Maternal mortality statistics for the United States have shown little improvement for 2 decades, and 20 countries have lower rates1,2: For 1980, the crude maternal mortality ratio within the United States was 9.2 deaths per 100,000 live births,3 whereas, for 1997, this ratio was 8.4.4 The Healthy People 2010 objective for maternal mortality is no more than 3.3 maternal deaths per 100,000 live births.5 To achieve this goal, we must identify potentially modifiable risk factors. Although attributes such as race, age, and parity have been associated with pregnancy-related death, the factors most mutable are health care services factors.
The Centers for Disease Control and Prevention (CDC), among others, recognized that pregnancy-related deaths are underreported, and epidemiologic studies are often hampered by incomplete ascertainment of cases.6 Ascertainment may be increased by using multiple sources. Linkage of death certificates of women of childbearing age with live-birth certificates recorded within a year of the woman's death has been reported to increase ascertainment by 36% to 150%.6,7
The traditional term “maternal death” includes deaths related to pregnancy occurring up to 42 days after the end of pregnancy. However, studies from the CDC have shown that 6–11% of pregnancy-related deaths occur more than 42 days after delivery.8,9 The term “pregnancy-related death” is used to describe deaths causally related to pregnancy occurring up to 1 year after the end of pregnancy.
We undertook this study to determine whether there is an association between pregnancy-related death and health care services variables including receipt of Maternity Care Coordination services; receipt of Women, Infants, and Children nutritional services; source of prenatal care; method of delivery; and level of use of prenatal care. We used enhanced methods of case ascertainment and expanded definitions, to include deaths up to 1 year after delivery.
MATERIALS AND METHODS
North Carolina uses a comprehensive approach to identify potential pregnancy-related deaths. This surveillance system identifies all death certificates with any mentioned cause of death related to a pregnancy complication (International Classification of Diseases, 9th Revision, codes 630–676) or with any other indication that the decedent was currently or recently pregnant. In addition, records of all deaths occurring in females between the ages of 10 and 50 are electronically matched with live-birth and fetal death files to identify all women who had had a delivery within 1 year of death. Matching variables include the decedent's name, date of birth, and social security number. The inpatient hospital discharge database is also used to identify all individuals with a primary or secondary discharge diagnosis relating to pregnancy and whose discharge status was reported as “deceased.” The surveillance program is a statewide population-based system covering around 7.6 million residents (75% white non-Hispanic, 22% black non-Hispanic, and 2% Hispanic). About 66,000 deaths and 110,000 live births occur in North Carolina each year. There are approximately 80 hospitals, including eight level three hospitals, with labor and delivery services. The number of annual deliveries ranges from approximately 30 to 6500 per hospital. Approval for this study was granted by the North Carolina State Registrar and the Wake Forest University Institutional Review Board.
We employed the definitions of pregnancy-associated, pregnancy-related, and nonpregnancy-related death adopted by the Maternal Mortality Study Group of The American College of Obstetricians and Gynecologists (ACOG) and the CDC.10 Those deaths that occurred during pregnancy or within 1 year of delivery or pregnancy termination, regardless of cause, were considered pregnancy associated. Pregnancy-associated deaths were reviewed by a panel of three obstetricians experienced in maternal mortality review. We examined the matching live-birth and death certificates and all autopsy or other medical information that had been obtained as part of the state maternal mortality review process. Each case was independently classified as pregnancy related or nonpregnancy related. Deaths were considered to be pregnancy related if the occurrence was within 1 year 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. Deaths were considered to be nonpregnancy related if the cause was not related to pregnancy. Discrepancies among reviewers were discussed, and unanimous agreement was reached on all cases.
For the 7-year study period 1992–1998, 400 pregnancy-associated deaths were identified by the surveillance system. Of these, 151 (37.8%) were pregnancy related and 249 (62.2%) were nonpregnancy related. Of the 151 pregnancy-related deaths, 33 (21.9%) did not have a live-birth outcome. Because all controls (described below) did experience a live birth, and because additional analysis variables were not available for cases without a live birth, these 33 deaths were excluded. The distribution of racial groups, age groups, level of education, and income categories did not differ among cases without a live-birth outcome and cases with a live-birth outcome. The remaining 118 deaths were the cases in the case–control analysis. Because all cases in the numerators in this study had a live-birth outcome and therefore are also represented in the denominators, the term rate rather than the ratio is used when discussing the case–control analyses.
From the 731,217 live births registered in the state for the 7-year study period, a one in 250 (0.4%) random sample was selected to serve as comparison subjects (controls). This provided us with 3697 controls. This sample size was sufficient to ensure that the standard errors for subgroup prevalences were less than 1%. To avoid overmatching and therefore overlooking significant associations, controls were not matched with cases but were sampled to be equally distributed over the 7-year study period. No controls were used more than once.
The primary predictor variables of interest were receipt of Maternity Care Coordination services; receipt of Women, Infants, and Children services; source of prenatal care; method of delivery; and level of use of prenatal care. These variables were selected because of their expected protective or harmful association with pregnancy-related death. Electronic Women, Infants, and Children; Maternity Care Coordination; and Medicaid files were merged with the live-birth certificate file for each case and control. The Maternity Care Coordination and Women, Infants, and Children files were used to code receipt of these services. The Maternity Care Coordination and Medicaid files were used to code source of prenatal care. Source of care was coded as public if care was received in a public health department or private for all other case patients and controls receiving prenatal care.
Method of delivery was classified as cesarean or vaginal as indicated on the birth certificate. Nine cases (7.6%) and 376 controls (10.2%) had missing information regarding method of delivery.
Level of use of prenatal care was determined from the birth certificate. If the subject had no prenatal visits, care was classified as “none.” If the subject had at least one prenatal visit, care was classified as “some.” Level of use of prenatal care was further categorized as adequate or inadequate based on the report of McDermott et al11 that indicated acceptable accuracy for a two-category scheme when using birth certificates as the data source. The classification scheme depends on the expected number of visits for gestational age at delivery based on ACOG's schedule.12 These variables were coded from the birth certificates.
Analyses were adjusted by mother's age, level of education, income, medical risk factors, and complications of labor and delivery. Level of education and income were selected to control for socioeconomic status. Assigned income was equal to the median per capita income for the ZIP code of residence. Median per capita income for each ZIP code was derived from US Census data available through the Web site http://venus.census-.gov/cdrom/lookup. This approach to estimating income has been validated elsewhere.13 Mother's age, level of education, ZIP code, medical risk factors, and complications of labor and delivery were derived from the birth certificate. Approval for this study was granted by the North Carolina State Registrar and the Wake Forest University Institutional Review Board.
We analyzed the differences in the distribution of the health care services variables between cases and controls using χ2 and the Fisher exact test. The association between maternal outcome and each of the primary predictor variables was assessed with univariate logistic regression analysis. We calculated the crude odds ratios (ORs) for the outcome of pregnancy-related death for exposure to the health care services variables of interest. For those health care factors found to have a significant association with pregnancy-related death, we made adjustments for covariates using multivariable logistic regression analysis to calculate the adjusted ORs. Analyses were performed using SAS software (SAS Institute Inc., Cary, NC). A significance level of .05 was selected.
For the population during this 7-year period, the pregnancy-associated mortality ratio was 54.7 deaths per 100,000 live births. The pregnancy-related mortality ratio was 20.7 deaths per 100,000 live births.
The distribution of primary predictor variables among cases and controls is shown in Table 1. The crude and adjusted ORs for pregnancy-related death associated with each of these variables are shown in Table 2. There was no association between Maternity Care Coordination or Women, Infants, and Children services and pregnancy-related death, nor was there an association with source of prenatal care, public versus private.
Cesarean delivery was associated with a significantly increased risk for pregnancy-related death. For the population, we determined the rate of pregnancy-related death among all cesarean deliveries and all vaginal deliveries for the 7-year study period. The pregnancy-related mortality rate among women with cesarean deliveries was 35.9 deaths per 100,000 cesarean deliveries with a live-birth outcome. The pregnancy-related mortality rate among women with vaginal deliveries was 9.2 deaths per 100,000 vaginal deliveries with a live-birth outcome. The resulting relative risk for pregnancy-related death associated with cesarean delivery for the population was 3.9 (95% confidence interval [CI] 2.7, 5.7) and was not statistically distinct from the computed OR of 5.6 (95% CI 3.8, 8.3) shown in Table 2. We reviewed all pregnancy-related deaths with cesarean delivery to determine if any were perimortem or postmortem procedures and identified one definite and four possible perimortem procedures. The rate of pregnancy-related death for the population by method of delivery was recalculated excluding these five cases. The relative risk for pregnancy-related death associated with cesarean delivery remained significant at 3.6 (95% CI 2.4, 5.2). Method of delivery was not coded on the birth certificate for 20 cases. After review of supplemental information and attempts to contact providers, we were unable to determine method of delivery for nine cases. We performed a sensitivity analysis assuming first that all nine were vaginal deliveries and second that all nine were cesarean deliveries. The resulting ORs for pregnancy-related death associated with cesarean delivery were 4.8 and 6.5, respectively, and did not change the conclusion that cesarean delivery is associated with a significantly increased risk for pregnancy-related death.
Because women with medical risk factors, preterm deliveries, or older age may be at increased risk for both cesarean delivery and pregnancy-related death, we explored the association between these factors and both method of delivery and case status. Six medical risk factors, age, and preterm delivery were found to be associated with both pregnancy-related mortality and cesarean delivery (Table 3). The adjusted OR (adjusting simultaneously for these medical risk factors, age, and preterm delivery) for pregnancy-related death associated with cesarean delivery versus vaginal delivery was 3.9 (95% CI 2.5, 6.1). This OR did not statistically differ from the unadjusted OR or the relative risk for the population.
To further examine the association between method of delivery and pregnancy-related mortality, we examined the pregnancy-related mortality rates by cause of death among cesarean deliveries and vaginal deliveries. The rates were higher among women with cesarean delivery for all causes of pregnancy-related death (Table 4).
Receipt of any prenatal care was associated with a significantly decreased risk for pregnancy-related mortality. Because women of lower socioeconomic status may be less likely to receive prenatal care and also at greater risk for pregnancy-related death, we explored the association between measures of socioeconomic status (income and education attainment) and both receipt of prenatal care and case status (Table 5). Women with an estimated annual income of less than $10,000 and women with less than a high school education were at significantly increased risk for both no prenatal care and pregnancy-related death. After adjusting for income and education attainment, the adjusted OR for pregnancy-related death associated with receipt of prenatal care was 0.2 (95% CI 0.1, 0.6) and did not differ from the unadjusted OR of 0.2 (95% CI 0.1, 0.4). Among women with at least one prenatal visit, there was no statistically significant association between pregnancy-related death and quantitatively adequate versus inadequate care, with an OR of 0.8 (95% CI 0.5, 1.3).
This study is unique in several respects. It comes from a contemporary and geographically defined population and is one of the few case–control studies examining the associations between pregnancy-related death and multiple health care services factors. Cases were identified using three methods, including electronic linkage of birth and death files. All pregnancy-related deaths were validated by an expert panel. Adjustments were made for confounders in the analysis.
We used birth certificate information to code the predictor variables. Buescher et al reported a follow-back study of the accuracy of information on birth certificates in North Carolina for the version implemented in 1988 and used throughout the 7 years of this study.14 The rates of agreement between the birth certificate and medical record were as follows: for method of delivery, 91.9%; for prenatal care variables, 79–82%; and for medical complications, 85.8%.
We conducted this study to identify modifiable health care services factors that are currently associated with pregnancy-related death. We hypothesized that Maternity Care Coordination or Women, Infants, and Children services may be protective, but this was not the case, suggesting that these services do not prevent very serious pregnancy-related complications. The absence of an association between sources of prenatal care, private versus public, is reassuring. Women receiving care in health departments may be expected to have more socioeconomic risk factors.
Method of delivery is an important health care services variable in intrapartum care. For the year 2000, the rate of cesarean delivery for the United States increased for the fourth consecutive year to 22.9% of all live births, a 4% increase from 1999. The primary cesarean delivery rate also increased by 4% to 16.1%.15 The Healthy People 2010 objective 16-9a is a rate of cesarean delivery of no more than 15% among low-risk women giving birth for the first time.5 In 1999, the national rate of cesarean delivery births to these women was 19%.16 Comparing recommended rates with national statistics suggests that the rate of cesarean delivery can be decreased without compromising neonatal outcomes; however, this concept is debated among obstetricians. We were unable to examine fetal indications or outcomes with our data source.
Case patients were much more likely than controls to undergo cesarean delivery. This association persisted after controlling for medical risk factors, age, and pre-term delivery. Most other studies examining the association between method of delivery and pregnancy-related mortality have also found an increased risk with cesarean delivery, even after adjusting for intercurrent diseases.17,18 An analysis of 3.5 million births by the Task Force on Cesarean Childbirth, sponsored by the National Institute of Child Health & Human Development, grouped cesarean deliveries by indication. The mortality rate was lowest in the group whose indication was previous cesarean delivery, presumed to be primarily elective repeat cesarean delivery in otherwise healthy women. The mortality rate in this population was 18.4 deaths per 100,000 cesarean deliveries, compared with 9.8 deaths per 100,000 vaginal deliveries with no complications. These authors concluded that cesarean delivery is probably neither less than two nor more than four times more hazardous than vaginal delivery.17 A report of 108 maternal deaths from Cape Town, South Africa, calculated maternal mortality rates (deaths per 100,000 procedures) by method of delivery and in different risk categories. The rate of direct maternal deaths among women with vaginal deliveries and without acute antenatal obstetric problems was six. For this same risk category undergoing elective cesarean delivery, the rate was 23.18
Our finding of an association between level of use of prenatal care and pregnancy-related death is in agreement with earlier studies.7,19 Five of the seven cases without prenatal care may have been prevented with prenatal care: Three died from preeclampsia or eclampsia, one died from sepsis after prolonged rupture of the fetal membranes, and one died from complications of hypertension. From these cases, the elements of prenatal care that would seem to be essential for improving maternal survival are blood pressure measurement, urine protein determination, and an obstetric-focused review of systems. Controlling for socioeconomic variables did not change the magnitude of the association between pregnancy-related mortality and prenatal care, suggesting that the impact of prenatal care is independent of socioeconomic status. This finding is in agreement with a study from 1975 through 1982 from Elkhart and Kosciusko Counties in northeastern Indiana of well-educated and prosperous Faith Assembly members who did not seek prenatal care and delivered at home without trained attendants.20 Compared with other women within the state giving birth, these women were more likely to be white, married, and between the ages of 20 and 34. There were six maternal deaths during the 8 years, which yielded a maternal mortality ratio of 872 deaths per 100,000 live births, or 92 times the maternal mortality ratio for the remainder of the state! All deaths occurred at home. Four of the six deaths were from hemorrhage, and the remaining two were from infection. Our findings also support the conceptual framework presented by Maine and Rosenfield.21,22 With experience in both developing countries and the United States and knowledge of the history of maternal mortality, these investigators contend that, unlike infant mortality, nutrition, education, and general standard of living are not major factors in maternal mortality; rather, socioeconomic factors affect maternal mortality by affecting access to health care.
Reducing the number of preventable deaths among pregnant and recently pregnant women continues to be a primary public health objective. The associations between antepartum and intrapartum health care services variables observed in the present study imply that the Healthy People 2010 objectives for reducing maternal mortality can be achieved through system changes. Our results suggest that improving use of prenatal care and lowering the cesarean delivery rate could potentially reduce pregnancy-related mortality in the United States.
1. Centers for Disease Control and Prevention. Maternal mortality—United States, 1982–1996. MMWR Morb Mortal Wkly Rep 1998;47:705–7.
2. World Health Organization. WHO revised 1990 estimates of maternal mortality: A new approach by WHO and UNICEF. Report no. WHO/FRH/MSM/96.11. Geneva: World Health Organization, 1996.
3. Centers for Disease Control and Prevention, National Center for Health Statistics. Vital statistics of the United States. Vol I–II. Washington: U.S. Government Printing Office, 1950–1993.
4. Centers for Disease Control and Prevention. Table 44. Maternal mortality for complications of pregnancy, childbirth and the puerperium, according to race, Hispanic origin, and age: United States, selected years 1950–99. Available at: http://www.cdc.gov/nchs/data/hus/tables/2001/01hus044.pdf
. Accessed 2002 Oct 1.
5. U.S. Department of Health and Human Services. Healthy People 2010: Understanding and improving health. 2nd ed. Washington: U.S. Government Printing Office, 2000.
6. Centers for Disease Control and Prevention. Enhanced maternal mortality surveillance—North Carolina, 1988 and 1989. MMWR Morb Mortal Wkly Rep 1991;40:469–71.
7. Atrash HK, Alexander S, Berg CJ. Maternal mortality in developed countries: Not just a concern of the past. Obstet Gynecol 1995;86:700–5.
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. Rochat RW, Koonin LM, Atrash HK, Jewett JF. Maternal mortality in United States: Report from the maternal mortality collaborative. Obstet Gynecol 1988;72:91–7.
10. Berg C, Danel I, Atrash H, Zane S, Bartlett L, eds. Strategies to reduce pregnancy-related deaths: From identification and review to action. Atlanta: Centers for Disease Control and Prevention, 2001.
11. McDermott J, Drews C, Green D, Berg G. Evaluation of prenatal care information on birth certificates. Paediatr Perinat Epidemiol 1997;11:105–21.
12. American College of Obstetricians and Gynecologists. Standards for obstetric-gynecologic services. 6th ed. Washington: American College of Obstetricians and Gynecologists, 1985:17–8.
13. McBean AM, Hebert P. Comparison of income information in the 1990 census with information in the Medicare Current Beneficiary survey. Baltimore: Health Care Financing Administration, 1995.
14. Buescher PA, Taylor KP, Davis MH, Bowling JM. The quality of the new birth certificate data: A validation study in North Carolina. Am J Public Health 1993;83:1163–5.
15. Martin JA, Hamilton BE, Ventura SJ, Menacker F, Park MM. Births: Final data for 2000. Natl Vital Stat Rep 2002;50(5):1–101.
16. Menacker F, Curtin SC. Trends in cesarean birth and vaginal birth after previous cesarean, 1991–1999. Natl Vital Stat Rep 2001;49(13):1–16.
17. Petitti DB, Cefalo RC, Shapiro S, Whalley P. In-hospital maternal mortality in the United States: Time trends and relation to method of delivery. Obstet Gynecol 1982;59:6–12.
18. Lilford RJ, Van Coeverden de Groot A, Moore PJ, Bingham P. The relative risk of caesarean section (intrapartum and elective) and vagina delivery: A detailed analysis to exclude the effects of medical disorders and other acute pre-existing physiological disturbances. Br J Obstet Gynaecol 1990;97:883–92.
19. Atrash HK, Koonin LM, Lawson HW, Franks AL, Smith JC. Maternal mortality in the United States, 1979–1986. Obstet Gynecol 1990;76:1055–60.
20. Kaunitz AM, Spence C, Danielson TS, Rochat RW, Grimes DA. Perinatal and maternal mortality in a religious group avoiding obstetric care. Am J Obstet Gynecol 1984; 150:826–31.
21. Maine D. How do socioeconomic factors affect disparities in maternal mortality? J Am Med Womens Assoc 2001; 56:189–90.
© 2003 The American College of Obstetricians and Gynecologists
22. Maine D, Rosenfield A. The safe motherhood initiative: Why has it stalled? Am J Public Health 1999;89:480–2.