Maternal morbidity and mortality is increasing in the United States and varies by race, ethnicity, and geography.1,2 This rise is in stark contrast to other countries with similar economic resources. The leading causes of pregnancy-related deaths in the United States between 2011 and 2014 were cardiovascular disease, venous thromboembolic events, infection, and hemorrhage.1 Obstetric hemorrhage is the leading cause of maternal mortality worldwide and the fourth leading cause in the United States.3,4 It accounts for 11.5% of pregnancy-related deaths in the United States and is considered the most preventable maternal complication.1,5,6
Existing research on rural maternity care in the United States is limited but suggests that rural residence is a significant driver of disparity in outcomes.7 A “rural mortality penalty” exists among adults, where, since the 1980s, those living in rural areas have increasingly higher mortality rates at every age than those living in urban areas in the United States.8,9 Rural women have challenges to their health and well-being with higher rates of poverty and unemployment. They have more limited access to health care and often drive long distances to seek prenatal care and hospital maternity services.10 These women have a higher risk of hospitalization due to pregnancy complications and higher rates of preterm and low birth weight neonates.11 In an analysis of data from the National Center for Health Statistics, the overall infant mortality rates were higher in rural areas and lowest in large urban areas of the United States.12
Between 2006 and 2015, there was a 45% increase in the number of hospitalizations in the United States involving severe maternal morbidity defined by the Centers for Disease Control and Prevention criteria.13 Because maternal mortality is rare, the study of severe maternal morbidity is crucial to improving outcomes. It is estimated that severe maternal morbidity is 100 times more common than maternal mortality.14 The revised U.S. birth certificate now includes some important measures of maternal risk and adverse outcomes. The availability of data from the revised birth record provides an opportunity for analysis of a near-complete U.S. birth cohort to generate a national picture of adverse outcomes. Transfusion is the most prevalent indicator of severe maternal morbidity, contributing to more than half of the Centers for Disease Control and Prevention–reported rates.15 The primary objective of our study was to determine whether the rate of peripartum blood transfusion for women in rural areas of the United States is different from their urban counterparts. We hypothesized that given their sociodemographic profile and less access to health care, women living in rural areas would have higher rates of transfusion around the time of delivery.
We performed a population-based retrospective cohort study using national vital statistics data to examine whether rates of blood transfusion during the peripartum period were higher in rural areas compared with more urban areas of the United States. The U.S. standard certificate of live birth was revised in 2003 to collect a set of maternal morbidity indicators including maternal transfusion, admission to intensive care unit, and unplanned surgical procedures after delivery.16 The National Center for Health Statistics made these data available for analysis in 2011. We obtained the 2014–2016 natality files through a data use agreement with the National Association for Public Health Statistics as compiled data provided by the 57 vital statistics jurisdictions through the Vital Statistics Cooperative Program. We linked individual records to the 2013 National Center for Health Statistics Urban-Rural Classification Scheme for Counties using the federal information processing code for mother's resident and delivery county.17 The National Center for Health Statistics scheme, outlined in Table 1, categorizes each county into one of six groups: large metropolitan-center, large metropolitan-fringe, medium metropolitan, small metropolitan, micropolitan, and noncore. The National Center for Health Statistics classification is based on the 2013 Office of Management and Budget delineations of metropolitan or micropolitan statistical areas, with an adjustment from 2012 postcensus estimates of the U.S. population. The classification scheme takes into account population size and whether the county is part of a metropolitan core. We studied the effect of both the county of residence and county of delivery because primary residence influences risks for transfusion that exist because of individual health status or health conditions, whereas county of delivery is more likely to be related to health care services. The study was deemed exempt by The University of Wisconsin Institutional Review Board because it does not involve human subjects as defined under federal regulation, CFR 46.102(e) (I).
We used 2014–2016 data because the vast majority of states (47 states and Washington, DC) were using the 2003-revised birth certificate by 2014. However, the transition for all 50 states was not completed until 2016. The three states still using the previous version as of 2014 were New Jersey, Rhode Island, and Connecticut. We excluded these three states from our primary analysis and subsequently conducted a sensitivity analysis including their available data. Mothers who delivered in the United States but did not reside in the 50 states or Washington, DC, were excluded because of missing information on their county of residence. Similarly, mothers who lived in the United States but delivered outside of the United States or Washington, DC, were excluded given the lack of information about county of delivery. We also excluded births of extreme size (the upper and lower 1% of gestational age-adjusted birth weight) to take into account possible mistakes in birth records. We included only hospital births. We further limited our primary analysis to mothers with nulliparous, term (37 weeks of gestation or greater), singleton, vertex births. The nulliparous, term, singleton, vertex cohort was chosen in an effort to define a comparable group of patients living and seeking maternity care across the rural–urban continuum.
In addition to the maternal county of residence and delivery, we also examined descriptive statistics for an array of control variables, including maternal socioeconomic and demographic characteristics, maternal health risks, and peripartum factors. The maternal socioeconomic and demographic variables we assessed were maternal age (in 5-year categories), race or ethnicity (non-Hispanic white, non-Hispanic black, Hispanic, non-Hispanic Native American and non-Hispanic Alaskan Native, and Asian American and Pacific Islander), nativity (mother foreign born or not), educational level (less than high school, high school graduate, some college or Associates degree, college degree or more), marital status (currently married or not), method of payment (private insurance, Medicaid, self-pay, and no insurance), and timing of initiation of prenatal care (first trimester or not). The maternal health risk factors included cigarette use during pregnancy, prepregnancy body mass index (BMI, calculated as weight in kilograms divided by height in meters squared; underweight [less than 18.5], normal weight [18.5–24.9], overweight [25–29.9], obese class I [30–34.9], obese class II [35–39.9], obese class III [40 and higher], diabetes, gestational diabetes, chronic hypertension, and pregnancy-related hypertension. Peripartum factors controlled for in the analysis were labor induction, labor augmentation, antibiotic use, chorioamnionitis, use of epidural, mode of delivery (cesarean, vaginal–forceps, vaginal–vacuum, and spontaneous vaginal), birth attendant (physician or others), and neonatal birth conditions such as postterm delivery and birth weight. We calculated the blood transfusion rate for each category of the variables, including the maternal county of residence and the county of delivery.
We then used step-wise logistic regressions to examine the associations between urbanicity and blood transfusion by entering groups of control and explanatory variables. Residential urbanicity was first entered and then included in all models of multivariate analyses. In the second model, we added maternal socioeconomic and demographic variables. The third model added maternal health factors and the fourth model incorporated labor and delivery conditions. In the fifth and final model, we added county of delivery.
We conducted four sensitivity analyses to determine whether the findings were affected by sample selection. First, we performed a sensitivity analysis for the entire U.S. birth cohort. Next, we conducted an analysis of the nulliparous, term, singleton, vertex cohort after removing the states of Hawaii and Alaska given their unique geographical locations and populations that might influence mobility between county of residence and delivery county. Then, we added back the available data from the three states that were still using the 1989 birth certificate in 2014–2016. In the last sensitivity analysis, we expanded the nulliparous, term, singleton, vertex cohort to include multiparous and nonvertex singleton pregnancies at term. All statistical analyses were conducted using STATA 15.
Between 2014 and 2016, there were 11,943,020 live births in 50 U.S. states and Washington, DC. Of these births, 11,912,448 were to mothers residing in the United States. The three states still using the previous birth certificate version as of 2014 were excluded from our analysis, which left 11,462,709 births, representing 96% of the live births in the United States from 2014 through 2016. After all the exclusions were applied, the final analytic sample in the nulliparous, term, singleton, vertex cohort included 3,346,816 women (Fig. 1). The majority of the cases were excluded based on prematurity, multiparity, multi-fetal gestation, or nonvertex presentation.
Demographic information, clinical characteristics, and transfusion rates for the nulliparous, term, singleton, vertex study cohort are outlined in Table 2. Their overall blood transfusion rate was 3.01 per 1,000 live births. The transfusion rate based on maternal county of residence across the National Center for Health Statistics geographic categories increased as the counties became more rural: large metropolitan-center (1.9/1,000 live births); large metropolitan-fringe (2.4); medium metropolitan (2.6); small metropolitan (2.6); micropolitan (4.5); and noncore rural area (5.3) (Fig. 2). The relationship was also noted when we examined the transfusion rates based on delivery county. The rate of 8.8 per 1,000 for women delivering in hospitals in the most rural counties was notable.
The positive association between transfusion rates and delivery in a rural county warranted a further examination of transfusion rates for mothers who did not deliver in their county of residence. Table 3 displays the rates for the 922,493 (27.6%) women who delivered in a county different from their county of residence. Some of this movement happened across counties at the same level of urbanicity; however, many of these women delivered in a more urban county, and some delivered in a less urban county. More than half (57%) of the women living in large metro-fringe counties delivered in a large metro-center county. Women living in medium metropolitan counties changed urbanicity level the least with 76% delivering in a medium metropolitan county. For women in the more rural counties, the majority delivered in a more urban county: 88% for women living in rural counties and 66% for women living in micropolitan counties. Women in all National Center for Health Statistics geographic categories had a higher rate of transfusion if they delivered in a more rural county than their county of residence. Women who lived in a rural county had a transfusion rate of 8.6 per 1,000 if they delivered in a noncore rural county, compared with only 2.5 per 1,000 if they delivered in a large metropolitan-center county. Women who lived in a large metropolitan-center county and delivered in a noncore rural county had the highest blood transfusion rate (18.5/1,000 live births), though this represents only a small portion of the women (0.1%).
The results of the step-wise multivariable regression analysis that controlled for factors that affect blood transfusion rates are outlined in Tables 4 and 5. Model 1 included only residential urbanicity based on the county of residence, model 2 added maternal background, model 3 incorporated antepartum health, and model 4 included labor and delivery conditions. In model 4, after adjusting for all covariates except county of delivery, the odds of transfusion remained higher among women living in micropolitan (adjusted odds ratio [aOR] 2.25, 95% CI 2.09–2.43) and noncore rural (aOR 2.59, 95% CI 2.38–2.81) counties when compared with women living in large metro-center counties. After adjusting for county of delivery (model 5), the odds of blood transfusion decreased for women living in micropolitan (aOR 1.39, 95% CI 1.19–1.63) and noncore rural (aOR 1.32, 95% CI l.12–1.55) counties, providing information that the driver of the increased rates of transfusion was more associated with county of delivery than county of residence.
The four sensitivity analyses showed similar results (Appendix 1, available online at http://links.lww.com/AOG/B744). The model fit when we analyzed the entire population including multiparous women and those with multi-fetal gestations, breech presentation, and preterm births. Including these women did not affect the results. When the three states with missing data (New Jersey, Rhode Island, and Connecticut) were added back in to the nulliparous, term, singleton, vertex restricted model, the aORs were unchanged, likely because this was only a small percentage of the total cohort. The same was true when we excluded Alaska and Hawaii from the regression analysis. The results were also the same when we only excluded preterm and multi-fetal gestations.
The need for transfusion of blood products is a known complication of pregnancy and is an indicator of maternal morbidity.14 A recent Healthcare Cost and Utilization Project 2018 Statistical Brief reported trends and disparities in U.S. delivery hospitalizations from 2006 to 2015.13 In it, the authors noted that blood transfusion was the most common form of severe maternal morbidity, performed in approximately 80% of the cases of severe morbidity. Our study results support our initial hypothesis that rural women may have higher transfusion rates. The magnitude of the difference we found was surprising. In our study, the odds of peripartum blood transfusion were higher for women living and delivering in more rural settings than those in more urban settings.
We considered several explanations for the increased rates of perinatal blood transfusion noted in this study. The results may be related to individual characteristics unique to rural women such as higher rates of antenatal anemia for women living or delivering in rural areas. Though we did not have data on the number of women with anemia, our analyses suggest that the higher transfusion rates were not directly related to antepartum anemia. Young age, nonwhite race, low income, and lack of insurance are all associated with iron deficiency anemia.18 Although many of the conditions that predispose to anemia cluster in rural areas, these individual-level factors did not explain the differences in transfusion rates in our study.
The majority of blood transfusions are a result of acute blood loss, and occur during the delivery admission, a time of significant risk for all pregnant women.18,19 We evaluated many of the individual maternal characteristics known to affect postpartum hemorrhage rates, including age, preeclampsia, chorioamnionitis, operative vaginal delivery, and cesarean delivery.20 Our multivariable analysis included socioeconomic and demographic factors associated with antepartum anemia as well as maternal health factors and characteristics of labor and delivery related to hemorrhage risk. These individual risk factors did not account for the transfusion differences seen in rural women.
Controlling for county of delivery, however, influenced the rate of transfusion, eliminating differences in transfusion rates across county of residence. This suggests that something at the health system level is associated with transfusion rates. A possible explanation is that increased rates of transfusion in rural birth facilities are a result of practice patterns such as a lower threshold for transfusion or less prompt diagnosis and management of postpartum hemorrhage. A prior study of maternal morbidity in low-risk pregnancy revealed that women delivering in rural, low-volume hospitals had higher rates of postpartum hemorrhage than women delivering in higher volume hospitals.10 The rural women in our study living and delivering in a rural county had more transfusions (8.5/1,000 live births) than women living in a rural county and delivering in a more urban one (2.5/1,000). Obstetric hemorrhage emergencies can occur without warning in pregnancy. Rural hospitals have maternity care workforce limitations and fewer resources such as dedicated obstetric nurses and extensive blood supply. Some smaller hospitals have only an “emergency supply” of blood products that could limit availability of blood products or the timely administration of transfusion. The availability of resources may influence practice patterns that ultimately result in the higher rates of transfusion observed in our study, because some providers may transfuse at a lower threshold to “stay ahead.”
The conclusion that health system factors seem to account for the higher transfusion rates is not totally unexpected. Access to maternity care for rural women is rapidly declining in the United States. In one study, the number of counties with hospital-based obstetric delivery services decreased by 9% from 2004 to 2014 and more than half of rural women travel more than 30 minutes to reach hospital maternity services.21 Only 9% of U.S. physicians practice in rural communities, 20% of the U.S. population lives in rural areas, and half of the rural U.S. counties have no practicing obstetrician gynecologist.11 Rural communities are burdened by both issues of access to maternity care as well as workforce problems.7,21 Disparities in maternal outcomes based on race have appropriately received significant attention in recent years. Maternity care issues in rural communities need similar focus and attention, because these communities face their own unique challenges.
This study makes a unique contribution to the existing literature on maternal morbidity among rural women in the United States, by using a national sample. A strength of the study is the large number of cases included and the uniform reporting of information on the birth certificate. However, this analysis is also limited in some ways by these data. Our analysis relies on the accuracy of birth record data with the inherent quality problems associated with any large administrative data set. Although consistent with some reports, the overall transfusion rate noted in our study is less than that of a study relying on detailed hospital records.22 In addition, some potentially significant clinical variables, such as preexisting anemia, are not recorded on the revised birth record. Also, the circumstances around delivering outside of one's county of residence were not clear and may be related to their risk of hemorrhage and requiring transfusion. Finally, our study may have actually underestimated the effect of delivering in a rural hospital on the risk of hemorrhage and blood transfusion as some smaller rural community hospitals have limited blood bank availability, so transfusion is not possible and may have occurred after maternal transport to a higher acuity setting after delivery.
In summary, our results demonstrate that the rate of blood transfusion is higher for women who live and deliver in noncore rural and micropolitan counties compared with those in more urban settings. Rural–urban differences in transfusion appear to be related to differences in maternity care practice rather than individual maternal characteristics. The study results are associations and do not prove that residence or delivery in a rural county caused the higher transfusion rate. Further research focusing on maternal health care delivery and obstetric outcomes in rural communities is needed to identify opportunities for improvement in care.
1. Creanga AA, Syverson C, Seed K, Callaghan WM. Pregnancy-related mortality in the United States, 2011–2013. Obstet Gynecol 2017;130:366–73.
2. MacDorman MF, Declercq E, Thoma ME. Trends in maternal mortality by socio-demographic characteristics and cause of death in 27 states and the District of Columbia. Obstet Gynecol 2017;129:811–8.
3. Postpartum hemorrhage. Practice Bulletin No. 183. American College of Obstetricians and Gynecologists. Obstet Gynecol 2017;130:168–86.
4. Chandraharan E, Krishna AJB. Diagnosis and management of postpartum haemorrhage. BMJ 2017;358:j3875.
5. Berg CJ, Harper MA, Atkinson SM, Bell EA, Brown HL, Hage ML, et al. Preventability of pregnancy-related deaths: results of a state-wide review. Obstet Gynecol 2005;106:1228–34.
6. Main EK, McCain CL, Morton CH, Holtby S, Lawton ES. Pregnancy-related mortality in California: causes, characteristics, and improvement opportunities. Obstet Gynecol 2015;125:938–47.
7. Kozhimannil KB, Hardeman RR, Henning-Smith C. Maternity care access, quality, and outcomes: a systems-level perspective on research, clinical, and policy needs. Semin Perinatol 2017;41:367–74.
8. James WL. All rural places are not created equal: revisiting the rural mortality penalty in the United States. Am J Public Health 2014;104:2122–9.
9. Singh GK, Siahpush M. Widening rural–urban disparities in all-cause mortality and mortality from major causes of death in the USA, 1969–2009. J Urban Health 2014;91:272–92.
10. Kozhimannil KB, Thao V, Hung P, Tilden E, Caughey AB, Snowden JM. Association between hospital birth volume and maternal morbidity among low-risk pregnancies in rural, urban, and teaching hospitals in the United States. Am J Perinatology 2016;33:590–9.
11. Health disparities in rural women. Committee Opinion No. 586. American College of Obstetricians and Gynecologists. Obstet Gynecol 2014;123:384–8.
12. Ely DM, Driscoll AK, Matthews T. Infant mortality rates in rural and urban areas in the United States, 2014. NCHS data brief, no 285. Hyattsville, MD: National Center for Health Statistics; 2017.
13. Fingar K, Hambrick M, Heslin K, Moore J. Trends and disparities in delivery hospitalizations involving severe maternal morbidity, 2006–2015: HCUP statistical brief# 243. Rockville, MD: Agency for Healthcare Research and Quality; 2006.
14. Callaghan WM, Creanga AA, Kuklina EV. Severe maternal morbidity among delivery and postpartum hospitalizations in the United States. Obstet Gynecol 2012;120:1029–36.
15. Main EK, Abreo A, McNulty J, Gilbert W, McNally C, Poeltler D, et al. Measuring severe maternal morbidity: validation of potential measures. Am J Obstet Gynecol 2016;214:643.e1–10.
16. Osterman M, Martin JA, Mathews TJ, Hamilton BE. Expanded data from the new birth certificate, 2008. Natl Vital Stat Rep 2011;59:1–28.
17. Ingram DD, Franco SJ. 2013 NCHS urban-rural classification scheme for counties. National Center for Health Statistics. Vital Health Stat 2014;2:1–73.
18. Anemia in pregnancy. ACOG Practice Bulletin No. 95. American College of Obstetricians and Gynecologists. Obstetrics Gynecol 2008;112:201–7.
19. Mei Z, Cogswell ME, Looker AC, Pfeiffer CM, Cusick SE, Lacher DA, et al. Assessment of iron status in US pregnant women from the national health and nutrition examination survey (NHANES), 1999–2006. Am J Clin Nutr 2011;93:1312–20.
20. Kramer MS, Berg C, Abenhaim H, Dahhou M, Rouleau J, Mehrabadi A, et al. Incidence, risk factors, and temporal trends in severe postpartum hemorrhage. Am J Obstet Gynecol 2013;209:449.e1–7.
21. Hung P, Henning-Smith CE, Casey MM, Kozhimannil KB. Access to obstetric services in rural counties still declining, with 9 percent losing services, 2004–14. Health Aff 2017;36:1663–71.
22. Ehrenthal DB, Chichester ML, Cole OS, Jiang X. Maternal risk factors for peripartum transfusion. J Womens Health 2012;21:792–7.