Secondary Logo

Journal Logo

Obstetric Anesthesiology: Research Report

Why Do Pregnant Women Die? A Review of Maternal Deaths from 1990 to 2010 at the University of Alabama at Birmingham

Frölich, Michael A. MD, MS*; Banks, Catiffaney BS*; Brooks, Amber MD; Sellers, Alethia MD*; Swain, Ryan PhD, MD; Cooper, Lauren*

Author Information
doi: 10.1213/ANE.0000000000000457
  • Free


During the past 20 years, the United States maternal mortality ratio has doubled from 7.4 per 100,000 live births in 1986 to 14.5 in 2005. The 2005 rate among African American women was 37.7, 3- to 4-fold higher than among Caucasian women.1 The concern about equality in the provision of health care is very timely in Birmingham because this city recently commemorated the 50th anniversary of the civil rights movement in the United States. In 1963, Martin Luther King, Jr, and the Southern Christian Leadership Conference led thousands of African American volunteers in a series of nonviolent protests that drew the world’s attention to racial segregation in the Southern United States, and led to the Civil Rights Act of 1964, which prohibited racial discrimination in hiring and public services in the United States. Social inequality translates into the field of health care, an ongoing area of concern. Experts in maternal health have called for actions to eliminate the racial disparity and reduce the rate of maternal deaths.1,2 While the need for action is clear, it is still somewhat obscure what exactly needs to be done in terms of public policy. The main limitation to our current understanding of maternal mortality in the United States is that national or statewide data are flawed by notorious inaccuracies in extracting causes of death and associated medical conditions from death certificates.3 We hypothesized that a disproportionate number of African American women died, relative to their prevalence in the overall population of women who delivered at our tertiary center and that socioeconomic or medical factors may help to distinguish women who died from those who did not. To investigate these concerns, we conducted a single center case-control study at University of Alabama at Birmingham (UAB) Hospital. Our primary study hypothesis was that the proportion of African American women among mothers who died (cases) would be significantly larger than the proportion of African American women who did not die (controls). Secondary study aims were to identify cofactors that would explain a possible racial disparity in maternal deaths and to categorize the causes of maternal death.


This was a case-control study of maternal deaths that occurred at the UAB Hospital during the years 1990 to 2010 and was approved by our IRB. UAB Hospital is located in the center of Birmingham and is the largest hospital in the state of Alabama with an average of 835 patients seen daily and 1,146 beds. The institution has an annual delivery volume of approximately 4500, covers all aspects of maternal–fetal medicine, and offers 24-hour access to subspecialty, adult, and neonatal intensive care services. Because there is an approximately equal proportion of African American and Caucasian patients delivering at UAB Hospital, the study population is ideal for the study of race as a possible determinant of health care outcomes.

The case ascertainment strategy was to first identify all women who died (all females with discharge status expired or autopsy in the electronic uniform billing database) between ages 11 and 50 years and subsequently to inspect all ICD-9 codes of the deceased patients for any codes related to pregnancy, childbirth, or the postpartum state. Maternal deaths were matched 2:1 with patients of the same age (±3 years) who delivered during the same week as the case (death) to which they were matched. Specifically, the 2 patients fitting the age-match criteria who delivered closest to the time of death of their associated case were extracted in a 2-step fashion. Possible controls from the same day were identified electronically, and the 2 closest matches were manually chosen. The medical record numbers of cases and controls identified in this manner provided the basis of a manual review of medical records. A research specialist conducted a preliminary review, and a physician then performed a more thorough secondary review. The ongoing medical record review was periodically discussed among co-investigators, and a third review was undertaken if there were remaining ambiguities concerning the accuracy or correct interpretation of the data.

Our medical record review was discussed and standardized by the investigators a priori and focused on the collection of factors contributing to maternal deaths based on research.4–6 The primary outcome variable was death. Predictive variables included race (categorized as Asian, African American, Decline/Refuse, Hispanic/Latino, American Indian, Hawaiian/Pacific Islander, Caucasian by patient self-report), body mass index (BMI) recorded closest to death, duration of hospital stay (days), salary, distance of residence from UAB Hospital, parity, gestational age (calculated at the time of death for those who died undelivered and at the time of delivery for those who died postpartum), insurance type, fetal outcome (per discharge record), pre-existing medical conditions, and hospital-acquired comorbidities. Some demographic information was available in electronic format but much of the medical data were retrieved from paper medical records (stored in microfiche format). Income level was estimated using the median annual salary for each zip code reported on the database.a The distance of a patient’s residence from the hospital was calculated using Google maps based on the patient’s residential address and the hospital address.b Causes of maternal deaths were categorized using the framework of the Centre for Maternal and Child Enquiries (CMACE),7 a report initiated by the National Health Service of the United Kingdom in 1952. The determination of the final cause of death, if unclear, was based on a consensus of the reviewing physician collaborators. A few basic definitions were adopted from the Eighth Report of the CMACE in the United Kingdom.4 In concordance with the CMACE, we included incidental causes of deaths in our analysis.

The variable that describes case versus control designation will from here on simply be referred to as “status.” Our a priori analysis plan was designed to quantify the strength of univariate associations of each explanatory variable and status. The relationship of categorical explanatory data and status was compared with a χ2 test or Fisher exact test as appropriate. The relationship of Gaussian-distributed continuous data and status was compared using a two-sided t test with unequal variances, and nonnormally distributed continuous data were compared using the Wilcoxon rank sum test.

A logistic regression model (with 2 explanatory variables) was added as a secondary analysis tool to investigate how the distance of a woman’s residence from the hospital might help to better characterize the relationship of race and status. This model had race and distance as explanatory variables and group (case versus control) as outcome. A race-by-distance interaction term would be included only if it proved to be statistically significant. There were very few missing data (less than 3%). Any missing data were excluded from the affected portion of the analysis. A comparison of distance to the hospital by race was performed with an analysis of variance after log transforming the non-Gaussian distributed distance data.


There was insufficient evidence to suggest racial disparity (P = 0.23); the proportion of African American women was 57% (42 of 77) in the maternal death group and 61% (94 of 154) in the control group. The estimated unadjusted odds ratio of maternal death for African American to Caucasian race was 0.66 (95% confidence interval [CI], 0.37–1.19). Results for our secondary analyses were as follows: There was no significant association of mortality status and insurance status, income, BMI, marital status, or parity (all P > 0.18, Table 1). However, there was a significant difference between case and control patients in the resident-hospital distance, gestational age, fetal survival, duration of hospital stay, lack of prenatal care, and cesarean delivery rate (all P < 0.042, Table 1). Distance of residence to the hospital differed by race; African American women lived significantly (P < 0.001) closer (14.0 miles [95% CI, 11.5–17.2]) to the hospital than Caucasian women (34.2 [95% CI, 26.0–45.0]). Overall, longer distance was associated with more frequent mortality.

Table 1
Table 1:
Cases Versus Controls

In the full logistic regression model, the race-by-distance interaction term of the logistic regression model was not significant (P = 0.919), and therefore only the main effects model was used. The likelihood ratio P value of the race effect, after adjusting for distance, was P = 0.11 (likelihood ratio

). The adjusted relationship of race and status was not statistically significant, but the direction of the point estimate and CI of the odds ratio for maternal death for African American to Caucasian race changed to 1.46 (95% CI, 0.73–3.01).


While African American women have been identified as being more likely to die during pregnancy by several independent researchers,6,8,9 our data do not support the hypothesis of racial disparity at UAB Hospital. There are several possible explanations for this rather interesting finding. First, the close proximity of African American patients may explain our findings. Second, national disparities in maternal mortality may reflect a women’s ability to access outpatient and inpatient medical care, rather than disparities of care within each hospital. State and nationwide mortality statistics that are based on death certificates have been recognized as being inaccurate,10,11 a shortcoming that we avoided by examining each maternal death in detail. This shortcoming of databases highlights the critical importance of an in-depth analysis of maternal deaths to get a better understanding of why women die. Third, the national disparity may reflect the fact that quality differs among hospitals rather than within hospitals. Hospitals that serve a predominantly African American population have been shown to have poorer outcomes for both African American and Caucasian patients when compared with hospitals that serve a predominately Caucasian population.12

The results of our secondary analyses give us some indication about the possible cause for this inequality because we have discovered that the distance of a woman’s residence to the hospital is a highly significant predictor of maternal death; women who died resided, on average, 90 miles from the hospitals, whereas women in the control group lived, on average, only 24 miles from the hospital (Table 1). The distance to the medical facility relates to 1 of 3 broad categories of impediments to receiving medical care described by Thaddeus and Maine who used a classic, pathways-based framework.13 Known as the “3 delays model,” it has been used extensively in studies of maternal mortality in developing countries.13–15 Phase 1 of this model describes delays in making the decision to seek medical care; phase 2 describes the process of identifying and reaching a medical facility; and phase 3 identifies delays in the receipt of appropriate care after having arrived at the medical facility. In addition to the distance to UAB Hospital, the lack of prenatal care was a significant predictor of maternal death in our study. Our findings suggest that we have to take a critical look at a potential delay in the decision to seek medical care and a delay in access to maternity care as the key impediments in accessing maternity care. If there are racial differences in these delays, these factors could also explain disparities in maternal mortality. Resources should be directed at identifying and preventing potential delays in the transfer of the very sick patient from a low-resource facility to a higher-acuity medical facility.

A few explanatory words on the observed difference of gestational age between cases and controls and the observed death proportions by race are warranted. Given that maternal death is a terminal event that can occur at any stage of the pregnancy, the observed larger proportion of low gestational age and decreased fetal survival in the group of maternal fatalities was expected. The observation of a higher rate of cesarean delivery in cases is also expected because in clinical practice the obstetrician would attempt to save the fetus by operative delivery in cases of impending maternal demise. Another interesting observation was the odds ratio estimate of 0.66 describing the odds of death for African American versus Caucasian race. While this estimate of race disparity was not statistically significant, the point estimate suggested lower odds of death for African American women. However, we also noted that African American women lived closer to the hospital and the point estimate of the odds ratio, after controlling for residential distance, was 1.46. Even after adjusting for distance, there was no statistically significant race disparity, but the direction of the point estimate changed to suggest higher odds of death for African American women. The point of discussion here is that the residential proximity had to be considered to get a better characterization of the odds of death by race.

A comparison of the causes of death between the CMACE and maternal deaths at UAB Hospital is provided in Table 2. Direct causes of death were 33% of maternal deaths at UAB Hospital versus 39% in the United Kingdom, and indirect causes of death contributed 55% at UAB Hospital versus 52% in the United Kingdom. Incidental deaths were 16% at UAB and 9% in the United Kingdom. The estimated maternal mortality ratio was 106 maternal deaths per 100,000 live births at UAB Hospital. The leading direct cause of maternal death in our 20-year-survey at UAB Hospital is related to preeclampsia and eclampsia, constituting 20% of all direct causes. By contrast, the leading direct cause of death in the United Kingdom, averaged over the years 1991 to 2008, has been thrombosis and thromboembolism, comprising 12% of cases, followed by preeclampsia and eclampsia (6%) and sepsis (5%).4 The observed difference in the incidence of thrombotic events reported at UAB Hospital and in the United Kingdom may be explained by the more consistent use of thromboprophylaxis at UAB Hospital. Another possible explanation is that thromboembolic causes of death are a category of sudden deaths that occur without warning signs, and it is likely that more of these types of deaths occur before a patient has entered the tertiary care hospital system. Thus, we may have missed these patients in the UAB Hospital analysis of maternal death. We observed a relatively small proportion of deaths due to maternal hemorrhage at UAB Hospital, which constitutes only 1% of cases at UAB Hospital (1990–2010), compared to 4% in the United Kingdom (1991–2008) and 36% in Africa and 30% in Asia.16 We attribute the low number of maternal deaths due to maternal hemorrhage at UAB Hospital to the round-the-clock availability of a multidisciplinary team of highly specialized health care providers, the immediate access to transfusion therapy, as well as surgical and other interventional treatment approaches that have been the cornerstone of most tertiary care centers in the United States.17 Causes of deaths that are unrelated to pregnancy are classified as incidental maternal deaths. In both databases, the majority of cases were due to motor vehicle accidents, but a significant number of cases have been attributed to domestic violence. The American College of Obstetricians and Gynecologists therefore recommends proper seat belt use by all pregnant women and screening for domestic abuse.18

Table 2
Table 2:
Direct and Indirect Causes of Maternal Deaths

An important topic of public health concern that may have an impact on maternal mortality is maternal obesity. In the most recent 2006–2008 CMACE report, 27% of women who died were obese (BMI > 30 kg/m2).4 At UAB, 40% (24/59) of women who died and 52% (77/146) of controls were obese, and there was no significant difference in BMI between the case and control groups. One factor to consider in this comparison is that the gestational age of maternal death cases was younger than that of their controls. The younger gestational age in the maternal death group may have biased the comparison. Reports on maternal death for the United States1 and the state of Alabama8 were not detailed enough to extract BMI information. The lack of an association between obesity and maternal deaths at UAB Hospital appears to contradict large-scale (>1 million patients) database reports suggesting that obesity is an independent risk factor for maternal deaths.19–21 This difference may be explained by the relatively larger proportion of high-risk pregnancies that are treated at a tertiary care facility (UAB Hospital) compared to a larger obstetrical population difference, which is also reflected by the relatively high maternal mortality ratio at UAB Hospital.

The impetus for performing a single-site case-control study was to overcome the problem of large studies that derive information from death certificates. However, this approach necessarily limits the sample size. A more complete list of covariates can be assessed in large-scale studies. This trade-off between data accuracy (single-center chart review study) and sample size (large-scale database review) should be considered when interpreting study findings.

In conclusion, we may revisit the social challenge posed above. At UAB Hospital, did we fail to address racial disparity in maternal deaths and default on the promise that all Americans will be treated equally? Data from this study provide insufficient evidence to support the hypothesis of racial disparity at our tertiary care medical center. We suggest that the next step towards understanding racial differences in maternal deaths reported in the United States should be directed at the health care delivery outside the tertiary care hospital setting and particularly at eliminating access barriers to health care for all women.


Name: Michael A. Frölich, MD, MS.

Contribution: This author helped in study design, conduct of the study, data collection, data analysis, and manuscript preparation.

Attestation: Michael A. Frölich approved the final manuscript.

Name: Catiffaney Banks, BS.

Contribution: This author helped in study design, conduct of the study, data collection, data analysis, and manuscript preparation.

Attestation: Catiffaney Banks approved the final manuscript.

Name: Amber Brooks, MD

Contribution: This author helped in conduct of the study, data collection, and manuscript preparation.

Attestation: Amber Brooks approved the final manuscript.

Name: Alethia Sellers, MD.

Contribution: This author helped in conduct of the study, data collection, and manuscript preparation.

Attestation: Alethia Sellers approved the final manuscript.

Name: Ryan Swain, PhD, MD.

Contribution: This author helped in conduct of the study, data collection, and manuscript preparation.

Attestation: Ryan Swain approved the final manuscript.

Name: Lauren Cooper.

Contribution: This author helped in conduct of the study, data collection, and manuscript preparation.

Attestation: Lauren Cooper approved the final manuscript.

This manuscript was handled by: Cynthia A. Wong, MD.


The authors would like to thank Mrs. Darleen H. Green for her invaluable contribution to this study.


a City Data Database. Available at: Accessed August 14, 2013.
Cited Here

b Google Maps. Available at: Accessed October 5, 2013.
Cited Here


1. Berg CJ, Callaghan WM, Syverson C, Henderson Z. Pregnancy-related mortality in the United States, 1998 to 2005. Obstet Gynecol. 2010;116:1302–9
2. Main EK, Menard MK. Maternal mortality: time for national action. Obstet Gynecol. 2013;122:735–6
3. 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
4. Cantwell R, Clutton-Brock T, Cooper G, Dawson A, Drife J, Garrod D, Harper A, Hulbert D, Lucas S, McClure J, Millward-Sadler H, Neilson J, Nelson-Piercy C, Norman J, O’Herlihy C, Oates M, Shakespeare J, de Swiet M, Williamson C, Beale V, Knight M, Lennox C, Miller A, Parmar D, Rogers J, Springett A. Saving Mothers’ Lives: reviewing maternal deaths to make motherhood safer: 2006–2008. The Eighth Report of the Confidential Enquiries into Maternal Deaths in the United Kingdom. BJOG. 2011;118(Suppl 1):1–203
5. Berg CJ, Callaghan WM, Henderson Z, Syverson C. Pregnancy-related mortality in the United States, 1998 to 2005. Obstet Gynecol. 2011;117:1229–30
6. Xu J, Kochanek KD, Murphy SL, Tejada-Vera B. Deaths: final data for 2007. Natl Vital Stat Rep. 2010;58:1–19
7. Lewis G. Saving Mothers’ Lives: the continuing benefits for maternal health from the United Kingdom (UK) Confidential Enquires into Maternal Deaths. Semin Perinatol. 2012;36:19–26
8. Chapman KCenter for Health Statistics Division of Statistical Analysis. . Alabama Department of Public Health (ADPH) Pregnancy-Related Mortality in Alabama 2005:1–16
9. Creanga AA, Berg CJ, Syverson C, Seed K, Bruce FC, Callaghan WM. Race, ethnicity, and nativity differentials in pregnancy-related mortality in the United States: 1993–2006. Obstet Gynecol. 2012;120:261–8
10. Lewis A. Improving accuracy in death certification, or, everybody dies of cardiopulmonary arrest. Tenn Med. 2012;105:37–8
11. Callaghan WM. Overview of maternal mortality in the United States. Semin Perinatol. 2012;36:2–6
12. Ly DP, Lopez L, Isaac T, Jha AK. How do black-serving hospitals perform on patient safety indicators? Implications for national public reporting and pay-for-performance. Med Care. 2010;48:1133–7
13. Thaddeus S, Maine D. Too far to walk: maternal mortality in context. Soc Sci Med. 1994;38:1091–110
14. Knight HE, Self A, Kennedy SH. Why are women dying when they reach hospital on time? A systematic review of the ‘third delay’. PLoS One. 2013;8:e63846
15. Combs Thorsen V, Sundby J, Malata A. Piecing together the maternal death puzzle through narratives: the three delays model revisited. PLoS One. 2012;7:e52090
16. Khan KS, Wojdyla D, Say L, Gülmezoglu AM, Van Look PF. WHO analysis of causes of maternal death: a systematic review. Lancet. 2006;367:1066–74
17. Walfish M, Neuman A, Wlody D. Maternal haemorrhage. Br J Anaesth. 2009;103(Suppl 1):i47–56
18. Brown HL. Trauma in pregnancy. Obstet Gynecol. 2009;114:147–60
19. Abbasi N, Balayla J, Laporta DP, Kezouh A, Abenhaim HA. Trends, risk factors and mortality among women with venous thromboembolism during labour and delivery: a population-based study of 8 million births. Arch Gynecol Obstet. 2014;289:275–84
20. Campbell KH, Savitz D, Werner EF, Pettker CM, Goffman D, Chazotte C, Lipkind HS. Maternal morbidity and risk of death at delivery hospitalization. Obstet Gynecol. 2013;122:627–33
21. Crane JM, Murphy P, Burrage L, Hutchens D. Maternal and perinatal outcomes of extreme obesity in pregnancy. J Obstet Gynaecol Can. 2013;35:606–11
© 2014 International Anesthesia Research Society