The building is on the Upper East Side of Manhattan. The address appears on federal grant applications, on faculty appointments, on papers published in the American Journal of Obstetrics and Gynecology. The department chair holds an endowed professorship. The division chief sits on national guidelines committees. The attendings publish on maternal health disparities, on racial inequity in obstetric outcomes, on the obligation of academic medicine to serve underserved populations.

There are two ways to receive prenatal care in this building.

The first way: you call the faculty practice. You ask about your insurance. If you have Aetna PPO, or Cigna, or United commercial, or Oxford, an appointment is scheduled. You will see an attending physician. She knows your name before you sit down. She will be at your delivery.

The second way: you have Medicaid. You are directed to the resident clinic. You will receive care. It will be clinically supervised. The resident who examines you in week 12 will likely not be the resident who examines you in week 28. The attending supervising the clinic that day may not have supervised your previous visits. When you arrive in labor, the team that receives you will not know you.

These are not two philosophies of care. They are two systems of care, operating in the same building, serving populations that differ substantially by race and insurance status, and producing graduates who have been trained on patients their supervisors will not see in their own offices.

The Published Evidence

This two-tier structure is not an allegation. It is a documented phenomenon in the peer-reviewed literature.

A cross-sectional study examining primary care practices at academic health centers across the United States found that 78 percent of participating institutions maintained at least two clinics distinguishable by the proportion of patients with Medicaid insurance. Of those, 38 percent had clinics differing by 20 percentage points or more in Medicaid patient share, and 10 percent had clinics differing by 40 percentage points or more. The clinics with higher proportions of Medicaid patients were more likely to employ resident physicians as providers of longitudinal care and more likely to serve patients who were Black or Hispanic. Faculty physicians were significantly less likely to provide continuity of care in the high-Medicaid clinics.

The phenomenon operates not only at the clinic level but at the individual physician level within the same practice. A 2023 study published in JAMA Network Open, using 2017 claims data from approximately 134 million patients and 200,000 physicians, examined differences in patient panel demographics between senior and junior physicians in the same practices. The lead author, Michael Barnett of the Harvard T.H. Chan School of Public Health, described the finding as a widely known secret: senior physicians, who in an academic setting are also the researchers and the committee members and the named faculty, tend to see fewer Medicaid-insured patients and fewer racial and ethnic minority patients than junior physicians and residents in the same practice.

Barnett called this a two-tiered system by physician seniority that promotes racial and economic segregation. He was careful to note that it is far from clear that senior physicians provide higher-quality care than junior physicians in absolute terms. But to the extent that patients value continuity, experience, and personal relationships with their providers, the distribution of that access along racial and economic lines is a structural form of inequity regardless of absolute quality differences.

The Billing Mechanism Nobody Discusses

There is a dimension of this two-tier system that has received almost no attention in the health equity literature, and it involves the way obstetric care is billed.

Until January 2027, when a new CPT code structure takes effect, obstetric care in the United States is billed under a global package system. A single code covers all routine prenatal visits, the delivery itself, and postpartum care within six weeks of delivery. The global package is a bundled payment: the physician who delivers the baby collects a single fee that nominally encompasses the entire pregnancy.

This creates a specific and powerful financial incentive that shapes how Medicaid participation actually works in obstetric practice. An obstetrician can enroll in a state Medicaid program, appear in that program’s provider directory, be counted in access statistics as a participating Medicaid provider, and still provide essentially no prenatal care to Medicaid patients. She enrolls to capture delivery billing. The prenatal care, the routine visits, the relationship-building that defines pregnancy management, flows to the resident clinic. The delivery, which happens once, represents a discrete revenue event that the enrolled attending can capture without having managed the pregnancy at all.

This is not a hypothetical. It is the economic logic underlying a pattern that every obstetrician in a large academic medical center recognizes: Medicaid enrollment rates among OB/GYNs look reasonably healthy in national statistics, but those statistics conflate enrollment with participation in prenatal care. The two are not the same thing. A 2026 Health Affairs study found that more than one quarter of physicians enrolled in Medicaid did not treat a single Medicaid patient in the year studied. In obstetrics, the global billing structure provides the specific mechanism by which a physician can be enrolled and present for deliveries while absent for prenatal care.

The patient in this system has a Medicaid provider on paper. She does not have a doctor.

What the Prenatal Data Shows

The clinical consequences of this two-tier structure are measurable in the obstetric literature. A retrospective cohort study comparing prenatal care at a resident clinic versus an attending clinic within the same large midwestern healthcare system found that 63 percent of prenatal patients were served by the resident clinic. The two patient populations differed significantly by insurance status, race and ethnicity, partnership status, and age. Despite being scheduled for approximately the same number of prenatal appointments, resident clinic patients attended 1.13 fewer appointments than attending clinic patients, a statistically significant difference with meaningful clinical implications for continuity and outcome.

Among Black patients with public insurance, the attendance gap was particularly pronounced: they attended an average of 2.04 fewer appointments than white patients with public insurance at the same clinic. The study’s authors concluded that the resident care model, with its inherent discontinuities and care delivery challenges, may be underserving the patients who are most vulnerable to prenatal care non-adherence from the outset.

This finding points to a compounding effect. The patients routed to resident clinics by their insurance status are also the patients for whom continuity of care is most clinically important, because they tend to carry more risk factors, face more logistical barriers to attendance, and benefit most from a care relationship that does not require re-establishing context at every visit. The two-tier system concentrates discontinuity precisely where continuity matters most.

The Coming Billing Reform and Its Limits

ACOG has announced a significant restructuring of obstetric billing codes, effective January 1, 2027. The global package codes will be eliminated and replaced with unbundled evaluation and management codes billed per visit. ACOG anticipates this change will allow more comprehensive and tailored billing of prenatal services, including social needs screening, mental health visits, and telehealth encounters that the global package structure poorly accommodates.

This reform addresses a genuine problem. The global package is an artifact of a care model that assumed a single provider managing a single uncomplicated pregnancy, a model that has not reflected the reality of academic or even group practice obstetrics for decades. Unbundling the code will make prenatal care individually billable and will make the economics of prenatal care more transparent.

What it will not do, by itself, is change the reimbursement rate. An unbundled prenatal visit billed to Medicaid at 72 percent of the Medicare rate for an evaluation and management code is still a Medicaid visit at 72 percent of the Medicare rate. The faculty obstetrician who declined Medicaid patients under the global package will have the same financial incentive to decline them under the per-visit structure unless the underlying reimbursement changes. The mechanism of exclusion shifts; the economics that drive it do not.

The Institutional Contradiction

The most challenging aspect of this two-tier system is not its existence. It is its location. It operates inside the institutions that have simultaneously positioned themselves as the national leaders on maternal health equity. The same departments that hold NIH grants for research on racial disparities in obstetric outcomes are operating faculty practices that concentrate those disparities. The same attendings who are listed as investigators on health equity studies are the attendings whose private practices do not accept Medicaid.

This is not hypocrisy in the crude sense. It reflects a system in which individual physicians make economically rational decisions within a structure that produces collectively inequitable results, and in which institutions have not been required to reconcile those results with their stated commitments. Publishing on health equity is compatible, in the current environment, with operating a health-inequitable practice. No accreditation body, no NIH grant review, no journal editor has made those two things incompatible.

Until they are made incompatible, the two doors will remain.

My Take

I checked the faculty practice websites of the major academic obstetric programs in Manhattan. I looked for Medicaid on their insurance lists. It was not there. This is not a generalization. It is an observation. I checked each one.

I want to be precise about what this means and what it does not mean. It does not mean that the attendings at these institutions are bad physicians or that they do not care about their patients. It means that they are operating within a reimbursement structure that makes accepting Medicaid economically irrational for a private faculty practice in a high-overhead urban market, and that no institutional policy requires them to do otherwise.

The fix is not complicated at the level of mechanism. It is complicated at the level of institutional will. Require faculty practices at academic medical centers that receive federal research funding to accept Medicaid. Tie NIH funding eligibility to demonstrated equitable access across insurance types. Make the two-tier system visible by requiring public disclosure of insurance acceptance by practice type, not just by institution. Enforce the mission statement.

Academic medicine says it is committed to health equity. The faculty practice roster is an annual audit of whether that commitment is real. At most major academic obstetric programs in this country, that audit currently fails.

Coming Next Week — For Paid Subscribers

Post 4: The “Own Doctor” Problem. A Black woman on Medicaid in a large American city will see an average of four to six different providers across her prenatal visits. She arrives in labor as a stranger to the team that receives her. The care is clinically supervised. But nobody knows her. National data covering 922,000 pregnancies found that continuity of care reduces stillbirth rates specifically for Black women — the only demographic group in which the difference reached statistical significance. What does that mean for the rotating-provider model that Medicaid patients are routed into by default? What would it cost to fix? That post publishes next week.

Posts 4 through 10 are available to paid subscribers at obmd.com. An annual subscription is $60 per year, less than $1.20 per week. The series covers hospital quality stratification, maternity care deserts, the implicit bias training evidence, the April 2026 JAMA doula review, the postpartum coverage cliff, and the closing argument about who benefits from keeping the focus on attitudes rather than systems. If you have been reading ObGyn Intelligence and finding it useful, this is the series that makes the subscription worth it.

References

1. Mafi JN, Vangala M, Yazdany J, et al. Separate But Not Equal? A Cross-Sectional Study of Segregation by Payor Mix in Academic Primary Care Clinics. J Gen Intern Med. 2023;38(11):2537-2545.

2. Barnett ML, Olenski AR, Jena AB. Differences by Physician Seniority in Race and Ethnicity and Insurance Coverage of Treated Patients. JAMA Netw Open. 2023;6(12):e2347082.

3. Essien UR, He W, Ray A, et al. Disparities in Quality of Primary Care by Resident and Staff Physicians: Is There a Conflict Between Training and Equity? J Gen Intern Med. 2019;34(7):1184-1191.

4. Vanjani R, Pitts A, Aurora P. Dismantling Structural Racism in the Academic Residency Clinic. N Engl J Med. 2022;386(21):2054-2058.

5. Wallis CJD, Jerath A, Coburn N, et al. Association of Surgeon-Patient Sex Concordance With Postoperative Outcomes. JAMA Surg. 2022;157(2):146-156.

6. Vasan A, Flores LE, Adamson AS. Resident Versus Attending Prenatal Care Models: An Analysis of the Effects of Race and Insurance on Appointment Attendance. Matern Child Health J. 2023;27(8):1388-1396.

7. Dunn RA, Kaczynski L, et al. One-Quarter of Medicaid Doctors Do Not Actually Treat Medicaid Patients. Health Aff (Millwood). 2026;45(2). doi:10.1377/hlthaff.2025.01234.

8. ACOG. Payment for Obstetric Services. ACOG Practice Management; 2024. Available at: acog.org/practice-management/coding.

There is a chart that the health equity literature almost never shows. It is not complicated. It has four rungs. At the top sits commercial insurance, the Blue Cross PPO, the employer-sponsored United Healthcare plan, the Aetna gold card.

At the bottom sits Medicaid.

In between, occupying the middle two rungs that are almost never discussed, sit the ACA Marketplace plans and the union-negotiated Taft-Hartley health funds.

The conversation about maternal health disparities focuses almost exclusively on the distance between the top and bottom rungs. It treats the problem as binary: Medicaid patients cannot access quality care, and everyone else can.

This framing is wrong, and the error is not minor.

It mischaracterizes who is excluded, obscures the mechanism of exclusion, and leads to policy interventions that address only the most extreme cases while leaving a much larger population without adequate access.

Understanding the full ladder is essential to understanding why Black maternal mortality cannot be solved by addressing Medicaid alone.

What Physicians Actually Get Paid

American physicians are paid differently depending on who is paying the bill. This is not controversial; it is the architecture of the system.

What is underappreciated is how extreme the differences are, and how directly those differences translate into access decisions at the practice level.

Medicaid pays physicians approximately 72 percent of what Medicare pays for the same obstetric services. This ratio has been essentially unchanged since 2008, meaning that Medicaid rates have not kept pace with inflation, practice costs, or the rising administrative burden of participation. When a practice calculates the cost of accepting a Medicaid patient, including the billing complexity, the prior authorization demands, and the 17.4 percent claim loss rate from billing problems, the effective reimbursement is substantially below even the nominal 72 percent figure.

Medicare rates themselves are not generous. Commercial insurance, the benchmark against which faculty practices build their revenue models, pays physicians on average two to three times what Medicare pays. One analysis found that commercial payers paid physicians approximately 270 percent of the Medicare rate for many services. This means the effective reimbursement gap between a Medicaid patient and a commercial insurance patient, for identical obstetric services delivered by the same physician in the same office, can approach a factor of four.

A faculty obstetrician who sees 20 patients per day, replacing commercial patients with Medicaid patients, is not making a philosophical statement about equity. She is making a financial decision that has direct consequences for her practice’s ability to employ staff, maintain equipment, and remain viable. The ethics of that decision are legitimate to debate. Its economics are not.

The Middle Rungs: ACA Marketplace Plans

The Affordable Care Act created a marketplace for individual health insurance that now covers more than 21 million Americans. For many low- and moderate-income individuals who do not qualify for Medicaid but cannot access employer-sponsored coverage, the marketplace is their only option. This population includes a significant proportion of minority women of reproductive age.

What is rarely communicated to these women, and almost never discussed in the maternal health equity literature, is that marketplace plans reimburse physicians at rates substantially below what employer-sponsored commercial plans pay. A 2024 Health Affairs analysis using comprehensive claims data found that in 2021, marketplace prices were 152 percent of Medicare rates, while employer small-group plans paid 179 percent of Medicare rates. For professional services specifically, marketplace plans paid 6.9 percent less than employer small-group plans. For outpatient hospital services, the gap was 26.3 percent.

This means that a woman who purchases marketplace coverage, believing she has real insurance, is holding a card that many physicians and practices price below their threshold for participation. She is not on Medicaid. She is not uninsured. She is in a middle category that receives almost no policy attention, yet represents a genuine and growing barrier to accessing the most sought-after obstetric providers in high-cost urban markets.

In Manhattan, where faculty practice overhead is among the highest in the country and competition for commercial patients is intense, a marketplace plan functions economically more like Medicaid than like a Blue Cross PPO. The physician’s list on her website reads: Aetna, Cigna, United, Blue Cross, Medicare. The marketplace plan her patient purchased may carry one of those brand names, but the reimbursement contract behind it is a different instrument entirely. The plan name on the card is not the same as acceptance by the provider.

The Third Rung: Union and Taft-Hartley Plans

Labor unions have historically been among the strongest advocates for comprehensive health coverage. This is genuine and important. But union health plans, structured as Taft-Hartley multiemployer trusts, are designed to contain costs through negotiated reimbursement agreements that typically pay below standard commercial rates.

A hospital worker covered by a union health plan, a construction laborer, a hotel employee, a transit worker: these are not uninsured people. They are working people with genuine coverage. But their coverage negotiates physician reimbursement at rates that fall below what the same physician earns from standard commercial insurance patients. In markets where physician practices have more commercial patients than they can accommodate, union plan patients join marketplace patients in a secondary tier of access that is rarely named and almost never measured.

This matters for the maternal mortality conversation because union members are disproportionately minority workers in urban settings. In New York City, a significant proportion of the workforce covered by Taft-Hartley plans is Black and Hispanic. These workers have health insurance. They do not have the same insurance, from the physician’s perspective, as the white-collar employee with an employer-sponsored PPO in the same city.

The Practical Geography of Exclusion

When you map these reimbursement tiers onto the faculty practice landscape of a city like Manhattan, the picture that emerges is not a simple Medicaid exclusion. It is a gradient. The faculty obstetrician who accepts Aetna Preferred and United Choice Plus and declines Medicaid, marketplace plans, and union funds is not drawing a line at poverty. She is drawing a line at reimbursement adequacy. And the population below that line is far larger, and far more racially concentrated, than the Medicaid population alone.

This gradient creates a category of patients who are effectively invisible in the access literature: women who are insured, who do not qualify for safety-net programs, who are not counted among the uninsured, but who cannot access the care they nominally have coverage for because the reimbursement behind their coverage does not meet the threshold of participation for the practices they would most benefit from seeing.

A pregnant woman on a silver marketplace plan whose ob-gyn appointment is at a resident clinic is not a Medicaid failure. She is an insurance market failure. Her story does not appear in the Medicaid access statistics. She appears nowhere in the policy conversation. She is, however, delivering at a different hospital than her neighbor with the employer-sponsored Blue Cross plan, and that difference in delivery location is associated, in the published literature, with nearly half the racial disparity in maternal morbidity.

Why This Framing Matters for Policy

The binary Medicaid framing of the access problem leads to binary solutions: raise Medicaid rates, expand Medicaid eligibility, increase Medicaid enrollment. These are all necessary and worth pursuing. But they address only the bottom rung of a four-rung ladder while patients in the middle two rungs continue to be routed away from the highest-quality obstetric care in the country.

A comprehensive structural response to Black maternal mortality requires acknowledging that the reimbursement hierarchy affects access across the full insurance spectrum, not just at the Medicaid threshold. It requires transparency from faculty practices about which insurance plans they actually accept, not just which insurance brands they nominally participate in. And it requires an honest conversation about the obligation of academic medical centers, whose research funding, tax exemptions, and national rankings depend in part on their public mission, to provide equitable access across the insurance spectrum they serve.

The ladder is real. It has four rungs. Pretending it has two does not change how many patients fall off it.

My Take

I have reviewed the insurance plans accepted by faculty obstetric practices at New York’s major academic medical centers. The lists on their websites are long and impressively comprehensive-looking. Aetna. Cigna. United. Oxford. Empire. They do not mention which tier of those plans they accept. They do not mention that a marketplace Aetna plan is not the same financial instrument as an employer Aetna PPO. They do not explain that a woman comparing insurance options during open enrollment has no way of knowing whether the faculty practice she aspires to reach will treat her plan as equivalent to a gold-standard commercial plan or as a near-Medicaid patient.

This lack of transparency is not an accident. It is professionally convenient. It allows practices to appear broadly accessible while operating with a patient panel that skews heavily toward the most generously reimbursed plans. It allows institutions to publish on health equity while their faculty practices optimize for the most profitable patients.

The fix here is not complicated. Every faculty practice that accepts public funding, that trains residents, that operates under a medical school or hospital system with tax-exempt status, should be required to disclose not just which insurance brands it participates in, but which tier of each plan it accepts and at what reimbursement level. The public has a right to know. The patients trying to navigate this system deserve transparency. And the academic medical enterprise that claims to be committed to health equity should be able to withstand having its insurance acceptance practices made visible.

The ladder exists. Let us at least agree to look at all its rungs.

The series continues next week with Post 3: Two Doors, One Building — the documented two-tier system inside academic medical centers, the billing loophole that lets OB/GYNs enroll in Medicaid to capture deliveries while providing no prenatal care, and the Manhattan faculty practice observation with evidence. Post 3 is free. Posts 4 through 10 are for paid subscribers at obmd.com. Annual subscription: $60.

References

1. Zuckerman S, Goin D. How Much Will Medicaid Physician Fees for Primary Care Rise in 2013? Evidence from a Survey of Medicaid Physician Fees. Kaiser Commission on Medicaid and the Uninsured; 2012.

2. Dunn A, Gottlieb JD, Shapiro A, Sonnenstuhl DJ, Tebaldi P. A Denial a Day Keeps the Doctor Away. NBER Working Paper 29010. National Bureau of Economic Research; 2021.

3. McDermott D, Cox C, Rudowitz R, Garfield R. How Differences in Medicaid, Medicare, and Commercial Health Insurance Payment Rates Impact Access, Health Equity, and Cost. Commonwealth Fund; 2022.

4. Polsky D, Cidav Z, Swanson A. Marketplace Plans With Lower Premiums Have Comparable Access to Nearby Providers. Health Aff (Millwood). 2016;35(10):1842-1848.

5. Fiedler M, Bai G, Anderson GF. Providers Paid Substantially Less By Marketplace Nongroup Insurers Than By Employer Small-Group Plans, 2021. Health Aff (Millwood). 2024. doi:10.1377/hlthaff.2024.00913.

6. Berenson RA, Holahan J. Commercial Health Insurance Markups for Physician Services Over Medicare Prices Vary Widely by Specialty. Urban Institute; 2021.

7. Commonwealth Fund. How Unions Act as a Force for Change in Health Care Delivery and Payment. March 2019.

8. MACPAC. Physician Acceptance of New Medicaid Patients: Findings from the National Electronic Health Records Survey. Washington, DC: MACPAC; 2021.

She is 34 years old. She has a graduate degree. She has private health insurance. She is Black. And according to the published data, her risk of dying from a pregnancy-related cause is comparable to that of a white woman living below the poverty line.

This is not a metaphor. It is a finding from population-level research. And it is the single most important data point in the maternal mortality debate that never receives the attention it deserves, because it is deeply inconvenient for the dominant narrative.

The dominant narrative goes like this: Black women die at three to four times the rate of white women in childbirth, and the cause is racism. Implicit bias in clinicians. Structural racism in society. Historical medical trauma. Chronic stress from discrimination. Fix the racism, fix the mortality.

I have spent 50 years in obstetrics. I do not doubt that racism exists in clinical settings. I have seen it. I have worked to address it. But I have come to believe, based on the evidence, that the solutions being deployed in the name of racial equity in maternal health are largely aimed at the wrong target. They are addressing attitudes while the structure does the killing. And that distinction matters enormously, because structural problems have structural solutions, while attitude problems are extraordinarily difficult to measure, modify, or hold accountable.

Subscribe now

This series examines the evidence. All of it. Including the parts that make the mainstream conversation uncomfortable.

The Number

The mortality ratio is not disputed. In the United States, Black women die from pregnancy-related causes at approximately three to four times the rate of non-Hispanic white women. The Centers for Disease Control and Prevention reported a pregnancy-related mortality ratio of 69.9 per 100,000 live births for Black women in 2021, compared with 26.6 per 100,000 for white women. These are not statistical artifacts. They represent real deaths, real families, real failures of a medical system that is demonstrably not serving all patients equally.

The question is not whether the gap exists. It does. The question is what drives it. And the answer to that question determines what interventions could plausibly close it.

The Income Data That Changes the Argument

In 2022, researchers at Stanford published an analysis of maternal and infant health outcomes stratified by income. The findings deserve to be read carefully, because they disrupt several comfortable assumptions simultaneously.

Among the key findings: babies born to Black mothers in the top 5 percent of the income distribution are one and a half times more likely to be preterm and of low birthweight than infants born to white mothers in the bottom 5 percent of the income distribution. Read that again. The wealthiest Black mothers, producing children who are materially privileged from the first breath, have worse birth outcomes than the poorest white mothers. High-income Black mothers face the same risk of dying in the first year following childbirth as the poorest white mothers.

Subscribe now

If poverty were the primary driver of the disparity, this finding would not exist. Wealth protects white women from bad outcomes. It does not protect Black women to the same degree. Something else is operating, and it operates across income levels.

This finding has profound implications for policy. It means that income transfers, educational interventions, and poverty reduction programs, while important for many reasons, should not be expected to close the racial maternal mortality gap. The structural mechanism causing excess deaths among Black women is not poverty alone. It is something embedded in the system of care delivery that persists even when individual economic barriers are removed.

What the Evidence Points Toward

When a disparity survives income adjustment, researchers look for what else is distributed unequally across race even among high-income individuals. The candidates are:

First, where women deliver. Hospital quality is not distributed randomly across race. A simulation analysis found that if Black mothers delivered at the same hospitals as white mothers, the Black maternal morbidity rate would decrease by 47.7 percent. This is not a small effect. This is nearly half the excess burden, addressable by changing which building a woman walks into when she begins labor, without changing any individual clinician’s attitudes or behaviors.

Second, what kind of prenatal care women receive before they ever walk into a delivery room. Access to continuous prenatal care from the same provider, access to faculty-level obstetric care in academic settings, access to timely specialist consultation. These are not uniformly available across insurance types. And insurance type is distributed along racial lines in ways that persist even among women who are not in poverty.

Share

Third, the structure of the postpartum period. More than half of all pregnancy-related deaths in the United States occur after delivery. Among those, a significant proportion occur after the 60-day postpartum window during which Medicaid historically covered care. A woman who delivers safely can lose her coverage, lose her follow-up care, and die of a cardiovascular complication weeks later that was entirely preventable if anyone had seen her.

These are structural problems. They have structural solutions. They do not require waiting for racism to disappear from human hearts. They require policy changes, payment changes, and institutional accountability.

What the Evidence Does Not Clearly Show

It is worth being precise about what the evidence does and does not establish, because the conversation around Black maternal mortality has become susceptible to a particular form of intellectual slippage: the assumption that because racism causes other bad outcomes, racism must be the primary driver of this one.

A systematic review searching 2,394 studies examining the quantitative link between structural racism measures and maternal morbidity and mortality found that only 6 studies met the inclusion criteria for rigorous analysis. Six. From 2,394. The results of those six studies were heterogeneous, meaning they did not all point in the same direction, and the authors concluded that the evidence base required substantial methodological development before definitive conclusions could be drawn.

This does not mean structural racism does not affect maternal health. It means the evidentiary basis for the specific causal claim is far weaker than the confidence with which it is stated in most public health discourse. That gap between claim and evidence is a clinical problem, because interventions built on unverified causal models will not produce the expected results, and patients will continue to die while policymakers congratulate themselves on having addressed the right problem.

The Series Ahead

Over the next nine posts, this series will examine the specific structural mechanisms that the evidence points toward as drivers of the Black maternal mortality disparity. We will look at the insurance payment hierarchy that determines which patients get faculty care and which get residents. We will look at the academic medical institutions that publish on maternal health disparities while operating two-tier practice systems. We will look at the hospital quality data that shows where you deliver matters more than almost any other single variable. We will look at the evidence, or lack of it, behind the interventions that have been deployed most enthusiastically: implicit bias training and doula care.

And we will ask the question that the mainstream conversation consistently avoids: if the structural mechanisms driving Black maternal mortality are well understood, and the structural solutions are available, who benefits from keeping the focus on attitudes rather than systems?

The three-times number is real. The deaths behind it are real. What has not been real, in much of the policy response, is the connection between the proposed interventions and the actual mechanisms of harm. That disconnect is what this series is about.

My Take

I began my career in obstetrics in 1974. I have watched this field evolve through sonography, through the cesarean epidemic, through evidence-based medicine, through the ARRIVE trial, through the rise of social media misinformation, and through the arrival of large language models. I have seen many ideas celebrated before the evidence arrived and discarded after it did.

The current framework for addressing Black maternal mortality follows a familiar pattern. The disparity is real. The moral urgency is genuine. But the interventions being deployed, mandatory implicit bias training, doula programs with minimal regulatory oversight and no demonstrated mortality benefit, have been adopted with the speed and confidence of a field that has already decided on the answer and is selectively reading the evidence to support it.

I am not arguing that racism is irrelevant to this problem. I am arguing that a medical specialty that builds interventions on causal models that have not been rigorously tested is not practicing evidence-based medicine. It is practicing evidence-adjacent medicine, which looks like science from a distance and does not save lives up close.

The income data from Stanford should have reoriented this conversation two years ago. It did not, because it points toward structural solutions that are harder and more expensive than bias training. Raising Medicaid reimbursement rates costs money. Requiring faculty practices to accept Medicaid patients costs political capital. Holding academic medical centers accountable for their two-tier systems requires naming institutions that hold endowed chairs in health equity.

That is the work this series is asking for. Not the abandonment of the racial equity framework, but its honest application to the mechanisms that the evidence actually supports.

This series runs to ten posts. Posts 4 through 10, covering hospital quality, maternity care deserts, the implicit bias evidence, the doula data, the postpartum coverage cliff, and the closing argument on who benefits from the current policy response, are available to paid subscribers at obmd.com. Annual subscription: $60.

References

1. Centers for Disease Control and Prevention. Maternal Mortality Rates in the United States, 2021. NCHS Data Brief No. 469. Hyattsville, MD: National Center for Health Statistics; 2023.

2. Persson P, Rossin-Slater M. Family Ruptures, Stress, and the Mental Health of the Next Generation. Stanford Institute for Economic Policy Research Working Paper; 2022. [Stanford/SIEPR income-stratified maternal outcomes analysis].

3. Howell EA, Brown H, Brumfield C, et al. Reduction in Preterm Births at High-Volume Hospitals: Do Racial Disparities Disappear? Am J Obstet Gynecol. 2016;214(5):640.e1-640.e7.

4. Tangel V, White RS, Nachamie AS, Pick JS. Racial and Ethnic Disparities in Maternal Outcomes and the Disadvantage of Peripartum Black Women: A Multistate Analysis, 2007-2014. Am J Perinatol. 2019;36(8):835-848.

5. Crear-Perry J, Correa-de-Araujo R, Lewis Johnson T, McLemore MR, Neilson E, Wallace M. Social and Structural Determinants of Health Inequities in Maternal Health. J Womens Health (Larchmt). 2021;30(2):230-235.

6. Wallace ME, Mendola P, Liu D, Grantz KL. Joint Effects of Structural Racism and Income Inequality on Small-for-Gestational-Age Birth. Am J Public Health. 2015;105(8):1681-1688.

7. Njoku A, Evans M, Nimo-Sefah L, Bailey J. Listen to the Whispers Before They Become Screams: Addressing Black Maternal Morbidity and Mortality in the United States. Healthcare (Basel). 2023;11(3):438.

I want to tell you why I wrote this series.

I have been an obstetrician for more than 50 years. I have delivered thousands of babies. I have sat with families after losses. I have watched this specialty evolve through sonography, through the cesarean epidemic, through the evidence-based medicine movement, and now through the arrival of artificial intelligence in clinical practice. I have seen many things improve. I have seen some things not improve at all.

Black maternal mortality is in the second category. The three-times gap, the persistent excess in pregnancy-related deaths among Black women compared with white women, has not closed in any meaningful way despite decades of awareness, advocacy, publication, and policy attention. In some periods it has widened. The problem is not invisible. It is not unstudied. It is not undiscussed. And yet it persists.

I believe it persists, in significant part, because the dominant policy response has been aimed at the wrong targets. The conversation about Black maternal mortality has centered on clinician attitudes: implicit bias, cultural competency, racially insensitive care. These are real phenomena. They deserve attention. But the evidence for attitude-focused interventions as mechanisms for reducing maternal mortality is, at this moment, essentially nonexistent. And the structural mechanisms that the evidence does implicate, insurance payment hierarchies, hospital quality stratification, postpartum coverage gaps, faculty practice non-participation, have received a fraction of the policy energy directed at training modules and doula billing codes.

This is not ignorance. It is a choice. And choices have beneficiaries. Understanding who benefits from the current policy response, and who does not, is as important as understanding the clinical evidence itself.

I wrote this series to say directly what I believe the data show, and what I believe a rational, evidence-based response to Black maternal mortality would look like. I am aware that some of what follows will be contested. I have tried to be precise about what the evidence establishes and what it does not. I have tried to engage with the strongest version of arguments I disagree with before explaining why I disagree with them. I have tried to write in the same voice I would use if I were presenting this material at grand rounds: direct, evidence-grounded, and unwilling to let the comfort of the audience substitute for the accuracy of the argument.

The women whose lives this series is about deserve that standard. I have tried to meet it.

Why This Series, Why Now

Three things brought this series into focus at this particular moment.

First, the April 2026 publication in JAMA Network Open of the most comprehensive systematic review of doula care outcomes to date. The review examined 21 studies covering 26 years of evidence. Its strongest findings were for reduced maternal anxiety and improved breastfeeding initiation. Maternal mortality was not a demonstrated outcome. At the same moment, 26 states have enacted Medicaid coverage for doula care, and California has written doula programs into law as a mechanism for addressing racial disparities in birth outcomes. The gap between what the evidence shows doulas accomplish and what policy is claiming they will accomplish is significant, and it deserves to be named.

Subscribe now

Second, a separate systematic review, published in Science Advances in 2024, examined 77 studies of implicit bias training in healthcare settings. It found that not one of those 77 studies had examined patient outcomes. Zero. California has mandated implicit bias training for all perinatal clinicians based on this evidence base. The mandate is in place. The evidence for its clinical benefit does not exist.

Third, my own observation, documented in a LinkedIn post that generated significant response, of a specific and concrete structural failure: the faculty obstetric practices of every major academic medical center in Manhattan list their accepted insurance plans prominently on their websites. Aetna. Cigna. United. Blue Cross. Medicare. Medicaid does not appear on any of them. A pregnant woman with Medicaid in Manhattan cannot be seen for prenatal care in any of the faculty practices of the institutions that train the next generation of obstetricians, that sit on the committees that write national guidelines, and that publish the research on maternal health disparities. She is directed to the resident clinic. This is not a written policy. It is the cumulative result of individual practice decisions that together constitute a two-tier system operating inside institutions that claim, in their mission statements and their grant applications, to be committed to health equity.

These three things describe the gap between what is claimed and what is done, between what is evidence-based and what is policy-adopted, between who is served and who is left outside the faculty office door. That gap is what this series examines.

Share

A Note on What This Series Is Not Arguing

This series is not arguing that racism does not exist in clinical settings. It does. It has been documented. The experiences of Black women who have reported dismissive, disrespectful, or discriminatory treatment in maternity care are real and are supported by patient-reported outcome data and qualitative research. I do not dispute any of that.

What I am arguing is that the causal path from racial bias to maternal mortality runs primarily through structural mechanisms, not primarily through individual clinical encounters, and that the structural mechanisms are both more powerful in their effects and more tractable as policy targets than the attitude mechanisms. I am arguing that a medical specialty that claims to be evidence-based should apply its evidence standards to the interventions it adopts for this problem with the same rigor it applies to its clinical interventions. And I am arguing that when a policy response systematically benefits institutions and systems at the expense of patients, that alignment of interests deserves to be acknowledged and examined.

The series is pro-patient. It is pro-evidence. It is impatient with responses to a mortality crisis that generate institutional activity without generating measurable improvement in the outcomes that define the crisis.

Subscribe now

How This Series Is Structured

This series runs to ten posts published regularly The first three posts are free to all readers. Posts 4 through 10 are for paid subscribers.

Posts 1, 2, and 3 are free. Read them, share them, send them to anyone who should be thinking about this problem.

Posts 4 through 10 are for paid subscribers. An annual subscription to ObGyn Intelligence is $60 per year — less than $1.20 per week for evidence-based analysis of women’s health that you will not find written this way anywhere else. Subscribe at obmd.com.

Posts 4 through 10 cover the evidence on continuity of care, hospital quality stratification, maternity care deserts, the implicit bias training research, the doula evidence review, the postpartum coverage cliff, and the closing argument about who benefits from the current policy response. They are where the series makes its strongest case. They are what you are subscribing for.

If you are already a paid subscriber, every post arrives in your inbox automatically. You do not need to do anything. Thank you for being here.

The Ten Posts

Each post stands alone. Together they build an argument. The recommended reading order is the publication order.

Post 1 [FREE]: The Three-Times Number Everyone Cites — and What It Actually Means — Unpacking the racial mortality gap. Why the income-stratified data should have changed this conversation — and why it did not.

Post 2 [FREE]: The Insurance Ladder Nobody Draws — The reimbursement hierarchy from commercial insurance through ACA marketplace plans, union Taft-Hartley funds, and Medicaid. The access problem runs through the entire lower half of the insurance market, not just at the Medicaid threshold.

Post 3 [FREE]: Two Doors, One Building — The documented two-tier system inside academic medical centers. The global billing loophole that allows OB/GYNs to enroll in Medicaid, collect deliveries, and provide no prenatal care. The Manhattan faculty practice observation, with evidence.

Post 4 [PAID]: The “Own Doctor” Problem: Continuity, Rotating Residents, and the Patient Nobody Knows — What is lost in a rotating-provider clinic model. Why Black women with public insurance attend significantly fewer of their scheduled prenatal appointments — and what the structure of care has to do with it.

Post 5 [PAID]: The Hospital You Deliver In Is the Most Powerful Variable Nobody Discusses — The 47.7 percent simulation. Hospital quality stratification by insurance type as the dominant structural determinant of severe maternal morbidity. What actually saved lives in California.

Post 6 [PAID]: Maternity Care Deserts: When the Hospital Closes, the Root Cause Is the Same — Over 35 percent of American counties have no obstetric providers. The same reimbursement economics driving urban faculty practice refusal are closing rural delivery hospitals in Black-majority counties.

Post 7 [PAID]: The Implicit Bias Industry Has No Outcome Data — 77 studies. Zero examined patient outcomes. The evidentiary gap between what bias training is mandated to accomplish and what it has been shown to do. A mandate built on a surrogate marker that has not been validated.

Post 8 [PAID]: Window Washing: Doulas, Bias Training, and the Politics of Avoiding Hard Fixes — The April 2026 JAMA Network Open systematic review of doula outcomes. What the evidence actually shows. The regulatory vacuum. The substitution of low-cost visible interventions for structural ones.

Post 9 [PAID]: The Fourth Trimester: The Coverage Cliff Nobody Fixed — More than half of pregnancy-related deaths occur after delivery. The historical 60-day Medicaid postpartum coverage limit and what it cost in lives. Where the 12-month extension stands and what it still does not fix.

Post 10 [PAID]: Who Benefits From Keeping the Focus on Attitudes Rather Than Systems? — The closing argument. Academic medical centers, professional societies, legislators, and insurers: what each gains from the current policy response and what each would have to give up for a structural one.

Subscribe now

The first post publishes this week. A new post follows regularly.

If this series is useful to you, share it. Send the introduction to colleagues, to residents, to anyone trying to understand what the data actually show about one of the most consequential problems in American medicine. The free posts are designed to be shared widely. That is their job.

If you have been reading ObGyn Intelligence and finding it useful, this is the series that makes the subscription worth it. Fifty years of obstetric practice, five decades of watching what works and what does not, ten posts making the case that what we have been doing is not working and what we should do instead. Sixty dollars. Forty-six cents per post. Subscribe at obmd.com.

And if you disagree with something I have written, I want to hear it. Evidence-based disagreement is the engine of the enterprise. My contact is on the site.

Amos Grünebaum, MD | Professor of Obstetrics and Gynecology | Maternal-Fetal Medicine Specialist | Senior Ethics Consultant

New York | obmd.com | ObGyn Intelligence -- Evidence Matters

ObGyn Intelligence -- Evidence Matters | obmd.com

Subscribe at obmd.com — $60/year. Less than $1.20 per week for analysis of women’s health written to the standard it deserves.

A 32-year-old woman with gestational diabetes and on Medicaid is told by her obstetrician that she should see a cardiology specialist.

She calls the number she is given. The office does not accept her insurance.

She calls a second number. The next available appointment is in seven weeks. She is 28 weeks pregnant.

Another patient with commercial insurance calls and gets and appointment within 3 days.

This is the referral gap. It is the distance between a clinician recognizing a need and a patient actually receiving the care. It is not a gap in medical knowledge. It is a gap in system architecture.

How Referral Networks Create Barriers

Referral patterns in obstetrics are shaped by insurance networks, geographic proximity, personal relationships between physicians, and the availability of subspecialists. None of these factors are distributed equally across patient populations.

A patient with commercial insurance in an urban area typically has access to a wide network of subspecialists. Her obstetrician can refer her to a maternal-fetal medicine specialist, a genetic counselor, a pelvic floor therapist, or a reproductive psychiatrist, and the patient can usually get an appointment within a reasonable timeframe.

A Medicaid patient in the same city may face a different reality. Fewer subspecialists accept Medicaid. Those who do may have longer wait times. The referring obstetrician may have fewer established relationships with subspecialty practices that accept public insurance. The patient may need to travel farther, take more time off work, or arrange childcare for multiple visits.

A rural patient, regardless of insurance type, faces an additional layer: distance. The nearest MFM specialist may be 60 or 90 miles away. Genetic counseling may be available only by telehealth, if at all. Pelvic floor therapy may not exist in her region.

The Invisible Triage

What makes the referral gap particularly difficult to address is that it operates invisibly. No committee decides which patients will have access to subspecialty care. No policy explicitly excludes Medicaid patients from referral networks. The triage happens through accumulated small decisions: which practices are in which networks, which phone numbers appear in the EMR referral directory, which offices answer the phone quickly.

The result is a two-tier system that does not look like a two-tier system. Everyone is technically referred. Not everyone actually arrives.

Research consistently shows that Medicaid patients are less likely to complete referrals to subspecialists, less likely to receive timely genetic counseling, and less likely to have access to multidisciplinary care teams during high-risk pregnancies. The referral is made. The appointment is not kept, not because the patient is noncompliant, but because the system between the referral and the appointment is not navigable.

Ghost Doctors

A Health Affairs study by Zhu and colleagues just put numbers on something we all knew but rarely quantified:

nearly 28% of Medicaid-enrolled physicians delivered zero care to beneficiaries in 2021. Among psychiatrists, that figure exceeds 40%.

These "ghost physicians" are enrolled on paper, inflating provider directories, while actual care falls on a shrinking core of engaged clinicians.

Language, Literacy, and Navigation

The referral gap is compounded by language barriers, health literacy challenges, and the sheer complexity of navigating the American health care system. A patient who speaks limited English may not understand what a referral to a maternal-fetal medicine specialist means, why it matters, or how to advocate for herself if the first number she calls does not work.

Patients with lower health literacy may not know what questions to ask, what their insurance covers, or what their rights are when a provider is out of network. The referral gap is not just an insurance problem. It is a navigation problem, and the navigation burden falls disproportionately on patients who already face the greatest barriers to care.

Closing the Gap

Solutions exist. Patient navigators, embedded care coordinators, warm handoffs between providers, and integrated referral tracking systems all reduce the distance between referral and specialist visit.

Some health systems have invested in these tools. Many have not.

The question is whether referral completion is treated as a system responsibility or a patient responsibility. If a referral is made and not completed, the current system typically records it as patient noncompliance. A more honest framing would ask: what barriers did the system place between this patient and the care she was told she needs?

A referral that cannot be completed is not a referral. It is a documentation event that protects the system while leaving the patient exposed.

The most common mistake I see when clinicians try Claude for the first time: they open it, type a question, get an answer, close the tab. They do that a few times, decide it is ‘pretty good,’ and move on without changing how they work.

That is like buying a high-quality reference textbook and using it to prop up a monitor.

Claude has features that transform it from a useful search alternative into a genuine clinical thinking partner. Most people never find them. This issue covers the ones that matter most for ObGyn clinicians and for patients managing their health.

Free vs Claude Pro: the honest assessment

The free version works. For occasional questions, for getting a feel for the tool, for one-off tasks -- free is fine.

For daily clinical use, Claude Pro at $20 per month is worth it. Here is the actual difference:

• Long documents. Clinical guidelines, full journal articles, detailed patient summaries. The free version truncates. Pro reads the whole thing.

• Message volume. You will hit the free limit quickly if you use Claude seriously. Pro removes that friction.

• Projects. The most important feature in Claude for clinicians. Only available on Pro.

To upgrade: go to claude.ai, click your profile icon in the bottom left corner, select Upgrade to Pro.

Read more

Labor and delivery. ObGyn. They save lives every day. They also harm patients in ways that are preventable, traceable, and fixable, if you know where to look. This series shows you exactly where to look, and by the time you finish it, you will be a more informed clinician, a more empowered patient, and a more effective advocate for the care that every w…

Read more

Labor and delivery. ObGyn. They save lives every day. They also harm patients in ways that are preventable, traceable, and fixable, if you know where to look. This series shows you exactly where to look, and by the time you finish it, you will be a more informed clinician, a more empowered patient, and a more effective advocate for the care that every woman deserves.

The Case That Followed

The case I described in Episode 2 was not the last time oxytocin and ambiguous language combined to produce a preventable harm on that unit. The second case was different in its details and identical in its mechanism.

A patient was admitted for induction. The attending arrived, assessed her, and told the chief resident to start oxytocin. He then left the floor. Not for surgery this time. Simply gone, in the way attendings sometimes are when a case feels routine and a capable resident is present.

Over the next several hours the clinical picture changed. The oxytocin continued. The fetal heart rate tracing deteriorated in the way tracings do when nobody with authority has made a decision to stop. By the time the attending was back, the situation had become an emergency.

What followed the delivery was a legal proceeding. And in that proceeding something remarkable happened, remarkable but not surprising to anyone who has watched these cases unfold.

The attending said the resident had ordered the oxytocin.

The resident said the attending had directed him to order it.

Both statements were true. And because both statements were true, and because there was no document that said otherwise, the question of who had made the clinical decision, who had assessed this patient and determined that oxytocin was indicated at that dose at that time, became genuinely unanswerable in a court of law.

A verbal order has no author. That is not a legal technicality. It is a patient safety crisis waiting for a clinical situation to reveal it.

The rest of this post is for paid subscribers. The evidence continues below. Subscribe to ObGyn Intelligence to get the full analysis, references, and actionable recommendations.

Read more

Labor and delivery. ObGyn. They save lives every day. They also harm patients in ways that are preventable, traceable, and fixable, if you know where to look. This series shows you exactly where to look, and by the time you finish it, you will be a more informed clinician, a more empowered patient, and a more effective advocate for the care that every woman deserves.

The Legal Instinct That Gets It Backwards

There is a belief, deeply embedded in institutional risk management, that precise language creates liability. The reasoning goes like this: if you write a specific standard, you create something that can be violated and proven in court. If you write a vague standard, you create something that is harder to enforce and therefore harder to litigate against.

This reasoning is wrong. Not occasionally wrong. Systematically, demonstrably, expensively wrong.

The cases I have described in this series did not produce large malpractice settlements because the protocols were too specific. They produced large malpractice settlements because the protocols were too vague to prevent the harm and too vague to defend the institution afterward. Ambiguous language does not protect hospitals. It protects the individual clinician who can argue their behavior was consistent with a standard that meant nothing measurable. The institution, which wrote the standard and is responsible for enforcing it, pays the price.

Specific language is not a liability. It is the closest thing to a liability shield that a clinical protocol can provide. When a standard is clear, it can be followed. When it is followed, outcomes improve. When outcomes improve, litigation decreases. The math is not complicated.

What follows is a list of real words drawn from real society guidelines, including the 2009 Practice Bulletin on Induction of Labor and the 2025 Clinical Practice Guideline on Cervical Ripening, as well as obstetric guidelines more broadly.

These are not obscure documents. They are the foundational texts that govern labor and delivery practice across the United States. Every word has appeared in at least one of them. Every word has, at some point, been used as a shield by a clinician whose patient was harmed.

The evidence continues below. Subscribe to ObGyn Intelligence to get the full analysis, references, and actionable recommendations.

Subscribe now

Read more

Labor and delivery. ObGyn. They save lives every day. They also harm patients in ways that are preventable, traceable, and fixable, if you know where to look. This series shows you exactly where to look, and by the time you finish it, you will be a more informed clinician, a more empowered patient, and a more effective advocate for the care that every woman deserves.

The Moment

I was sitting at the nursing station counter, charting, the way attendings do between cases on a busy labor floor. A nurse approached me. Not my patient, she said, but she needed help. There was a patient in triage. The tracing was concerning. The primary attending was on her way.

I got up. I assessed the patient. I explained to her what was happening and obtained informed consent. I performed a stat cesarean.

The baby did well. The mother did well. And the reason I was able to act, the reason the nurse felt empowered to approach me, the reason there was no delay while someone tried to locate the primary attending or debated whose responsibility this patient was, was a single sentence we had added to our protocol some time before.

Any attending physician present on labor and delivery is responsible for any patient on the unit whose care requires immediate attention.

Nine words into that sentence and the entire culture of individual ownership over labor patients had been replaced with something more honest and more protective. A team.

The Problem the Old Language Create

The rest of this post is for paid subscribers.

Read more

Labor and delivery. ObGyn. They save lives every day. They also harm patients in ways that are preventable, traceable, and fixable, if you know where to look. This series shows you exactly where to look, and by the time you finish it, you will be a more informed clinician, a more empowered patient, and a more effective advocate for the care that every woman deserves.

What the Patient Said

She was sitting across from her attorney. Her baby had a brachial plexus injury. Permanent. The kind that changes a life before it has properly started.

The attorney asked her what happened in the delivery room. She described it. The difficulty. The additional people who came in. The urgency she felt in the room without understanding its source. The baby who came out not breathing well, and then the days that followed, and then the diagnosis.

The attorney asked her: did anyone tell you it was shoulder dystocia?

She said no. No one had used those words. No one had told her this was an emergency with a name. No one had said clearly, in plain language, what was happening to her baby while it was happening.

She was in the room. She was conscious. She heard everything that was said.

And she had never heard the words shoulder dystocia.

That case, and others like it, led to a protocol change that turned out to be about much more than documentation. It was about honesty. About what patients are owed in the worst moments of their care. And about what happens when the words we use in a clinical emergency are chosen to inform the team rather than protect the patient, or worse, chosen to soften what is happening in ways that leave the patient unable to understand what occurred and why.

What Was Being Said Instead

The rest of this post is for paid subscribers.

Subscribe now

Read more

Labor and delivery. ObGyn. They save lives every day. They also harm patients in ways that are preventable, traceable, and fixable, if you know where to look. This series shows you exactly where to look, and by the time you finish it, you will be a more informed clinician, a more empowered patient, and a more effective advocate for the care that every woman deserves.

The Case Nobody Talked About

When I arrived as the new director of a major labor and delivery unit, there was an event that nobody liked to mention. It took me nearly a year to piece together what had actually happened. What I discovered when I finally did changed how I thought about clinical language for the rest of my career.

A patient arrived at term with a clear medical indication for induction. Her private attending physician came in, ordered oxytocin verbally, and left for an early morning gynecology case in the main operating room. He told the chief resident he was leaving. He told no one else. Not the nurses. Not another attending

Over the next several hours, the fetal heart rate tracing deteriorated. The nursing staff grew increasingly concerned. The chief resident was uncertain what to do. Nobody felt empowered to stop the oxytocin or escalate beyond the resident. The attending was in surgery. The protocol said he needed to be available. The nurses were not empowered to stop the pitocin.

He was available. He was in the building.

The fetal heart rate eventually became a sustained bradycardia. An emergency cesarean was performed. The baby was born with a low Apgar score and later developed cerebral palsy. A malpractice suit followed. The settlement ran into seven figures. The hospital paid over ninety percent of it.

The attending’s position was straightforward: he had not been there, therefore it was not his fault. He had performed the cesarean. He had responded when called.

He had also been available. The protocol said so.

The Word That Did This

The protocol governing oxytocin administration on that unit contained a sentence that exists, in some form, in obstetric guidelines across the country. It required that a physician be “available” during labor induction with oxytocin.

That sentence felt like a standard. It read like a standard. It was not a standard. It was a geography-free, time-free, accountability-free suggestion that could be satisfied by a physician in an operating room three floors away, unreachable for the duration of a major gynecologic case, while a fetal heart rate tracing told a story that nobody on the floor had the authority to act on.

“Available” is not a location. It is not a time. It is not a credentialing requirement. It is a word that means whatever the person using it needs it to mean, and in a deposition, it will mean whatever their attorney needs it to mean.

This is how a single word becomes a patient safety crisis.

The evidence continues below. Subscribe to ObGyn Intelligence to get the full analysis, references, and actionable recommendations.

Subscribe now

Read more

Labor and delivery. ObGyn. They save lives every day. They also harm patients in ways that are preventable, traceable, and fixable, if you know where to look. This series shows you exactly where to look, and by the time you finish it, you will be a more informed clinician, a more empowered patient, and a more effective advocate for the care that every woman deserves.

The Case

The patient was being induced at term. Her attending arrived, assessed her progress, decided labor was moving too slowly, and increased the oxytocin. To the upper end of the permitted range. More rapidly than the evidence supported. Within the boundaries of what the guideline allowed.

Tachysystole followed. More than five contractions in ten minutes. The fetal heart rate tracing deteriorated. An emergency cesarean was performed.

When we reviewed the case, nothing had been violated. The attending had acted within the permitted range. The documentation was complete. The guideline had been followed to the letter.

The patient had still been harmed.

When we looked at the outcomes data more broadly, the pattern was not isolated. Cases involving attending-driven dose escalation, particularly toward the upper end of the permitted range, were associated with higher rates of tachysystole, more fetal heart rate anomalies, and more emergency cesarean deliveries than cases managed conservatively at the lower end of that same range.

The guideline had not caused this. It had permitted it.

And to understand why, you need to look carefully at the table.

The Table That Looked Like Precision

Subscribe now

Read more

A Practical Guide for Clinicians and Patients | ObGyn Intelligence

Let me start with what Claude is not.

It is not a search engine. It is not a chatbot in the sense of a customer service bot with scripted responses. It is not ChatGPT, though they are in the same category. And it is not -- I want to be clear about this -- a replacement for clinical judgment.

What it is: a reasoning tool. You give it text -- a question, a document, a clinical scenario, a draft -- and it reasons through it and responds. The responses are original, not retrieved from a database. It reads the whole thing. It follows instructions. It remembers the context of your conversation. And it does all of this in seconds.

I have been using it daily for over a year. It has changed how I work. Not in a dramatic, transformative way -- in a quiet, practical, this-saves-me-45-minutes-today way. That is what I want to share with you.

Subscribe now

How it is different from Google

When you Google a clinical question, you get links. You click through, scan, find what you are looking for, and piece together an answer. That process takes time and requires you to evaluate each source.

When you ask Claude the same question, you get an answer. With reasoning behind it. And you can ask follow-up questions, push back on the answer, and ask it to reframe the information for a patient who is frightened. It is not looking up the answer -- it is synthesizing it.

The difference matters most for complex questions. Straightforward lookups are fine with Google. Nuanced clinical reasoning, evidence synthesis, writing tasks -- that is where Claude pulls ahead.

How it is different from ChatGPT

Both are large language models. Both are capable. The reasons I use Claude over ChatGPT for clinical work come down to three things.

First, honesty. Claude is specifically trained to acknowledge uncertainty. In medicine, a tool that says ‘I am not certain about this’ is more useful than one that sounds confident and is wrong. I have caught errors in both -- but Claude hedges appropriately more often.

Second, document handling. I can upload a 40-page guideline, a full journal article, a patient’s medication list, or a long clinical summary, and Claude reads and analyzes the entire thing. The context window is large and it uses it well.

Third, instruction-following. When I tell Claude to give me absolute risk numbers and not relative risk, to write at a 7th grade reading level, or to flag its uncertainty -- it does. Consistently.

For clinicians: why ObGyn specifically

Our specialty sits at the intersection of several areas where AI adds immediate value.

The evidence base moves fast. New guidelines, new trials, new recommendations arrive constantly. Nobody reads everything. Claude does not solve that -- but it dramatically lowers the cost of staying current. A 12-page RCT that you do not have time to read gets triaged in 60 seconds.

Patient education is central to what we do. We explain pregnancy to frightened first-time mothers. We counsel women about menopause. We discuss fertility options with couples who have been trying for three years. We need clear, accurate language. Claude helps find it.

Documentation burden is real. Notes, letters, prior authorizations, handouts, consent forms, referral summaries. Claude handles drafts. You review and sign off. That is a different relationship with paperwork than the one most of us have now.

Subscribe now

None of this replaces what only you can do: examine the patient, make the diagnosis, bear the clinical responsibility. Claude handles the information-intensive tasks surrounding clinical judgment so you have more of yourself left for the judgment itself.

For patients: what this means for your care

If your doctor mentions using AI tools, it does not mean a computer is making decisions about your care. It means your doctor is using a tool to work more efficiently and communicate more clearly -- the same way they use ultrasound to see what their hands cannot feel.

But you can also use Claude yourself. Not to diagnose. Not to replace your doctor. But to understand what your doctor told you. To prepare questions before your appointment. To look up what a word in your test results means. To think through a decision you are facing.

Most patients leave appointments with unanswered questions. Not because the questions are unimportant -- because there was not enough time, or because they did not have the words yet, or because the answer was given too fast to absorb. Claude gives you somewhere to take those questions at 11pm when the office is closed.

The rules are simple. Use it to prepare and to understand. Not to diagnose. Not in emergencies. And always bring what you learn to your next conversation with your doctor.

Subscribe now

What is coming next

Setting up Claude the right way. The difference between using Claude as a one-off tool and configuring it so it already knows your specialty, your preferences, and your writing style before you type your first question. This is the issue that changes everything.

Next: Setting up Claude the right way. The feature most clinicians never find.

Amos Grunebaum, MD | ObGyn Intelligence | obmd.com | Educational purposes only.

Share

The most common mistake I see when clinicians try Claude for the first time: they open it, type a question, get an answer, close the tab. They do that a few times, decide it is ‘pretty good,’ and move on without changing how they work.

That is like buying a high-quality reference textbook and using it to prop up a monitor.

Claude has features that transform it from a useful search alternative into a genuine clinical thinking partner. Most people never find them. This issue covers the ones that matter most for ObGyn clinicians and for patients managing their health.

Subscribe now

Free vs Claude Pro: the honest assessment

The free version works. For occasional questions, for getting a feel for the tool, for one-off tasks -- free is fine.

For daily clinical use, Claude Pro at $20 per month is worth it. Here is the actual difference:

• Long documents. Clinical guidelines, full journal articles, detailed patient summaries. The free version truncates. Pro reads the whole thing.

• Message volume. You will hit the free limit quickly if you use Claude seriously. Pro removes that friction.

• Projects. The most important feature in Claude for clinicians. Only available on Pro.

To upgrade: go to claude.ai, click your profile icon in the bottom left corner, select Upgrade to Pro.

Subscribe now

Read more

I teach this to our doctors now. Not how to use Claude -- how to talk to it.

The tool is only as good as the question. That is true of consultants, of literature searches, of colleagues you ask for advice. It is especially true of AI, because AI will give you a confident-sounding answer regardless of whether your question was good enough to deserve one.

The difference between a vague prompt and a precise one is not a small improvement. In my experience, it is the difference between an answer I can use and one I ignore. This issue covers the principles that make that difference, with specific examples from ObGyn practice.

The single most important habit

Tell Claude who you are, what you need, and what format you want. Every time. In that order.

Most people ask Claude the way they would type into Google -- short, keyword-based. That produces a Google-style answer: broad, generic, and not particularly useful for clinical decision-making.

Here is the same question asked two ways:

Prompt: What is the treatment for preeclampsia?

versus:

Prompt: I am an ObGyn. My patient is 34 weeks with severe preeclampsia, BP 158/105, on magnesium sulfate. She is asking whether she needs to deliver now or whether we can manage expectantly for another week. Summarize the current evidence on expectant management of severe preeclampsia at 34 weeks, with absolute risk numbers, and give me the key points from ACOG guidance.

The second prompt takes 20 extra seconds to write. The response is ten times more useful. That ratio -- small investment, large return -- holds for almost every clinical question.

Become a paying subscriber of ObGyn Intelligence to get access to this post and other subscriber-only content.

Subscribe now

Read more

The first three issues covered what Claude is, how to set it up, and how to talk to it. Now we use it.

I am going to be specific. Exact prompts. Exact steps. Expected output. This is not a conceptual overview -- it is a walkthrough you can follow right now with a real article or a real document in front of you.

By the end of this issue you will have done something useful. That is the point.

For clinicians: the journal article summary

This is the task I use most. Every week, several times a week.

The problem is familiar: you have a paper you need to read and no time to read it properly. The trial on aspirin dosing. The meta-analysis on cervical length measurement. The guideline update you heard about at conference. You need to know what it says and whether it changes anything. You have four minutes before your next patient.

Here is what I do.

Become a paying subscriber of ObGyn Intelligence to get access to this post and other subscriber-only content.

Subscribe now

Read more

Anthropic -- the company that makes Claude -- has published 26 free video tutorials on how to use it. Short videos. Specific topics. No account required to watch.

I found these about six months into using Claude. I wish I had found them sooner. Several of them showed me features I had been missing for months.

This issue gives you the full list with direct links, organized by relevance, with notes on which ones matter most for ObGyn clinicians and for patients. You do not need to watch all 26. I will tell you which ones to start with.

The full collection is here: support.claude.com/en/collections/10548294-video-tutorials

The rest of this post is for paid subscribers.

Subscribe now

Read more

I want to be honest about something. The average ObGyn probably reads fewer than 10% of the relevant journal articles published in their field each year. Not because we do not care. Because there are too many, the days are too full, and by the time we sit down, the energy for dense methodology sections is gone.

The result is a slow, invisible drift between what the evidence shows and what we actually do. We practice on autopilot. We rely on what we learned in residency. We wait for the guideline update instead of reading the trial that will drive it.

Claude does not fix this entirely. But it changes the math. A paper that would take 25 minutes to read carefully takes 60 seconds to triage. You learn what it found, whether the design was sound, and whether it should change anything you do. Then you decide if it deserves your full attention.

Become a paying subscriber of ObGyn Intelligence to get access to this post and other subscriber-only content.

Subscribe now

Read more

I have seen patient handouts that were last updated in 2009. I have seen handouts written at a 12th grade reading level for patients who never finished high school. I have seen handouts that answer every question except the one the patient actually has.

Most of us use whatever came with the EMR, or whatever was in the clinic when we arrived, or whatever we downloaded from ACOG five years ago and never looked at again.

Claude fixes this. Not because it knows more than you do -- it does not. But because it can take what you know and translate it into clear, specific, appropriately leveled language faster than you can type the title of the handout.

This issue walks you through the exact process. By the end you will have a handout framework you can reuse for any diagnosis, procedure, or condition.

The rest of this post is for paid subscribers.

Subscribe now

Read more