← Archive

The Referral Nobody Owns: How 26 States Built a Doula Benefit That Barely Reaches Any Births

The map is not the revenue. The map is not even the access. The investable asset is the billing-and-referral infrastructure that converts a state benefit into a matched appointment before a patient goes into labor.

Pediatric Health Dispatch | Deep Dispatch | September 11, 2026

The Number and the Gap

PHD published a piece in May arguing that the 26-state Medicaid doula coverage milestone, while real, was not a revenue thesis by itself. Reimbursement rates for labor and delivery support range from $459 to $1,500 across those states. Several enacted programs were still waiting on rate-setting or federal approval. The piece called this out: the map is not the revenue.

The Virginia data, now peer-reviewed in Health Affairs Scholar and published in January 2026, says the situation is worse than that piece implied. This is the correction.

The study, led by researcher Desirae Leaphart Mensah and drawing on Virginia Medicaid claims from January 2022 through December 2024, found that fewer than 1% of pregnant women with a live birth and comprehensive Medicaid coverage received doula services during the benefit's first two years. The denominator is not a technical edge case. It is every eligible Medicaid patient in the state with a live birth during a period when the benefit was funded, reimbursed, and available. The numerator is approximately 600 patients. Virginia had 130 state-certified doulas enrolled in Medicaid at the time; fewer than half ever submitted a claim. Of those who did bill, postpartum visits showed up more often than labor support, which is the service with the strongest outcomes evidence.

One state is one state, and that caveat belongs near the top of any honest reading. Virginia's design is genuinely restrictive. The program requires a physician referral. It requires the patient to use a state-approved doula rather than a community doula the patient found on her own. Those design choices likely suppressed utilization relative to a less-restricted program. What the comparison-state evidence shows is that the restricted-design explanation, while real, does not account for all of what is happening.

Minnesota has covered doula services since 2013, longer than almost any other state in the country and under a substantially less restrictive design. Its 2025 legislatively mandated report on doula services, drawing on Medical Assistance enrollment and claims data, found that doula-service receipt among enrollees rose from 0.31% in 2017 to 2.79% in 2022. That is better than Virginia. It is not a success story. The report also notes that the 2.79% figure likely undercounts actual use, because many doulas in Minnesota do not bill managed-care organizations directly. Instead, they bill through clinics or community organizations, and those encounters are only partially captured in Medicaid claims. The real number may be somewhat higher. It is probably not meaningfully higher. The structural pattern is the same: even in the best-documented, longest-running state Medicaid doula program in the country, fewer than three out of 100 eligible births involved a doula.

Oregon, New Jersey, and Rhode Island did not surface clean public birth-share utilization figures in this research pass, and that absence is itself a finding. Oregon's 2025 workforce analysis documents a doula registry that doubled from 86 to 183 certified providers between 2022 and 2025, but explicitly says more data are needed to know whether that workforce growth translated into more Medicaid-supported births. New Jersey's response to the utilization problem was to build a state-level Doula Assistant tool in 2026 to help doulas and families navigate enrollment, eligibility, and reimbursement paperwork. That a state with an active doula benefit needed to construct a navigator for the navigator is a finding that requires no interpretation.

Arkansas is different in kind from these retrospective comparisons, and it went live eight days ago. Arkansas Medicaid began reimbursing doula and lactation care on September 1, 2026, under a state plan amendment approved in July. Roughly 60% of Arkansas births are Medicaid-funded. Before the payment switch, UAMS reported a statewide doula workforce expansion collaboration, meaning supply-side capacity was built first rather than after. That inverts the usual sequence, where coverage arrives and workforce development scrambles to catch up. Arkansas also runs a two-gate enrollment process: certification through the Department of Health, then a separate Medicaid provider application to DHS. That administrative gap between certified and enrolled providers is exactly where utilization goes to die in other states. Arkansas is, starting now, a prospective natural experiment: a state that did the supply-side work in advance, with a known start date and a Medicaid birth share large enough to produce real numbers. If Arkansas underperforms despite that preparation, the thesis advances from "restrictive designs fail" to "the administrative layer is the binding constraint even when a state does the workforce work first." The first checkpoint is around October 1.

The clinical evidence on doula care is not in dispute. ACOG said so formally in July, in a committee statement co-authored by ob-gyns and practicing doulas, recommending that clinicians build doulas into care teams, citing lower cesarean rates, shorter labor, better breastfeeding initiation, and documented benefits for Black and Indigenous women who bear the largest burden of maternal mortality in this country. The outcomes evidence is settled. It is not the interesting part of this story.

The interesting part is that a benefit with strong clinical evidence, active state investment, established reimbursement rates, and a committed workforce is reaching fewer than 3% of the births it is supposed to serve. That is not a clinical failure. It is a systems failure, and it has a specific location.

Where the Gap Lives

The administrative layer between a state benefit and a birth is the constraint. Not the coverage law. Not the clinical evidence. Not the rates, although rates matter at the margin for individual doulas deciding whether to participate. The constraint is the credentialing process a doula must complete to receive Medicaid payments; the prior-authorization requirements some states attach to individual encounters; the billing infrastructure required to submit claims to a managed-care organization that may accept a different format than the state fee-for-service system; and the referral workflow that is supposed to route a patient from an OB, midwife, or FQHC to a matched doula before the patient goes into labor.

None of those steps is insurmountable on its own. Together they represent a sustained administrative burden that falls on the party least equipped to carry it: an individual community-based doula who may be billing Medicaid for the first time, managing her own certification renewal, operating as an independent contractor without billing staff, and also attending births.

The friction inventory is visible in public program documents, not just in founder complaints. Virginia requires a physician referral and state approval of the individual doula. The published study identified certification attrition and geographic maldistribution as named access barriers, meaning doulas drop out of the Medicaid credentialing process before completing it, and the ones who do complete it are not distributed where the patients are. Minnesota's state report finds that billing through managed-care organizations is technically complex enough that many doulas rely on host organizations as billing intermediaries, an arrangement that suppresses the visibility of utilization in claims data while still requiring a coordination layer that someone has to manage.

New Jersey built a state tool dedicated to navigating the paperwork. Maryland, as of July 2026, is requiring all Medicaid MCOs to standardize their doula-service contracting in value-based arrangements, specifying minimum service components and documenting reimbursement expectations. The fact that Maryland had to mandate the contract structure, not just the coverage, is the clearest single piece of evidence that managed-care plans left to their own devices were not implementing the benefit consistently.

The mechanism is what matters. Plans are not failing to build doula referral pipelines because they made a considered risk judgment that the investment does not pencil out. They have not built a savings case at all. The Commonwealth Fund's 2025 review found 21 states now require doula services in Medicaid managed-care contracts, up from 8 in 2021 and 2022. States are mandating the benefit into managed care. Plans are complying with the mandate. No one inside the plan is measuring doula utilization rates or is financially accountable for whether a covered patient receives a service the plan is required to cover.

The strongest published program-economics figure surfaces from a hospital running a program, not a plan underwriting a policy. Boston Medical Center's Best Beginnings for Babies estimated average savings of $1,152 per patient and 18% return on investment under proposed MassHealth reimbursement rates, driven mostly by lower preterm birth and cesarean rates. That analysis reflects what the hospital learned from running the program. No Medicaid MCO has published a comparable actuarial analysis as a business justification for investing in doula referral infrastructure. That absence is the mechanism. When a plan's actuaries have not built a savings case, the operational work of connecting a patient to a doula before labor is nobody's job.

The binding constraint is not the benefit. It is the infrastructure between the benefit and the birth, and the current payment structure does not compensate anyone for building it.

What the Fix Looks Like

Flourish Care's partnership with NeighborHealth, announced August 24, is the clearest current example of a company taking on exactly that unbundled administrative work as a product.

The operating structure is worth understanding precisely. Flourish handles referral intake, doula matching, care coordination, charting, and insurance billing. NeighborHealth's clinicians refer and do nothing else. NeighborHealth is one of the largest federally qualified health centers in the country, serving more than 120,000 people in Greater Boston; its patient population carries meaningful MassHealth maternity volume. MassHealth has covered doula services since late 2023. The coverage existed before Flourish arrived. What Flourish is selling is the operational layer that converts coverage into an appointment.

The press release frames this as "a roadmap for FQHCs and hospital systems looking to make that coverage real." That is a category claim and it is testable.

The FQHC channel makes strategic sense for this product. HRSA reported 914 health centers providing on-site prenatal care in 2024, serving more than 611,000 prenatal-care patients. A peer-reviewed analysis of health-center delivery data finds that FQHC maternity capacity is concentrated in urban and high-poverty counties: precisely the populations where state doula benefits were designed to close maternal-outcome disparities, and precisely the institutions with the least administrative slack to build doula credentialing and billing functions internally. A doula infrastructure company that can credibly embed into FQHC referral workflows does not need to persuade every OB practice in the country. It needs to persuade the institutions that already control the Medicaid maternity volume.

This changes what the right diligence question is. PHD has tracked doula companies by network scale: states covered, doulas enrolled, families reached. Those metrics describe supply. The constraint documented above is operational and administrative, not supply. The metrics that matter for the infrastructure thesis are different: claims submitted per credentialed doula (a measure of whether enrolled doulas are actually billing), referral-to-first-visit conversion rates (a measure of whether the matching workflow produces appointments), and the number of Medicaid MCOs in which the company holds an active billing relationship rather than routing through a host organization.

Flourish Care is the clearest current example of a company explicitly building for that product category. Malama Health is building toward a payer negotiation through NIH-backed evidence development and remote-monitoring infrastructure, which is a longer path to the same underlying question of whether any MCO will ever fund the fix. The May piece was directionally right: the map is not the revenue. The Virginia study, and the Minnesota comparison, and the Oregon workforce data with no utilization numbers attached, and the New Jersey Doula Assistant tool, and the Maryland transmittal mandating MCO contract structure say the reframe needs to go further. The map is not the access. A state decision to reimburse doula services is a necessary condition for doula care to reach Medicaid births. It is nowhere close to sufficient.

What We're Watching

• Whether Maryland's July 2026 MCO doula-contracting requirements produce measurably better utilization than states relying on voluntary managed-care implementation. Maryland is the clearest natural experiment: every MCO must now cover community-based doula services in value-based arrangements and document a defined minimum service scope. If Medicaid claims in Maryland show meaningfully higher birth-share utilization than peer states without equivalent contract mandates by mid-2027, it will be the best available evidence that contracting specificity is the policy lever that coverage law alone cannot be.
• Whether any Medicaid managed-care organization publishes a doula program cost-benefit analysis based on plan-level actuarial data. The Boston Medical Center Best Beginnings estimate of $1,152 average savings per patient is the strongest figure in the public literature, and it comes from a hospital, not a payer. Every health care category that has seen a plan-authored savings case has subsequently seen utilization rise. The doula category has not seen that analysis from the plan side yet, which explains exactly why the referral infrastructure problem remains nobody's job to solve.
• Whether Flourish Care's FQHC model in Massachusetts produces utilization data that can be compared to statewide MassHealth doula claims. If the embedded administrative-layer approach produces measurably better referral conversion than the baseline, the 914 FQHCs providing prenatal care nationally, concentrated in the counties with the highest Medicaid maternity volume, become a rollup thesis. The channel is large enough. The question is whether the model scales.
• Arkansas provider enrollment counts at the 30, 60, and 90-day marks. ADH certification and DHS Medicaid enrollment are two separate gates, and the gap between certified doulas and enrolled Medicaid providers on go-live day is the leading signal that the administrative-layer pattern is reproducing in a state that did the supply-side work first. First checkpoint around October 1.