Four million infants are born each year in the United States, and the associated healthcare costs are substantial. In 2009, 7.6% of all hospital costs were attributable to maternity and newborn care, totaling over $27 billion.1 Almost half of childbirth-related hospital stays (47%) were covered by private health insurance; 45% of stays were billed to Medicaid programs.1 Maternity and newborn care is the top expenditure category for payments made to hospitals by both public payers and private health insurance companies.2 The average total costs of maternity (prenatal, labor and delivery, and postpartum) and newborn care for commercial payers was $27,866 for a cesarean delivery and $18,329 for a vaginal delivery in 2009.3 While payments by Medicaid programs were less overall, cesareans remain about 50% more costly than vaginal deliveries, at $13,590 for a cesarean delivery and $9131 for a vaginal delivery.3 Ensuring access to evidence-based, high-value care during childbirth is a clinical and financial imperative for healthcare providers, healthcare delivery systems, and health insurers.
A growing evidence base suggests that continuous labor support confers measurable clinical benefits to both mother and baby.4-6 Continuous labor support is the care, guidance, and encouragement provided by those who are with a pregnant woman in labor that aims to support labor physiology and mothers’ feelings of control and participation in decision making during childbirth.4 In a meta-analysis of randomized controlled trials, women who received continuous labor support reported greater satisfaction,7,8 had higher rates of spontaneous vaginal birth,9-11 higher infant Apgar scores,8 shorter labors,7,8 and lower rates of regional anesthesia (eg, epidural labor),12 cesarean deliveries,7,12 and forceps or vacuum deliveries.4,11,13 While many different individuals can and commonly do provide continuous labor support (including obstetric nurses, husbands and partners, close friends, and family members), the strongest results were achieved when continuous labor support was provided by someone who was not part of the woman’s family or social network or employed by the hospital.4
Doulas are trained professionals who provide continuous, one-onone emotional and informational support during the perinatal period. They are not medical professionals and do not provide medical services, but work alongside nurses, obstetricians, midwives, and other healthcare providers. A core function of the work of a doula is the provision of continuous labor support.14 Use of doula care is rising in the United States,4,15,16 but remains low: approximately 6% of women who gave birth in 2011 and 2012 reported receiving care from a doula.17 There are substantial barriers to access to doula care, especially for low-income women and women in minority communities. 5,6,15 The cost of birth doula services varies widely, but averages between $300 and $1200 and may include 1 or more prenatal or postpartum visits in addition to support during labor and birth.18,19 As health insurance programs do not typically offer coverage for these services,15 many women who would benefit from doula care are unable to access it.5,15,20 In addition, with a few notable exceptions (eg, HealthConnect One, International Center for Traditional Childbearing, and Everyday Miracles), most doulas are white upper-middle class women serving other white upper- middle-class women.15 These organizations employ doulas from underserved communities and also offer doula services to lower-income women and women of colot. The lack of diversity in the doula workforce is likely exacerbated by lack of third-party reimbursement and payment for doula care, further disadvantaging underrepresented groups who may be best served by a doula who shares their language, culture, or background.20
Women of color and low-income women are at greater risk of delivery-related complications and have higher rates of adverse birth outcomes than white, privately insured women.21 However, when low-income and women of color have access to doula care, they experience better outcomes than Medicaid recipients in general, with lower cesarean delivery rates and higher breastfeeding initiation rates.5,6 Recent research on the potential benefits of doula care, especially among low-income women, has ignited discussion regarding reimbursement of doula care by health insurance programs, including Medicaid programs. The state of Oregon has implemented a program for Medicaid coverage of birth doulas, and Minnesota passed legislation. in May 2013 that lays the groundwork for Medicaid reimbursement for trained doulas starting July 1, 2014.22,23
The goal of this study was to characterize women who used doula services and those who desired but could not access doula support among a representative sample of US childbearing women. We also explored the relationship between doula support, desire for doula support, and cesarean delivery, distinguishing nonindicated cesareans. If desire for doula services is related to higher rates of nonindicated procedures, this could serve to identify opportunities to better serve at-risk women who may benefit from access to continuous labor support.
Data are from the Listening to Mothers III (LTM3) survey, a nationally representative sample of women who gave birth to a single infant in a US hospital between July 1, 2011, and June 30, 2012 (N = 2400). The survey was commissioned by Childbirth Connection, funded by the Kellogg Foundation, and conducted online by Harris Interactive using validated procedures.17,24 Women aged 18 to 45 years who were participating in one of several online panels maintained by Harris Interactive formed the pool of potential respondents, with checks to ensure that each respondent only participated once. After data collection was complete, responses were weighted by propensity to be online as well as several demographic variables to enhance comparability with the national population of women who gave birth in 2010, the most recent year for which birth certificate data were available for this purpose.17
The Listening to Mothers surveys are the only nationally representative samples of childbearing women that contain information about doula care alongside selfreported clinical experiences, perceptions, and decisions about childbirth. In addition to asking whether a woman had support from a doula, the survey also asked about awareness of and level of familiarity with this type of care,
and whether women who knew about doula care would have wanted to have this type of care. The latter question is particularly useful as it may help at least partially address selection issues in who chooses to have a doula.
The 2 main predictors of interest were having doula support and, among those who did not have doula support but had a clear understanding of what a doula is, desire for doula support. Women were categorized as having doula support if they reported receiving supportive care during labor from a “doula or trained labor assistant.” Those who did not use doula support during labor were asked if they had heard of doulas and whether they had a clear understanding of this type of caregiver. Those with a clear understanding of doulas were then asked whether they would have liked to have doula support during their most recent birth; those who responded affirmatively were categorized as reporting “desire for doula support” in this analysis.
Measurement of cesarean birth was based on selfreported mode of delivery (vaginal or cesarean). Women with cesarean deliveries were asked to provide the main reason for the cesarean, which we categorized as a definitive medical indication for this procedure or a nondefinitive indication. We based these categorizations on professional standards used for accreditation measures25 and confirmation by our clinician co author (DKG). The following reported reasons for cesarean were considered definitive medical indications: baby being in the wrong position for birth, problems with the placenta, fetal monitor showing fetal distress during labor, and maternal health condition that called for cesarean delivery. All other reasons cited were categorized as being potential reasons, but not definitive medical indications for cesarean; these included prior cesarean, labor taking too long, provider concern regarding the size of the baby, fear of labor and vaginal delivery, being past the due date (for women whose pregnancies are <41.5 weeks gestation at delivery), having a narrow pelvis, or citing no medical reason for their cesarean. The term nonindicated cesarean refers throughout the manuscript to this type of delivery. Detailed information about the proportion of women with each of the reasons for cesarean delivery is provided in the eAppendix (available at www. ajmc.com). We conducted multiple sensitivity analyses around the classification of reasons for cesarean as medical indications, and results were substantively unchanged when we categorized any combination of the following reasons as definitive indications: labor taking too long, provider concern regarding the size of the baby, and having a narrow pelvis.
Sociodemographic covariates included age, race/ethnicity (white, black, Hispanic, or other/multiple race), education (high school or less, some college or associate’s degree, 4-year college degree, graduate education/ degree), 4-category census region (Northeast, Midwest, South, West), nativity (foreign- or US-born), partnership status at the time of the LTM3 survey (unmarried without partner, unmarried with partner, or married). Pregnancy characteristics included parity (first-time vs experienced mother), pregnancy intention (unintended pregnancy or not), agreement with the statement “birth is a natural process that should not be interfered with unless medically necessary,” and primary payer for maternity services (private, public [ie, Medicaid or other government programs], or none reported). We also conducted sensitivity analyses around the inclusion of control variables for labor support from a partner, spouse, family member, or friend, and results were robust to these specifications.
We first examined the descriptive statistics for the overall sample (N = 2400) with 1-way tabulation. We also explored doula care and desire for doula care (among those without access) by sociodemographic and pregnancy characteristics, using 2-way tabulation with x2 tests to identify significant differences. We then conducted multivariate logistic regression analyses to identify characteristics predicting use of and desire for doula care, and to estimate the adjusted odds of cesarean delivery overall (vs vaginal birth) and nonindicated cesarean delivery (vs vaginal birth) by use of doula support and desire for doula care.
We built 3 models to test these relationships: 1) comparing women with doula support to those who did not have doula support, 2) comparing women with doula support to those who expressed a desire for doula care but did not have a doula, and 3) among women who did not have doula support but did have a comprehensive understanding of this type of caregiver, comparing women who had an expressed desire for doula support with those who did not. All analyses were conducted using Stata v.12 and weighted to be nationally representative. This study was granted exemption from review by the University of Minnesota Institutional Review Board (Study Number 1011E92983).
Characteristics of the study population are reported in Table 1. Approximately 6% of women in the sample gave birth with doula support. Among those without doula support, 59% were aware of doula care; among women aware of doula care, 27% reported wanting a doula, but did not have one. Just over 30% of women in the sample had a cesarean delivery, and 10% of women with no definitive medical indication for a cesarean reported that they delivered via cesarean. Nearly half the sample had private health insurance coverage for their birth (45.5%). Other characteristics are broadly representative of the US childbearing population.
Table 2 reports doula support and desire for doula support by sociodemographic and pregnancy characteristics. A higher percentage of younger women (18-25 years) reported doula care, compared with women aged 35 and older (9.5% vs 1.9%). Younger mothers were also more likely to desire doula support, with 37.1% of women aged 18 to 24 years expressing this view, compared with 22.5% of women aged 35 and older. Having doula support did not differ significantly by race/ethnicity, but there were strong racial/ethnic variations in desire for doula support, with 21.6% of white women, 38.8% of black women, 29.8% of Hispanic women, and 43.5% of other/mixed race women reporting that they would have liked to have doula support. First-time mothers (vs experienced mothers) had higher rates of both doula support (8.8% vs 4.0%) and desire for doula support (33.5% vs 22.5%). While there were no differences in doula support by primary payer, there were significant differences in desire for doula support, with 39.3% of uninsured women and 32.6% of women with public coverage wanting doula support, vs 21.1% of privately insured women.
Multivariate logistic regression results for doula support and desire for doula care by sociodemographic and pregnancy characteristics are shown in Table 3. Adjusted odds largely reflect similar patters as the crude estimates presented in Table 2. Women with lower odds of doula support included: aged 25 to 29 years and over 35 years (vs aged 18-24 years) (AOR = 0.47, 95% CI, 0.24-0.91; and AOR = 0.19, 95% CI, 0.07-0.48), experienced mothers (vs first-time mothers) (AOR = 0.57, 95% CI, 0.34-0.98), and women whose pregnancies were unintended (AOR = 0.53, 95% CI, 0.28-0.99). Similar patterns emerged in predictors of desire for doula support: women aged 30 to 34 years (vs women aged 18-24 years) had lower odds of desiring doula care (AOR = 0.49, 95% CI, 0.28-0.84), as did experienced mothers (vs first-time mothers) (AOR = 0.67, 95% CI, 0.46-0.98). Factors associated with higher odds of desire for doula support were black race (vs white) (AOR = 1.77, 95% CI, 1.03-3.03), public or no health insurance coverage (vs private coverage) (AOR = 1.83, 95% CI, 1.17-2.85; and AOR = 2.01, 95% CI, 1.07-3.77), having a college degree (vs high school or less) (AOR = 1.79, 95% CI, 1.02-3.16), and having a planned cesarean delivery (AOR = 1.83, 95% CI, 1.14-2.93).
Table 4 presents the unadjusted (crude) and adjusted odds of cesarean delivery and cesarean without definitive medical indication by doula support and desire for doula support, controlling for sociodemographic and pregnancy-related characteristics. In each comparison, unadjusted results were similar in direction and magnitude to results from the adjusted models. Doula support was associated with a nearly 60% reduction in odds of cesarean delivery (AOR = 0.41, 95% CI, 0.18-0.96) and 80% lower odds of nonindicated cesarean delivery (AOR = 0.17, 95% CI, 0.07-0.39), compared with not having doula support. When comparing women who had doula support with those who indicated a desire for doula support but did not have it, women who had doula support had substantially lower odds of cesarean delivery overall (AOR = 0.31, 95% CI, 0.06-0.33) and of nonindicated cesarean delivery (AOR = 0.11, 95% CI, 0.03-0.36), compared with those who expressed a desire for doula care. Additionally, women who wanted doula support but did not have it had higher odds of cesarean delivery (AOR = 1.48, 95% CI, 1.00-2.19) and nonindicated cesarean delivery (AOR = 1.73, 95% CI, 1.10-2.73), compared with women who did not express a desire for doula support.
This analysis found that, among a nationally representative sample of US women who gave birth in 2011-2012, women with doula support had substantially lower chances of having a cesarean delivery and even lower rates of nonindicated cesarean, compared with women without support from a birth doula. This is consistent with prior research.4,5,26 However, prior observational research has noted the challenge of selection bias; that is, disentangling the desire for doula care from birth outcomes, given that measured and unmeasured characteristics associated with choosing a doula may also impact choices about delivery mode.27,28
A unique contribution of this analysis is that we are able to distinguish that doula support during labor and birth, not the desire for doula support, is associated with lower odds of nonindicated cesarean, compared with nonsupported births. Two key findings support this contribution: first, women who desired but did not have doula support had almost 50% greater chances of delivering via cesarean and more than 70% higher odds of having a nonindicated cesarean delivery, compared with women who did not desire doula care. This indicates that women who would like to have had a doula are not necessarily those who have fewer obstetric interventions, but that they may benefit from greater counseling and support before and during labor about the use of these interventions, especially when there is no definitive medical indication. Secondly, we show that the association between doula care and reduced chances of cesarean delivery and nonindicated cesarean delivery was relatively stable when comparing women with doula care to women who wanted but did not have doula care, who may be a more similar comparison group than women without doula care overall. Given the current clinical and policy focus on the potential maternal and neonatal risks of nondefinitively indicated caesarean deliveries,29,30 these findings have immediate and actionable implications.
There is a large unmet demand for doula care among American women, many of whom would likely benefit substantially from the evidence-based benefits associated with continuous labor support.4,15 Only 6% of women reported having support from a doula when they gave birth in 2011 or 2012, up from 3% of women in 2005.16 However, our findings indicate that over 40% of women are not aware of doula care, which translates into approximately 1.6 million women of the 4 million US women who give birth each year. Of those who are aware of what a doula is and the type of care they provide, 27% indicated that they would definitely want this type of support, which would mean an additional 1 million US women using doulas each year. Based on the findings from this analysis, if these women’s odds of nondefinitively indicated cesarean were lowered by 80% rather than elevated by 70%, the result could be an improvement in quality, safety, and a decrease in costs of childbirth. Identifying barriers to doula access is a crucial step in addressing this unmet need. While the survey data used in this analysis did not contain details on why women who wanted a doula did not have access to this service, prior research indicates several potential barriers and challenges; the most salient of which is concern about the out-of-pocket expense.5,15,20,22 Especially for families with low incomes or limited savings, doula services at costs ranging from several hundred to several thousand dollars,18 may be perceived as unaffordable in the context of other expenses related to childbirth and infant care (eg, car seats, diapers, feeding supplies) as well as changes such as loss of income during unpaid maternity leave.18,20 Additional barriers might include logistical challenges, such as distance from a doula for rural women, objections from husbands/partners or family members, or cultural issues, such as seeking but not finding a doula with a similar heritage or linguistic background.5,15,20
This analysis shows that 10% of women with no definitive medical indication for cesarean delivered by cesarean, representing potentially modifiable risks and costs. Cesarean delivery is more costly than vaginal birth (approximately $28,000 vs $18,000 for commercial payers), and 31.3% of US births in 2009 to 2011 were via cesarean delivery.31 From the perspective of a payer, including doula care as a covered benefit would require an investment in professional doula services, and the financial impact would depend on cesarean rates and risk factors in the covered population as well as reimbursement rates related to these services. However, the potential value for this investment is substantial. For example, while fees for doula care vary widely, they average around $1000, and with an approximate $10,000 mean difference between the cost of a vaginal and cesarean delivery, the decision to cover 10 doula-supported births would be cost-neutral if 1 nonindicated cesarean were avoided among these. Of course, continuous labor support is important for women who have cesarean deliveries and offers quantifiable benefits to these women as well.4 Further, the positive outcomes associated with doula support may accrue over time, so the financial rationale for insurance coverage of doula care is strong, especially since cost is a known barrier to access.5,15
Women who report that they would like to have doula care are the same women who stand to benefit most from the known effects of continuous labor support.4,5 Black women (vs white women), women with public health insurance (Medicaid and other government-funded programs which primarily serve low-income women, vs private insurance), and women without health insurance (vs those with private insurance) have higher risks of adverse birth outcomes, but are often least able to afford doula care or access culturally competent care.20 Our findings show that these same groups of women are more likely to report desiring but not having access to doula care, with limited resources being a likely explanation (although this is not directly assessed). While the associations identified in this analysis cannot be interpreted causally, our findings indicated that women who reported wanting a doula but not having one experienced higher cesarean rates than women who did not report wanting doula care, and lower rates than women who had a doula. This suggests that the association between doula support and lower cesarean rates is unlikely due to selection bias (ie, the idea that women who choose to have doulas are those who would have had lower rates of cesarean anyway), which is consistent with findings from randomized controlled trials.4 Our study extends these findings to a broader, nationally representative population. However, more and better data are needed to replicate these findings in a community and policy context. Facilitating access to doula care through health insurance benefits or coverage policies may be an opportunity for research on this topic, by utilizing randomization or staggered starts in implementation.
Not surprisingly, a majority of certified doulas (89.4%) believed that doula care should be reimbursed through health insurance,15 but there are real barriers to a wide implementation of reimbursement to a new category of services, especially services that are provided in a medical context but not by a healthcare professional. The state of Oregon has addressed this challenge by adapting language about reimbursement for nontraditional health workers to include trained, certified doulas.22
Our findings must be considered in light of limitations. First, the retrospective nature of the self-reported results carries the risk of recall and social desirability bias, particularly when women were asked whether they would have liked to have had a doula in their recent birth. Women’s actual birth experiences may have influenced their response to this question; also, the reasons that women desired but did not have a doula are not directly assessed. Second, while the LTM3 contains unique information about doulas and childbirth for a nationally representative sample of women, it is based on self-report, and does not include diagnostic or clinical data. As such, our categorization of medically indicated versus nonindicated cesarean sections was not confirmed by medical record data. However, we conducted extensive sensitivity analyses around these definitions, all of which produced consistent results. The survey was conducted online, though it uses validated methodologies and the weighted sample is consistent with data on the US childbearing population.17 Future prospective studies may help to examine this issue more fully.
Finally, sample size was limited, inhibiting our ability to detect smaller differences between groups. For example, the impacts of doula care for minority populations (eg, Native American or Asian women) or on less frequent outcomes (eg, preterm birth) could not be assessed in this sample because only several women may fall into these categories, which is not enough data to generate stable estimates. Nonetheless, this analysis provides the first nationally representative data comparing a quality-of-care outcome (cesarean without definitive medical indication) based on access to and reported desire for doula care. In summary, we found that women with doula support had lower odds of nonindicated cesareans compared with women without doula support and compared with women who desired but did not have doula support. Additionally, women who desired but did not have doula support had higher odds of cesarean without definitive medical indication, compared with those who did not desire doula care. These results, which should be confirmed by future prospective studies, suggest that increasing access to doula care for at-risk women who desire intrapartum doula support (eg, black, uninsured, or publicly insured women) may facilitate decreases in rates of nonindicated cesareans.