pregnant at the time of death or within varying lengths of time before death.11 In addition, during the 1990s, many state health departments began to use computers to link death certificates of women of reproductive age to birth and fetal death certificates. This linkage and the presence of pregnancy check boxes on the death certificates allowed states to identify more deaths with a temporal relationship to pregnancy, from which those that were pregnancy related (ie, also causally related to pregnancy) could be found. This is reflected in the fact that the percent of the death certificates sent by state health departments to the Pregnancy Mortality Surveillance System that were determined, upon review, to be pregnancy related decreased from 97% for the years 1979–19867 to 89.9% for 1987–19908 to 59.9% for the period of the current report.
The striking disparity in the risk of pregnancy-related death between black women and white women—a 3.8-fold difference—is the largest black/white gap of any indicator used in the field of maternal and child health and one that has persisted at the same magnitude for more than 60 years.12 Black women have a greater risk of pregnancy-related death regardless of age, live-birth order, marital status, or trimester of onset of prenatal care. However, with the exception of increasing maternal age, the relative risk of pregnancy-related death for black women compared with white women was greatest among those groups whose absolute risk of death was lowest. A more detailed case-control analysis of Pregnancy Mortality Surveillance System data from 1979–1986 had similar findings. The excess risk of a pregnancy-related death after a live birth for black women was greatest among women with the absolute lowest risk, that is, women of low parity, with more education, with adequate prenatal care and who delivered a normal birth weight infant at term.13 This higher relative risk of an adverse outcome among apparently low-risk black women and their offspring has been found in studies of other important reproductive health indicators, such as very low birth weight, low birth weight, and both neonatal and infant mortality,14 for reasons that remain unclear.
A mortality rate is the product of the frequency, or prevalence, of a condition by the risk of death if the condition is present, or the case-fatality rate. If black women are more likely to die of pregnancy complications, it is because they either have higher rates of the complications or have a greater risk of dying from the conditions if they occur. We hope to use this paradigm in special studies as we try to understand why black women have an increased risk of pregnacy-related death by asking the relevant questions. Do they have more complications, such as hypertensive disorders of pregnancy, hemorrhage, or pulmonary embolism? If they develop a complication, is it more severe? Do they have more comorbidities? And do they have access to and receive the needed level and intensity of care?
Some of the changes in the distribution of the causes of pregnancy-related death from 1979 to 1997 are striking. Although the use of proportionate mortality has inherent limitations, several of the changes in proportionate mortality found are consistent with advances in technology and improvements in clinical care, as well as in better case ascertainment. Perhaps most dramatic is the relative decrease in deaths from hemorrhage, particularly between the periods 1986–19908 and 1991–1997, when the percent of deaths caused by hemorrhage declined from 25.8% to 18.2%. Two factors may have contributed to this decrease. First, between these two time periods, the proportion of pregnancy-related deaths associated with an ectopic pregnancy decreased from 10.7%8 to 5.6%, most likely because of improved ability to diagnose and treat ectopic pregnancies early.15 Second, advances in the management of hemorrhage in general,16 including use of prostaglandin analogues and surgical techniques, such as embolization, are probably reflected in the decline in hemorrhage deaths associated with all pregnancy outcomes.
Increased use of linkages and check boxes indicating the occurrence of pregnancy leads to increased identification of pregnancy-related deaths, particularly those from cardiomyopathy and other medical conditions.17 The relationship between these deaths and pregnancy can easily be missed, as many occur more than 42 days postpartum, and the causal relationship between pregnancy and the death may not be reflected in the cause of death information on the death certificate and, thus, not be reflected in the cause of death code. The increase in deaths from other medical conditions may also be affected by changes in the age distribution of women giving birth. The prevalence of chronic medical conditions increases with age, and more women in the United States are becoming pregnant at older ages. Between 1976 and 1997, the pregnancy rate for women 35–39 years increased 74% (from 35.3 pregnancies per 1000 women to 61.3 per 1000) and that for women 40 years and older increased 41% (from 9.9 per 1000 women to 13.7).18,19 The percent of births to women 35 years and older went from 4.5% to 12.6% of all live births. In addition, more women with serious medical conditions who previously would not have become pregnant might be doing so now.
Pregnancy-related death, although too frequent, is a relatively rare event. With data from all states and the District of Columbia, the Pregnancy Mortality Surveillance System provides a national, population-based data set, which allows us to view pregnancy-related mortality at a national level, with numbers sufficient to look at trends and major risk factors. In addition to the official cause of death codes on the certificate, the Pregnancy Mortality Surveillance System also allows use of notes written on death certificates, information on the matching birth or fetal death certificates if available and, in some cases, reports from maternal mortality committees and other sources. This expanded information, along with review by clinically experienced epidemiologists, allows the cause of death and the outcome of pregnancy to be reported separately and the cause of death to be reported in a more clinically meaningful way. The increase in the numbers of pregnancy-related deaths reported to the Pregnancy Mortality Surveillance System indicates that new methods to identify deaths during or within 1 year of pregnancy being used by state health departments are, in fact, improving case ascertainment.
However, the Pregnancy Mortality Surveillance System still relies heavily on vital record data and, thus, is limited in its ability to identify and to describe pregnancy-related deaths. Even with the improved case finding over the current study period, the Pregnancy Mortality Surveillance System receives data from state health departments, which are not uniform in their use of methods of case finding such as computer linkages, check boxes, and periodic queries. Death and birth certificates contain very limited clinical, social, and behavioral information, and those variables that are available do not help explain the racial disparities.13 The detailed and varied chain of events that led to the death of the woman cannot be determined from these data sources. However, the purpose of the Pregnancy Mortality Surveillance System is to provide an overview of pregnancy-related mortality and associated risk factors on a national level; it is the purview of state maternal mortality review committees to investigate pregnancy-related deaths in detail and seek ways to reduce them and so improve maternal health in their states.11
Since 1982, there has been no improvement, with the maternal mortality ratio remaining around eight deaths per 100,000 live births.20 Although some may claim this is an irreducible minimum in the number of deaths caused by pregnancy, three lines of evidence would indicate otherwise. First, significant gaps exist between racial and ethnic groups,21 without any biologic reason for these disparities. Second, a score of other countries have maternal mortality ratios lower than those of the United States.4 Third, studies show that many pregnancy-related deaths are preventable through changes in patient, provider, and system factors.22
More than 25 years ago, the concept of sentinel events was introduced, as “an unnecessary disease, disability or untimely event which justified carefully controlled scientific search for remediable underlying causes.”5 The authors cited maternal deaths as an example of such an event. Today, even though they are uncommon, pregnancy-related deaths deserve to be identified and carefully reviewed at the state level. We must go beyond the identification of pregnancy-related deaths to their review and analysis of the data. There lie lessons to be learned and actions taken to improve maternal health.
1. Bruce FC, Safe Motherhood Summit Working Group. Highlights of the national summit on safe motherhood: Investing in the health of women. Matern Child Health J 2002;67:67–9.
2. Centers for Disease Control and Prevention. Healthier mothers and babies. MMWR Morb Mortal Wkly Rep 1999;48:849–57.
3. Centers for Disease Control and Prevention. Maternal mortality—United States, 1982–1996. MMWR Morb Mortal Wkly Rep 1998;48:705–7.
4. World Health Organization. Maternal mortality in 1995: Estimates developed by WHO, Unicef, UNFPA. Geneva, Switzerland: WHO/RHR01.9, 2001.
5. Rutstein DD, Berenberg W, Chalmers TC, Child CG, Fishman AP, Perrin EB. Measuring the quality of medical care—a clinical method. N Engl J Med 1976;294:582–8.
6. Atrash HK, Rowley D, Hogue CJR. Maternal and perinatal mortality. Curr Opin Obstet Gynecol 1992;4:61–71.
7. Atrash HK, Koonin LM, Lawson HW, Franks AL, Smith JC. Maternal mortality in the United States, 1979–1986. Obstet Gynecol 1990;76:1055–60.
8. Berg CJ, Atrash HK, Koonin LM, Tucker M. Pregnancy-related mortality in the United States, 1987–1990. Obstet Gynecol 1996;88:161–7.
9. National Center for Health Statistics. Natality data, public-use data files, detail files (1989–1997). NCHS, CDC. Available via the Internet at: http://www.cdc.gov/nchs/products/catalog/subjects/natality.htm
. Accessed 17 Oct 2001.
10. World Health Organization. Manual of the International Statistical Classification of Diseases, Injuries, and Causes of Death (ICD-9). Vol. 1. Geneva, Switzerland: World Health Organization, 1977.
11. Berg C, Danel I, Atrash H, Zane S, Bartlett L, eds. Strategies to reduce pregnancy-related deaths: From identification and review to action. Atlanta, GA: Centers for Disease Control and Prevention, 2001.
12. Centers for Disease Control and Prevention. Differences in maternal mortality among black and white women—United States, 1990. MMWR Morb Mortal Wkly Rep 1995;44:6–7,13–14.
13. Saftlas AF, Koonin LM, Atrash HK. Racial disparity in pregnancy-related mortality associated with livebirth: Can established risk factors explain it? Am J Epidemiol 2000; 152:413–41.
14. Berg CJ, Wilcox LS, d'Almada PJ. The prevalence of socioeconomic and behavioral characteristics and their impact on very low birthweight in black and white infants in Georgia. Matern Child Health J 2001;4:75–84.
15. Lipscomb GH, Stovall TG, Ling FW. Nonsurgical treatment of ectopic pregnancy. N Engl J Med 2000;343:1325–9.
16. American College of Obstetricians and Gynecologists. Postpartum hemorrhage. ACOG educational bulletin no. 243. Washington: American College of Obstetricians and Gynecologists, 1998.
17. Centers for Disease Control and Prevention. Pregnancy-related mortality—Georgia, 1990–1992. MMWR Morb Mortal Wkly Rep 1995;44:93–5.
18. Ventura SJ, Mosher WD, Curtin SC, Abma JC, Henshaw S. Trends in pregnancies and pregnancy rates by outcome: Estimates for the United States, 1976–96. Vital Health Statistics, Vol. 21, No. 56. Hyattsville, Maryland: National Center for Health Statistics, 2000.
19. Ventura SJ, Mosher WD, Curtin SC, Abma JC, Henshaw S. Trends in pregnancy rates for the United States, 1976–97: An update. National Vital Statistics Reports, Vol. 49, No. 4. Hyattsville, Maryland: National Center for Health Statistics, 2001.
20. Eberhardt MS, Ingram DD, Makuc DM, Pamuk ER, Freid VM, Harper SB, et al. Urban and rural chartbook. Health, United States, 2001. Hyattsville, Maryland: National Center for Health Statistics, 2001.
21. Centers for Disease Control and Prevention. Pregnancy-related deaths among Hispanic, Asian/Pacific Islander, and American Indian/Alaska Native Women—United States, 1991–1997. MMWR Morb Mortal Wkly Rep 2001;50:361–4.
22. Mertz KJ, Parker AL, Halpin GJ. Pregnancy-related mortality in New Jersey, 1975–1989. Am J Public Health 1992;82:1085–8.