Abstract
Objectives. To assess the relationship between doula utilization and health outcomes of females enrolled in Medicaid-affiliated plans in the United States.
Methods. In this retrospective, observational cohort study, we used Medicaid claims data from a national health insurer to compare health outcomes between females who used and who did not use a doula (2014–2023). We conducted propensity score matching using a 1:1 case‒control match, without replacement, and fit logistic regressions to analyze the relative risks for maternal health outcomes.
Results. The study population included 722 matched pairs with and without a doula. Results indicate females with doulas had a 47% lower risk of cesarean delivery and a 29% lower risk of preterm birth, and were 46% more likely to attend a postpartum checkup (all differences P < .05).
Conclusions. Doula care is associated with improved health outcomes among Medicaid enrollees.
Public Health Implications. Doulas have garnered increasing interest from policymakers as a strategy to address increasing trends in maternal morbidity and persistent health disparities. This study provides evidence from Medicaid enrollees across the United States that doula care can improve maternal health. (Am J Public Health. 2024;114(11):1275–1285. https://doi.org/10.2105/AJPH.2024.307805)
Health insurance coverage for birth doulas in the United States has recently gained traction at the federal level as a strategy for improving persistent challenges in maternal health, such as increasing trends in maternal morbidity and mortality and maternal health disparities.1–3 The Surgeon General recommended in their 2020 Call to Action coverage of doulas to help broaden access to quality care.4 Similarly, the White House recommended coverage for doula services to help expand the perinatal workforce to address provider shortages and increase provider diversity in its 2022 Blueprint for Addressing the Maternal Health Crisis.5
According to DONA International, a doula is “a trained [nonclinical] professional who provides continuous physical, emotional, and informational support … before, during, and shortly after childbirth to help them achieve the healthiest, most satisfying experience possible.”6 Evidence thus far links doulas with decreased rates of cesarean deliveries, birth complications, postpartum depression and anxiety, low birth weight, and preterm birth.7–9
Some research suggests that integrating doulas into maternal health care could also help reduce maternal health disparities.10 Black women have 3 to 4 times higher risk of maternal mortality, relative to White women, and are more than twice as likely to experience severe maternal morbidity, regardless of the amount of prenatal care received, socioeconomic status, insurance type, or preexisting conditions.10,11 Doulas may help mitigate maternal health disparities by facilitating the delivery of culturally relevant care and helping empower their patients to advocate for themselves.11–13
Although research linking doulas to improved maternal health appears promising, the evidence base on doulas in the United States is just being developed. Data collected through administrative claims has only recently become available because of the passage of legislation in several states allowing for health insurance coverage of doulas through Medicaid. In 2018, for example, only Minnesota and Oregon provided Medicaid coverage for doula care. As of May 2024, however, 14 states and Washington, DC, provide coverage for doula services through Medicaid.14
Most previous evaluations of doula care, therefore, come from a single hospital or within a single state, affecting the broader generalizability of study findings. Other methodological challenges include self-reported data, which introduce self-reporting bias into analyses; small sample sizes, which can impede the ability to discern statistical significance on maternal health outcomes that are not highly prevalent; and a lack of standardization of how or whether a doula was trained and certified, which impacts the validity of inferences made about doulas.10
The objective of this study was to provide insight on the relationship between doula care covered through Medicaid managed care and health outcomes. This study addresses limitations of previous studies by using an adjudicated source of clinical claims and doula data, including into analysis only certified doulas, and analyzing a geographically diverse, multistate study population. This study also provides insight into how doula care affects race disparities in maternal health through race-specific analysis on the relationship between doula utilization and maternal health.
A unique contribution of this study is the analysis of area-level measures, including number of hospitals in a county providing obstetric care (a measure of health care access), neighborhood-level socioeconomic status (an indicator of socioeconomic disadvantage), and infant mortality rate (a sensitive measure of the overall health of a population).15,16 Integration of these area-level measures into analyses is important because maternal health is affected not only by individual-level factors but also through area-level influences.17 The relationship between doula care and maternal health outcomes, accordingly, is impacted by the community in which someone lives. These area-level factors, however, have been minimally addressed in earlier studies.
In this retrospective cohort study, we used administrative claims data from the Healthcare Integrated Research Database, a proprietary repository of medical and pharmacy claims data from more than 88 million individuals enrolled in health plans across the United States.18 A limited data set was used that was stripped of individual member identifiers to comply with the laws and regulations included in the Health Insurance Portability and Accountability Act (HIPAA) and HIPAA Privacy Rule (45 CFR 164.514[e][3]i).
The study population included self-reported females with at least 1 pregnancy outcome diagnosis or procedure code (International Classification of Diseases, Ninth Revision [ICD-9; Geneva, Switzerland: World Health Organization; 1980] and International Classification of Disease, Tenth Revision [ICD-10; Geneva, Switzerland: World Health Organization; 1992]) and at least 1 day of Medicaid eligibility between January 1, 2014, and June 30, 2023. (See Table A, available as a supplement to the online version of this article at https://ajph.org, for definitions and lists of relevant ICD-9, ICD-10, Current Procedural Terminology, and Healthcare Common Procedure Coding System codes.) These years were selected based on data availability (doula utilization data were available in at least 1 state starting in 2014).
We excluded members who were not of reproductive ages (15 to 49 years) from the study. The study population was further restricted to members in 9 states (California, New Jersey, Missouri, Texas, Wisconsin, and 4 other states located in different US regions) where doula care was provided (either as a legislated benefit or alternatively funded via pilots, grants, etc.; Figure 1). Some states could not be specified because of compliance rules that prohibit us from identifying specific states without their approval (these states were contacted, but they did respond to our request).
FIGURE 1— Study Population Attrition of Women Enrolled in Medicaid-Affiliated Plans in the United States: 2014–2023
aThe study population was restricted to members in 9 states (CA, NJ, MO, TX, WI, and 4 other states located in different US regions; some states could not be named because of health plan compliance rules that prohibit us from identifying specific states without their approval.) where doula care was provided (either as a legislated benefit or alternatively funded via pilots, grants, etc.).
Health plan members received access to doula care through a few different approaches. Some individuals requested a doula of their own volition, while others received access because they were identified as high risk through their health plan’s care management teams. Alternatively, some individuals were connected to doulas as a result of community outreach. Specific communities outreached to about doulas varied by state. Programs in some states, for example, focused outreach on Black women, while programs in other states focused outreach on orthodox Jewish women, adolescent pregnancies, women in rural settings, or women experiencing substance use disorder.
Health plan members who received doula care were identified in claims through 3 strategies. In states with Medicaid coverage of doulas, doula utilization was identified through the Centers for Medicare & Medicaid Services taxonomy code for doula services (374J00000X). Unique National Provider Identifier codes for doulas also were used to identify members who used doulas. In states where Medicaid does not yet cover doulas, we received masked identifier codes of members who received doula care from Elevance Health‒affiliated health plans. Representatives from all affiliated health plans confirmed that the number of members identified as receiving doula care was consistent with their knowledge of doula utilization in their state. All doulas were required to be trained and certified, although states varied on who provided the certification.
Maternal health characteristics and outcomes were evaluated during pregnancy (from estimated start of pregnancy until the day before delivery), at delivery, and during the postpartum period—from the day after birth through 84 days following birth (i.e., 12 weeks postpartum or the “fourth trimester”)—when many physiological and psychological changes occur as individuals recover from childbirth).19 The start of pregnancy was estimated by subtracting the gestational age (identified from ICD-10 Z3A codes) from the date of the pregnancy outcome. In rare instances when no Z3A codes were identified on the date of the pregnancy outcome, the last or closest Z3A code was used.
Health outcomes included cesarean delivery, vaginal birth after cesarean delivery, preterm birth, emergency department (ED) visit within 30 days of delivery, inpatient admission within 30 days of delivery, attendance at a postpartum visit between 7 and 84 days after delivery, prevalence of postpartum depression and anxiety, and prevalence of severe maternal morbidity, which is unintended outcomes of labor and delivery that can significantly impact women’s health. These outcomes were selected because they are relatively more prevalent indicators of maternal morbidity and because of previous evidence suggesting their association with doula care. We did not evaluate maternal mortality, because we lacked statistical power to evaluate the small number of maternal deaths that occurred. We also did not assess the relationship between doula care and prenatal visit adherence given that many women did not learn about the availability of doula services until after they had attended a prenatal visit.
We used propensity score matching (PSM) because it allows an observational, unrandomized study to simulate a randomized control trial to enable causal inference.20 The advantages of PSM include that it helps reduce the potential impact of selection bias,21 and it is an effective methodology when the study population has a large pool of “unexposed” individuals (without doulas) compared with those “exposed” (with doulas).22
We estimated the propensity score for members receiving doula care with a multivariable logistic regression model that incorporated the following sociodemographic and clinical variables: age at the time of pregnancy outcome, race/ethnicity (identified through self-report in either electronic health records or patient enrollment files or derived from the RAND imputation algorithm),23 state of residence, neighborhood socioeconomic status (Agency for Healthcare Research and Quality‒validated socioeconomic status index),24 National Center for Health Statistics Urban‒Rural Classification Scheme for Counties (developed to assess associations between urbanization level of residence and health),25 and the presence of any one of the following commonly occurring pregnancy complications: gestational diabetes, gestational hypertension, pre-eclampsia or eclampsia, anemia, placental abruption, thrombocytopenia, placenta previa accreta spectrum disorder, short cervix, or infection. Cases (with doula) were matched to the controls (without doulas) using a 1:1 case‒control match greedy algorithm, without replacement. Only 1 pregnancy per female (their most recent was selected for consistency purposes) was included into the matching algorithm so that an individual could not be matched more than once. Given the large pool of controls, this method tends to result in estimates with minimal bias and less variance relative to other matching approaches.26
We assessed the quality of the match by analyzing the standardized mean differences between the baseline characteristics of the cases and controls. The 2 groups were considered appropriately balanced with standardized differences less than 0.10. The models were further adjusted by year of delivery outcome and the number of hospitals providing obstetric care in a county. The models were not adjusted for individual comorbid conditions (which in some cases had standardized differences > 0.10) because the overall composite pregnancy complication measure was well balanced.
We conducted a series of subgroup analyses on the significant findings from the full PSM analyses. Specifically, PSM analyses were stratified by race (Black vs White), because of well-established maternal health disparities among Black women. Low sample sizes of other race/ethnicity groups prohibited further analyses. PSM analyses also were conducted by area-level infant mortality rate (IMR; counties in the highest quartile of IMR vs lowest quartile). Results were stratified by IMR because it is an indicator of maternal and infant health and of the socioeconomic and environmental conditions where people live.15 Similar PSM methods were used for the subgroup analysis as for the whole study population, except the PSM models stratified for Black females and White females did not match on race.
We identified 1 094 874 pregnancies insured through an affiliated Medicaid plan between January 2014 and June 2023. Of these pregnancies, 869 were assisted by doula, most (69%) of which occurred between 2020 and 2023. The majority of doula-supported pregnancies occurred during this window because state legislation enabling Medicaid reimbursement for doulas in most states occurred after 2020.27
Recipients of doula services resided in more than 350 different zip codes in 9 states, although some states had more doula recipients than others. In 2 states, for example, we identified fewer than 10 members who received doula care, while in 3 states, we identified more than 100 members who received doula care (Table 1). Variance in the distribution of doula recipients was largely attributable to differences in the length of time doula care has been available to Medicaid enrollees in those states.
TABLE 1— Baseline Characteristics of Women Enrolled in Medicaid-Affiliated Plans in 9 States Across the United States, Unmatched Sample: 2014–2023
| Unmatched Sample | ||||
| Doula (n = 869), Mean ±SD or No. (%) | No Doula (n = 1 094 005), Mean ±SD or No. (%) | Pa | Standardized Biasb | |
| Age, y | 28.0 ±5.4 | 27.2 ±6.1 | < .01 | 0.1 |
| Statec | ||||
| 1 | 74 (8.5) | 236 316 (21.6) | < .01 | 0.4 |
| 2 | 7 (0.8) | 53 460 (4.9) | < .01 | 0.3 |
| 3 | 130 (15.0) | 103 450 (9.5) | < .01 | 0.9 |
| 4 | 35 (4.0) | 27 419 (2.5) | < .01 | 0.1 |
| 5 | 401 (46.1) | 103 450 (9.5) | < .01 | 0.9 |
| 6 | 70 (8.1) | 318 956 (29.2) | < .01 | 0.6 |
| 7 | 28 (3.2) | 47 944 (4.4) | .1 | 0.1 |
| 8 | 122 (14.0) | 122 623 (11.2) | < .01 | 0.1 |
| 9 | 2 (0.2) | 97 466 (8.9) | .004 | 0.1 |
| Urbanicity | ||||
| Urban | 575 (66.2) | 574 474 (52.5) | < .01 | 0.3 |
| Suburban | 223 (25.7) | 374 627 (34.2) | < .01 | 0.2 |
| Rural | 9 (1.0) | 85 178 (7.8) | < .01 | 0.3 |
| Unspecified | 62 (7.1) | 59 726 (5.5) | .03 | 0.7 |
| Patient’s race/ethnicity | ||||
| Asian/Pacific Islander | 36 (4.1) | 65 896 (6.0) | .02 | 0.1 |
| Black | 379 (43.7) | 244 835 (22.4) | < .01 | 0.5 |
| Hispanic/Latino | 148 (17.0) | 405 088 (37.0) | < .01 | 0.5 |
| White | 261 (30.0) | 306 771 (28.0) | .19 | 0.0 |
| Other | 33 (3.8) | 68 415 (6.3) | .7 | 0.0 |
| Socioeconomic status index | ||||
| Q1 (worst) | 210 (26.5) | 228 682 (22.4) | .006 | 0.1 |
| Q2 | 185 (21.3) | 208 373 (19.0) | .09 | 0.1 |
| Q3 | 190 (21.9) | 248 831 (22.7) | .54 | 0.0 |
| Q4 (best) | 208 (23.9) | 335 771 (30.7) | < .01 | 0.2 |
| Missing or unknown | 76 (8.7) | 72 348 (6.6) | .011 | 0.1 |
| No. of hospitals with obstetric units in county | 4.3 ±3.1 | 6.6 ±7.8 | < .01 | 0.2 |
| Comorbid conditionsd | ||||
| Obesity | 178 (20.5) | 180 039 (16.5) | < .01 | 0.1 |
| Gestational hypertension | 125 (14.4) | 104 126 (9.5) | < .01 | 0.2 |
| Substance use disorder | 44 (5.1) | 64 926 (5.9) | .28 | 0.0 |
| Pregnancy anemia | 59 (6.8) | 40 470 (3.7) | < .01 | 0.1 |
| Gestational diabetes | 55 (6.3) | 84 024 (7.7) | .14 | 0.1 |
| Any pregnancy complicatione | 303 (34.9) | 290 404 (26.5) | < .01 | 0.2 |
a We used the χ2 test and 2-sample t test to calculate P values.
b Standardized bias for continuous variables (where treatment = doula and control = non-doula):
Standardized bias for categorical variables (where treatment = doula and control = non-doula):
c Some states could not be named because of health plan compliance rules that prohibit us from identifying specific states without their approval. All states are listed as state number to prevent identification of any particular state.
d Conditions are defined using International Classification of Diseases, Ninth Revision, and International Classification of Diseases, Tenth Revision, codes, which are included in Appendix Table A (available as a supplement to the online version of this article at https://ajph.org).
e Any pregnancy complication is defined as the presence of at least 1 of the following conditions: gestational diabetes; gestational hypertension, preeclampsia, or eclampsia; anemia; placental abruption; thrombocytopenia; placenta previa accreta spectrum disorder; short cervix; or infection.
The doula group differed from the nondoula group on several characteristics. The mean age was higher in the doula group (28.0 vs 27.2); a larger proportion of members reported their race as Black (43.7% vs 22.4%); a larger proportion lived in urban settings (66.2% vs 52.5%); and individuals with doulas delivered in counties with relatively fewer available hospitals with obstetric units (mean number 4.3 vs 6.6). All differences were statistically significant at P < .01 and standardized mean differences greater than 0.1. Some of these differences were expected. Females who were at greater risk for adverse maternal health outcomes were more likely to be contacted about doula services through care management because of their potential for benefiting from doula services (Table 1).
A total of 722 females who received doula care were matched to females who did not receive doula care. Each matched pair included unique individuals whose most recent pregnancy was included in analysis. Therefore, we excluded 147 pregnancies from the total sample that received doula care because they represented multiple pregnancies associated with the same woman. Upon matching by propensity score for baseline sociodemographic characteristics, results indicated balance between the doula and nondoula females, with standardized mean differences less than 0.1 (Table 1).
The clinical characteristics of females with doulas compared very similarly to females with no doulas in the propensity score‒matched sample. The only significant differences were that females with doulas had a lower prevalence of gestational diabetes (6.4% vs 10.3%; P < .01) and a lower prevalence of substance use disorder during pregnancy (5.1% vs 10.3%; P < .01). The doula and nondoula groups were not matched on these individual conditions and instead were matched on the presence of any pregnancy complication, for which the groups showed strong balance (Table 2).
TABLE 2— Baseline Characteristics of Women Enrolled in Medicaid-Affiliated Plans in 9 States Across the United States, Matched Sample: 2014–2023
| Matched Sample | ||||
| Doula (n = 722), Mean ±SD or No. (%) | No Doula (n = 722), Mean ±SD or No. (%) | Pa | Standardized Biasb | |
| Age, y | 28.4 ±5.4 | 28.4 ±5.5 | .97 | 0.0 |
| Statec | ||||
| 1 | 62 (8.7) | 62 (8.6) | .99 | 0.0 |
| 2 | 7 (1.0) | 6 (1.0) | .78 | 0.0 |
| 3 | 113 (16.4) | 118 (15.7) | .7 | 0.0 |
| 4 | 18 (2.5) | 18 (2.5) | .99 | 0.0 |
| 5 | 346 (47.8) | 344 (47.8) | .96 | 0.0 |
| 6 | 49 (6.8) | 49 (6.8) | .99 | 0.0 |
| 7 | 27 (3.8) | 26 (3.6) | .88 | 0.0 |
| 8 | 96 (13.1) | 94 (13.3) | .89 | 0.0 |
| 9 | 2 (0.3) | 3 (0.4) | .66 | 0.0 |
| Urbanicity | ||||
| Urban | 463 (64.2) | 465 (64.4) | .93 | 0.0 |
| Suburban | 202 (27.9) | 199 (27.6) | .88 | 0.0 |
| Rural | 6 (0.8) | 7 (1.0) | .78 | 0.0 |
| Unspecified | 51 (14.2) | 51 (14.2) | .99 | 0.0 |
| Patient’s race/ethnicity | ||||
| Asian/Pacific Islander | 28 (3.9) | 28 (3.9) | .99 | 0.0 |
| Black | 294 (40.8) | 295 (40.9) | .99 | 0.0 |
| Hispanic/Latino | 126 (17.5) | 126 (17.5) | .98 | 0.0 |
| White | 233 (32.3) | 232 (32.3) | .98 | 0.0 |
| Other | 41 (5.7) | 41 (5.7) | .99 | 0.0 |
| Socioeconomic status index | ||||
| Q1 (worst) | 171 (23.6) | 170 (40·2) | .99 | 0.0 |
| Q2 | 155 (21.5) | 157 (21.8) | .92 | 0.0 |
| Q3 | 160 (22.2) | 162 (22.4) | .92 | 0.0 |
| Q4 (best) | 172 (23.9) | 171 (23.7) | .94 | 0.0 |
| Missing or unknown | 64 (8.9) | 62 (8.6) | .67 | 0.0 |
| No. of hospitals with obstetric units in county | 4.2 ±2.9 | 4.8 ±3.8 | < .01 | 0.1 |
| Comorbid conditionsd | ||||
| Obesity | 150 (20.8) | 124 (17.2) | .08 | 0.1 |
| Gestational hypertension | 104 (14.4) | 86 (11.9) | .15 | 0.1 |
| Substance use disorder | 37 (5.1) | 74 (10.3) | < .001 | 0.2 |
| Pregnancy anemia | 52 (7.2) | 46 (6.4) | .52 | 0.0 |
| Gestational diabetes | 46 (6.3) | 74 (10.3) | .008 | 0.1 |
| Any pregnancy complicatione | 255 (35.4) | 255 (35.4) | .99 | 0.0 |
a We used the χ2 test and 2-sample t test to calculate P values.
b Standardized bias for continuous variables (where treatment = doula and control = non-doula):
Standardized bias for categorical variables (where treatment = doula and control = non-doula):
c Some states could not be named because of health plan compliance rules that prohibit us from identifying specific states without their approval. All states are listed as state number to prevent identification of any particular state.
d Conditions are defined using International Classification of Diseases, Ninth Revision, and International Classification of Diseases, Tenth Revision, codes, which are included in Appendix Table A (available as a supplement to the online version of this article at https://ajph.org).
e Any pregnancy complication is defined as the presence of at least 1 of the following conditions: gestational diabetes; gestational hypertension, preeclampsia, or eclampsia; anemia; placental abruption; thrombocytopenia; placenta previa accreta spectrum disorder; short cervix; or infection.
In adjusted analyses, doula care was associated with a 47% lower risk of cesarean delivery (relative risk [RR] = 0.53; 95% CI = 0.43, 0.66) and 29% lower risk of preterm birth (RR = 0.71; 95% CI = 0.51, 0.98). Doula recipients also were 116% more likely to have a vaginal birth after cesarean delivery (RR = 2.16; 95% CI = 1.10, 4.24) and 46% more likely to have timely attendance at a postpartum visit (RR = 1.46; 95% CI = 1.31, 1.61). We observed no significant differences between doula care and ED visits, inpatient admissions, prevalence of postpartum depression and anxiety, or prevalence of severe maternal morbidity within 30 days of delivery (Table 3).
TABLE 3— Relative Risk of Health Outcomes by Receipt of Doula Care Among Propensity Score‒Matched (PSM) Women Enrolled in Medicaid-Affiliated Plans: United States, 2014‒2023
| Doula (n = 722) No. (%) | No Doula (n = 722) No. (%) | PSM RR (95% CI) | |||||
| All Women With Doula | Black Women With Doula | White Women With Doula | Women With Doula in Counties With Low IMR | Women With Doula in Counties With High IMR | |||
| Cesarean delivery | 172 (23.9) | 316 (43.9) | 0.53 (0.43, 0.66) | 0.74 (0.51, 1.1) | 0.79 (0.48, 1.30) | 0.56 (0.27, 1.19) | 0.43 (0.23, 0.81) |
| Vaginal birth after cesarean section | 27 (3.8) | 12 (1.4) | 2.48 (1.29, 4.76) | …a | …a | …a | …a |
| Preterm birth | 61 (8.5) | 83 (11.5) | 0.71 (0.51, 0.98) | 0.70 (0.43, 1.14) | 0.70 (0.28, 1.77) | 0.67 (0.19, 2.36) | 0.71 (0.34, 1.48) |
| ED visit within 30 d postpartum | 42 (5.8) | 42 (5.8) | 0.85 (0.55, 1.31) | …a | …a | …a | …a |
| Inpatient admission within 30 d postpartum | 38 (5.3) | 35 (4.9) | 1.04 (0.65, 1.66) | …a | …a | …a | …a |
| Severe maternal morbidity or mortality | 24 (3.3) | 20 (2.8) | 1.25 (0.70, 2.24) | …a | …a | …a | …a |
| Postpartum anxiety or postpartum depression | 82 (11.4) | 74 (10.3) | 1.12 (0.82, 1.53) | …a | …a | …a | …a |
| Postpartum visit within 7–84 d of birth | 465 (64.6) | 322 (44.6) | 1.46 (1.31, 1.61) | 1.33 (1.12, 1.58) | 1.54 (1.23, 1.92) | 1.54 (0.99, 2.38) | 1.71 (0.95, 1.44) |
Note. CI = confidence interval; ED = emergency department; IMR = infant mortality rate; RR = relative risk. PSM models were matched on maternal age, race/ethnicity, state of residence, socioeconomic status quartile, and urban‒rural classification. PSM models were additionally adjusted for year of pregnancy outcome and the number of hospitals in the county providing obstetric care. Separate PSM models were run on health outcomes with significant results and sufficient sample size for stratified categories (i.e., Black vs White, low IMR vs high IMR). The PSM models stratified by race/ethnicity were not matched on race/ethnicity. Conditions are defined using International Classification of Diseases, Ninth Revision; International Classification of Diseases, Tenth Revision; and Healthcare Common Procedure Coding System codes, which are included in Appendix Table A (available as a supplement to the online version of this article at https://ajph.org).
a Outcomes that were not statistically significant in the overall (nonstratified) PSM model and were not evaluated in the stratified models.
Results indicate both White and Black recipients of doula care were more likely to have timely attendance at a postpartum visit, although the RR was greater for White females (RR = 1.54; 95% CI = 1.23, 1.92) than Black females (RR = 1.33; 95% CI = 1.12, 1.58). No other differences observed between Black and White females were statistically significant. In counties with high IMR, doula care was associated with a 57% lower risk of cesarean delivery (RR = 0.43; 95% CI = 0.23, 0.81) but was not statistically significant among counties with low IMR (Table 3).
Results from this analysis of propensity score‒matched pairs of females enrolled in Medicaid show doula care was associated with a 47% lower risk of cesarean delivery and 29% lower risk for preterm birth. These findings are consistent in effect size with other studies showing similar associations between doulas and decreased risk of cesarean delivery (∼41%) and preterm birth (∼22%).10,13,28
This study contributes unique evidence on the relationship between doula care and attendance at a postpartum visit, demonstrating that females supported by doulas were 46% more likely to have timely attendance at a postpartum visit. Other novel findings from this study include how the effects of doula care on RR of cesarean delivery varies by county-level IMR. Cesarean deliveries, while sometimes necessary, are associated with increased risk of maternal morbidity.29 This study’s finding that doulas were linked with a 57% reduction in cesarean delivery risk among counties with high IMR suggest that utilization of doulas could be an effective strategy for mitigating disparities (and their sequelae of maternal morbidity risk).
Other findings from this study’s subgroup analyses that may be considered for addressing maternal health disparities include the lack of significant differences in cesarean deliveries and preterm births between Black and White doula recipients. Previous studies have reported worse outcomes for Black women—even after adjusting for maternal characteristics such as age, education, insurance status, and clinical conditions.30,31 In addition, the finding that doula care was positively correlated with postpartum visits for Black females is important because most incidences of maternal morbidity and mortality occur in the weeks following birth. Attendance at postpartum visits is critical, therefore, for identifying potential health problems before their onset or increase in severity occurs.32
Results, of course, are contingent on who receives doula care. Individuals who received doula care in this study, for example, were at higher risk for adverse maternal health outcomes based on sociodemographic and clinical characteristics. As doula utilization in the Medicaid population grows, further research may consider exploring the relationship between doula care and maternal health outcomes among populations that are at relatively lower risk for adverse outcomes.
Results may also depend upon when doula care is initiated and utilized. Previous research, for example, indicates lower odds of a postpartum depression or anxiety diagnosis when doula care is utilized during labor and delivery, specifically.8 As doula programs mature in currently covered states and legislation enabling Medicaid coverage of doulas expands to new states, a more nuanced analysis of the timing of doula utilization and the relationship with maternal outcomes using medical claims data may be more feasible.
The following limitations should be considered when interpreting this study’s findings. First, while this study offers one of the most geographically diverse study populations on doula care in the United States, doula-assisted pregnancies represent a very small (< 1%) proportion of the total number of births covered by Medicaid. We have used a robust methodology to reduce the effects of potential selection bias, but unmeasured confounders could remain. The primary study findings, however, are consistent with results previously reported.
Second, because we do not have the complete medical histories of any member in the study, we do not know which pregnancy (e.g., first, second, third) was assessed. The direction of effect of pregnancy number on health outcomes is unclear, however, because risk could increase or decrease depending on the outcome. The way in which pregnancy number affects the relationship between doula utilization and maternal health outcomes should be explored in future research.
Lastly, results from this study generalize only to the Medicaid-insured population. The relationship between doula care and health outcomes may differ in commercially insured populations, which account for more than half of all births in the United States.33
As part of the blueprint for addressing the maternal health crisis in the United States, the federal government is actively making efforts to grow and diversify the doula workforce.34 Results from this study provide evidence to support this endeavor, by showing the positive impact doulas can make toward decreasing maternal health risks. This study demonstrates how partnership with doulas was associated with reductions in cesarean deliveries and preterm births, and improved adherence to postpartum care recommendations. This study’s stratified analyses further provide support for how doula care may reduce maternal health disparities between Black and White females and may be an especially effective strategy for reducing cesarean deliveries in communities at greatest risk for poor maternal and infant health outcomes.
Because doula services are an optional benefit through Medicaid, however, it is at the discretion of the state to reimburse doulas for providing services to Medicaid enrollees. Access to doula care, therefore, varies substantially among Medicaid programs, with some states operating a fully funded Medicaid doula benefit, some states offering small doula programs not widely available to all Medicaid beneficiaries, and some states where the only option for individuals is to self-pay for doula services. As more states begin to operationalize a Medicaid doula benefit and uptake of doula services increases, future research will be able to evaluate how the differences in these state policies affect the relationship between doula care and maternal health outcomes. Causal mechanisms will be able to be explored through analysis of, for example, timing and frequency of doula utilization and how it relates to maternal health. In addition, there is growing momentum for doula coverage mandates through private and commercial insurance coverage, and more employer groups are adding doula coverage to their health plans. This growth in doula coverage will allow for a better understanding of the relationship between doula care and maternal health across lines of business and a broader, more diverse population.
See also Horan, p.
ACKNOWLEDGMENTS
We are grateful to the individuals we corresponded with from the various state health plans about the doula programs in their states. These individuals helped support us with data and provided us with a rich context about the people, challenges, and successes associated with these programs.
CONFLICTS OF INTEREST
All authors were employed by Elevance Health or its subsidiary, Carelon Research, at the time of study and writing of the article.
HUMAN PARTICIPANT PROTECTION
This observational study, conducted under the research exception provisions of Privacy Rule 45 CFR 164.514(e), was exempt from institutional board review because researchers accessed a limited data set for analysis that was devoid of individual patient identifiers and complied with all relevant provisions of the Health Insurance Portability and Accountability Act. Institutional review board exemption was not necessary because the study was an analysis of the managed care organization’s membership data for the purposes of health plan treatment, planning, and operations. All data were anonymized before being used.
