The head delivers.

Then it stops.

The shoulders do not follow.

The room changes instantly.

Someone announces shoulder dystocia.

The legs are flexed.

Suprapubic pressure is applied.

Additional maneuvers follow.

Seconds become very long.

Eventually the baby is delivered.

Then one arm does not move normally.

A brachial plexus injury is diagnosed.

Months later, a question appears:

Who pulled too hard?

Sometimes that is the wrong question.

Shoulder Dystocia Is an Emergency, Not a Diagnosis We Can Reliably Predict

Read more

On September 17, the subject line read: “Video: Nurse Attacked; Doc’s High-Speed Arrest; Fake Ob/Gyn Pleads Guilty.”

Subscribe now

I have read MedPage for years.

Lately it felt different.

More arrests.

More killings.

More doctors in handcuffs.

But a feeling is not data. So I counted.

What I measured

MedPage sends a newsletter called “Exclusives & Features” twice a week. The Thursday edition is a roundup of three legal stories. I went through every Thursday edition still in my inbox, from 2023 to this week. That came to 54 issues and 150 headline items. I sorted each item into one of five groups: a clinician accused or convicted of a crime, violence against a clinician, a malpractice suit or other civil case, a licensing action, and everything else.

Share

In August 2023, crime and violence made up 1 in 3 items. Half were malpractice suits and verdicts. By 2025, crime and violence had risen to 6 in 10. This year they sit at about half. Malpractice news, the part a practicing clinician can actually learn from, fell to about 1 in 4.

Two more things stood out. About half of the criminal items describe an accusation, an arrest or an investigation, not a conviction.

Some headlines end in a question mark, such as “Doc Hit Dementia Patient?” A question mark is not a verdict. It reads like one.

The Monday edition changed too.

From April to September 2025, none of its 23 lead stories was a medical oddity. Most were policy reporting.

From June to September 2026, 13 of 17 were oddities, with headlines like “Eyeball Tattoo Goes Sideways.”

One finding cuts the other way, and it belongs here. MedPage’s Weekly Review, the editors’ pick of the week’s best stories, featured these crime roundups a handful of times in 2024 and 2025. In 38 issues this year, it featured none. MedPage is not pushing crime onto its front page.

My data have limits. This is my inbox, not MedPage’s full archive, and some months are missing. The 2023 sample is small. I read headlines, not the articles, and I coded them alone. The direction is clear. The exact percentages are not precise.

The best case for the column

Accountability journalism matters. Impostors who pose as doctors, clinicians who harm patients, licensing boards that fail to act: the profession needs to know. Violence against nurses is a real workplace danger. Silence protects the wrong people.

I agree with every word.

That is exactly why this column fails.

Why it fails

A professional news service has a different job than a tabloid. Coverage of a colleague’s wrongdoing earns its place when it teaches. How did the impostor get hired? Why did the board miss the warning signs? What did the hospital change after the attack? A three-item list of arrests teaches none of that. “Doc Accused of Luggage Theft” does not make anyone a better doctor.

Journalists have their own code of ethics. It tells reporters to “avoid pandering to lurid curiosity, even if others do.” It asks them to weigh a suspect’s right to a fair trial and to think hard before naming someone who has not been charged. It asks them to update stories as facts change (1). Charges get dropped. Cases get dismissed. My sample includes both. A one-line accusation stays online long after the case ends.

In our work, beneficence means seeking the greatest good for the person who trusts us. A news service for clinicians owes its readers the same thing: stories that make them better at their jobs. Spectacle does the opposite. It spends clinicians’ limited attention on the worst colleague in the country, not on the lesson that might protect their own patients.

Conclusion

MedPage’s flagship original-reporting newsletter has drifted from accountability journalism toward spectacle.

It names clinicians before adjudication, emphasizes lurid detail, and offers no systemic lesson.

I do not know why the mix changed, and I will not guess. I do know what a professional news service owes its readers. Report a crime when it exposes a system failure, and explain the failure. Wait for charges before naming anyone. Follow up when charges are dropped. And give Thursday back to the malpractice verdicts that actually teach.

Check your own inbox this Thursday and count the arrests. Then share this with a colleague who still calls it must-read.

If careful counting is what you want from medical news, subscribe to ObGyn Intelligence.

References

1. Society of Professional Journalists. SPJ Code of Ethics [Internet]. Society of Professional Journalists; 2014 [cited 2026 Sep 25]. Available from: https://www.spj.org/spj-code-of-ethics/

Data source: author’s analysis of MedPage Today newsletter emails received August 2023 to September 2026 (Thursday “Exclusives & Features” roundups, Monday editions, and Weekly Review issues). Headline items coded by a single reviewer.

Biology does not.

A woman delivers.

Her baby is healthy.

She walks out of the hospital carrying a car seat, discharge papers, medications, and instructions.

Everyone feels that the dangerous part is over.

Sometimes it has barely begun.

Delivery Is an Event. Postpartum Is a Process.

The United States recorded 649 maternal deaths under the National Center for Health Statistics definition in 2024, a maternal mortality rate of 17.9 per 100,000 live births. Importantly, that definition captures maternal deaths during pregnancy or within 42 days after pregnancy and therefore does not capture the entire year of pregnancy-associated risk.

Clinical risk does not disappear at discharge.

Hypertension can worsen.

Infection can evolve.

Venous thromboembolism can occur.

Cardiomyopathy may first become clinically apparent.

Bleeding can recur.

Surgical wounds can deteriorate.

Psychiatric illness can become life-threatening.

A national claims-based cohort followed nearly 460,000 deliveries for one year. Women who experienced severe maternal morbidity during delivery remained at increased risk of readmission throughout that entire year. Sepsis and hypertensive disorders were among the major reasons for subsequent readmission.

That makes postpartum care an essential part of obstetric safety, not an epilogue.

The Dangerous Handoff Is Often to Nobody

Inside the hospital, responsibility is visible.

There is an obstetrician.

A nurse.

An anesthesia team.

A charge nurse.

A blood bank.

An operating room.

Then the patient goes home.

Who owns her worsening blood pressure at 9 PM three days later?

Who owns the wound photograph she sends through the portal?

Who owns the headache mentioned to an answering service?

Who sees the emergency department note from another hospital?

Who makes sure the patient actually receives the follow-up that was recommended?

This is where postpartum care can become fragmented.

The patient has left the building, but she has not left the disease process.

Read more

That sentence appears in too many serious adverse-event reviews.

The nurse was worried.

The resident was worried.

Anesthesia knew.

The attending had been called.

The tracing was deteriorating.

The patient’s condition was changing.

Everyone possessed part of the information.

But nobody owned the escalation.

Share

Communication Is Not the Same as Escalation

Clinicians communicate constantly.

“FYI.”

“Doctor aware.”

“I sent a message.”

“I told the resident.”

“I mentioned it during sign-out.”

Information moved.

But escalation requires more.

Escalation means that information reaches someone with the authority and ability to make the next decision, and that the communication produces a defined response.

That distinction matters enormously in obstetrics.

Labor and Delivery Is a Team Sport Played at High Speed

Read more

The ultrasound looked reassuring.

The bleeding stopped.

Two years later, she bled again.

Someone wrote:

“Previously evaluated.”

Six months after that:

endometrial cancer.

The first evaluation may have been entirely appropriate.

The mistake may have been believing that it was valid forever.

Subscribe now

Postmenopausal Bleeding Is a Symptom, Not a Diagnosis

Most women with postmenopausal bleeding will not have endometrial cancer.

That matters.

Medicine should not turn every episode of bleeding into catastrophe.

But postmenopausal bleeding is clinically important because endometrial cancer frequently presents with bleeding.

Evaluation therefore aims to identify who requires endometrial assessment while avoiding unnecessary invasive procedures in everyone.

Read more

The Society for Maternal-Fetal Medicine published a new statement this month. It deals with a hard problem: how to care for women with dangerous preg…

Read more

The form is complete.

Her name is correct.

The physician signed it.

The risks are listed.

The date and time are present.

Legally and administratively, everything may look excellent.

There is only one problem.

The patient may still not understand what she agreed to.

That is not a paperwork problem.

It is an informed consent problem.

Consent and Informed Consent Are Not the Same Thing

A signature proves remarkably little.

It proves that a person signed something.

It does not prove that she understood the probability of success.

It does not prove that she understood the alternatives.

It does not prove that she understood the consequences of declining treatment.

It does not prove that the clinician explained the information in a way she could understand.

ACOG’s guidance is clear that valid informed consent requires more than disclosure. The patient must receive adequate, accurate, understandable information and be able to understand and reason through it while making an intentional and voluntary choice. [1]

That is a much higher standard than obtaining a signature.

Obstetrics Makes Consent Especially Difficult

Consent becomes harder when decisions occur during labor.

Read more

What was announced

On September 10, 2026, Olivia and Tom Walton committed an initial $100 million over five years to Healthy Moms Healthy Babies America (HMHBA), a program housed within Heartland Forward’s Maternal and Child Health Center for Policy and Practice, with Neel Shah as president and Robin Reck as executive director.

The stated goal is to cut the US maternal mortality rate in half within five years.

The money is described as catalytic capital for state matching grants, “high-impact partnerships,” and infrastructure, with four pillars:

1. prenatal care (virtual care, community health workers, doulas, midwives),

2. “protecting maternity units as more units close,”

3. a first year of family support (postpartum home visits, mental health screening, paid leave), and

4. better data (2,3).

On closures the language is careful.

The press release says hospitals “are closing obstetric units at an accelerating pace,” and the Call to Action says “hundreds of U.S. hospitals, disproportionately in rural communities, have closed maternity wards,” attributing this to liability cost and workforce shortage and stating that families “must travel farther,” disrupting continuity of care. Neither document states that closures caused the rise in maternal deaths; the causal inference is left to the reader, and the press coverage has supplied it (2,3). On midwives, the Call to Action cites the Lancet Global Health modelling estimate that midwifery could avert 41% of maternal deaths, and on doulas it cites a 52.9% reduction in cesarean odds and 57.5% reduction in postpartum depression or anxiety (3). Each of these claims is examined below against the primary literature.

The baseline matters for a “halve it” target. NCHS reports 649 maternal deaths in 2024, a rate of 17.9 per 100,000 live births, not significantly different from 18.6 in 2023; the non-Hispanic Black rate was 44.8, the non-Hispanic White rate 14.2, the Hispanic rate 12.1, and the rate at age 40 and over 62.3 (1). HMHBA’s own website uses a pooled 2019–2023 figure of 23.5, which includes the pandemic years. Halving 17.9 means roughly 9 per 100,000 by 2031; halving 23.5 means roughly 12, a number the country is arguably within measurement error of already. Which baseline is chosen will determine whether the program is later judged a success.

Claim 1: obstetric unit closures drive maternal mortality

The closures are real and well counted. Among 4,964 short-term acute-care hospitals, the share without obstetric services rose from 35.2% in 2010 to 42.4% in 2022; 537 hospitals lost obstetric care (238 rural, 299 urban) and 138 added it (5). Kozhimannil’s original rural analysis found that 179 rural counties lost hospital obstetric services between 2004 and 2014 (4). March of Dimes counts 96 labor and delivery closures across 35 states from January 2024 to May 2026, 58% of them the county’s sole birthing facility, and estimates 1,113 counties (34.6%) as maternity care deserts (6). None of this is in dispute.

What is in dispute is the outcome. No quasi-experimental study of closures in the United States has demonstrated an effect on maternal death. The closure literature measures place of birth, travel time, prenatal visit counts, preterm birth and, less often, maternal morbidity, and the results are summarized in the table below.

The only positive severe morbidity signal comes from New Jersey, a largely urban and suburban setting (10). The most rigorous rural analysis, covering 605 closures over three decades, found maternal morbidity slightly better after closure, which the authors attribute to women delivering at higher-volume, higher-quality hospitals (7). That interpretation is consistent with the volume literature: among low-risk women, delivery at rural hospitals with 10 to 110 births a year carried an adjusted SMM risk ratio of 1.65 compared with rural hospitals above 460 births (14). Keeping the smallest units open is therefore not self-evidently protective; it may trade one risk for another.

The mortality signal that advocacy documents cite comes from cross-sectional comparisons of places, not from closures. Atwani and colleagues, using county-level data on 14.8 million births from 2018 to 2021, found a maternal mortality rate of 32.25 per 100,000 in desert counties versus 23.62 in full-access counties (adjusted incidence rate ratio 1.36, 95% CI 1.21–1.54), with no difference for low- or moderate-access counties (12). In Louisiana, desert residence carried an adjusted relative risk of 3.37 for pregnancy-related mortality on 112 deaths (13). Driving distance to the delivery hospital in Pennsylvania was associated with a composite of transfusion, ICU admission and hysterectomy (aRR 1.53 at 80 km), but the authors themselves ask whether distance is a cause or “simply a marker of social deprivation” (11). Desert counties are poorer, sicker, more rural and more often in the South; the same counties had high maternal mortality before their units closed.

The honest summary is this: closures worsen access, lengthen travel by about 14 to 25 minutes on average, push a small fraction of births out of hospital, and in rural counties may modestly increase preterm birth (4,6). There is no evidence that they have contributed measurably to the national maternal mortality rate, and the best-designed study could not detect an effect because maternal death in closure counties is too rare to study. Any statement that closures are “responsible for” US maternal mortality outruns the data. The more defensible claim is that they are a threat to timely care in specific communities, which is a legitimate reason to fund rural obstetric readiness and transfer systems, but not a mechanism that will halve the national rate.

Claim 2: more midwives and doulas will reduce maternal deaths

This claim rests on three literatures, none of which measures maternal mortality in a high-income country.

The 2024 Cochrane review of midwife continuity models pooled 17 randomized trials with 18,533 women in Australia, Canada, China, Ireland and the United Kingdom (no US trial). It found moderate-certainty increases in spontaneous vaginal birth (RR 1.05) and reductions in cesarean (RR 0.91) and regional analgesia; the 2016 review’s apparent reductions in preterm birth and fetal loss did not persist in the update (preterm RR 0.95, 95% CI 0.78–1.16; neonatal death RR 0.85, 0.43–1.71, low certainty). Maternal death was a prespecified outcome: no deaths occurred in the three trials reporting it, so the effect is unestimable rather than null (15). The 2017 Cochrane review of continuous labor support (27 trials, 15,858 women) found a cesarean RR of 0.75 and better birth experience; maternal mortality does not appear among its outcomes (16).

The US doula evidence is observational Medicaid claims work. The Minnesota comparison behind the original cost argument found cesarean 22.3% versus 31.5% (17). The eClinicalMedicine study cited by HMHBA compared 298 propensity-matched pairs across three states and reported cesarean OR 0.47 and postpartum depression or anxiety OR 0.43; it did not report maternal death or severe morbidity (18). The 2024 follow-up on 722 matched pairs found lower cesarean and preterm rates and more postpartum visits, but no difference in emergency visits, admissions, depression or severe maternal morbidity within 30 days, and the authors state they “did not evaluate maternal mortality, because we lacked statistical power” (19). A 2026 systematic review of midwifery and doula care specifically for Black women located 16 studies and concluded the literature “remains severely limited”; none reported mortality or SMM (58). Both the Elevance studies were authored by the insurer that funds the benefit, and propensity matching cannot remove the selection of motivated, engaged women into doula programs.

The “41% of maternal deaths” figure is a Lives Saved Tool model of 88 low- and middle-income countries that together account for 98% of the world’s maternal deaths. It estimates what would happen if midwife-deliverable interventions (uterotonics, antibiotics, antihypertensives, skilled attendance) were scaled from current low coverage; the authors describe the results as “indicative and directional rather than exact.” No high-income country is modelled, and the interventions being credited are already nearly universal in US hospitals (20). Citing this number for the United States is a category error.

Ecological comparisons are the remaining support. The Commonwealth Fund notes the United States has about 4 midwives and 12 obstetrician-gynecologists per 1,000 live births against 61 and 13 in the United Kingdom and 81 and 5 in Sweden (23). Vedam’s state integration index correlated with vaginal birth, breastfeeding and neonatal outcomes, not maternal mortality (21). But the country with the most midwife-led system in Europe, the United Kingdom, reports a maternal mortality rate of 12.80 per 100,000 maternities for 2022–2024, described by MBRRACE as roughly 20% higher than in 2009–2011 and rising (25). The Netherlands, the archetype of community midwifery, had a confidential-enquiry rate of 12.1 in 1993–2005 that fell to 6.2 in 2006–2018 while home birth declined and hospital care was reorganized (26). Nothing in either series attributes change in mortality to midwifery workforce size.

Midwives and doulas have a solid evidence base for what they do: fewer cesareans, fewer instrumental births, less regional analgesia, better experience, and, for doulas in Medicaid populations, plausibly fewer preterm births. These are worthwhile ends and they may reduce the morbidity that follows surgery.

But a program that promises to halve maternal deaths and then invests in workforce categories whose trials have never recorded a maternal death is promising an outcome the intervention has not been shown to deliver.

The one caveat on the other side is that a US planned home birth attended by a midwife carries a roughly threefold higher neonatal mortality than hospital birth (22); “more midwives” is only safe as “more midwives in integrated hospital and birth-center systems.”

The European comparison: how much of the gap is real

The claim that US mortality is “more than double” or “three times” the European rate rests on comparing the US vital-statistics number with OECD civil-registration counts for Europe. Both sides of that comparison are biased, in opposite directions.

On the European side, Euro-Peristat declines to publish routine maternal mortality at all because certificate data are “unreliable.” The eight-country enhanced-surveillance study in the BMJ found that vital statistics “underestimated maternal mortality by 36% or more in all countries except Denmark,” and produced rates per 100,000 live births of 2.7 in Norway, 3.4 in Denmark, 4.3 in Finland, 5.3 in the Netherlands, 6.3 in Italy, 8.0 in France, 9.6 in the United Kingdom and 10.9 in Slovakia (24). Germany, which has no confidential enquiry, reports 3.5 through Destatis, but a certificate-by-certificate review in Berlin found 7.8 to 9.1, “over double the official estimate” (27). The Dutch audit figure of 6.2 is more than double the OECD figure of 2.8 used in the Commonwealth Fund exhibit (23,26).

On the US side, the pregnancy checkbox inflates the count. NCHS’s own linkage found that in 2014 and 2016 more than half of certificates with a positive checkbox had no pregnancy-related hospitalization, and that if the checkbox were ignored entirely the 2018 rate would have been 8.7 rather than 17.4 (28). Rossen and colleagues attributed 9.6 deaths per 100,000 to the checkbox and found no significant trend once it was removed (29). Joseph and colleagues showed that when pregnancy must appear among the recorded causes of death rather than only on the checkbox, the rate is 10.2 in 1999–2002 and 10.4 in 2018–2021, with no increase; their bias-corrected sensitivity estimate is about 16 to 17 in both periods (30). Declercq and Thoma found that for the same 42-day window, NCHS runs 44% to 63% above the CDC Pregnancy Mortality Surveillance System (31). The rebuttal from Janevic, Declercq and Howell is that the checkbox also has false negatives: 62% of decedents with a pregnancy hospitalization in the prior year were not flagged, and New York’s review committee identified 386 potential deaths of which only 129 were identifiable from certificates alone (32). Both positions have merit; the debate is about the size of the correction, not whether one is needed.

A like-for-like comparison therefore places the United States at roughly 16 to 18 per 100,000 against 13 in the United Kingdom, 8 to 9 in France and Germany, 6 in the Netherlands and 3 to 4 in Scandinavia. The gap is real, but it is a factor of 1.3 to 3, not 5 to 10, and the “two-thirds lower” figure that HMHBA and others use is an artefact of comparing a checkbox-inflated numerator with under-ascertained European ones. One consequence for the Waltons’ target deserves saying plainly: if NCHS ever adopts Joseph’s multiple-cause criterion, the national rate will “halve” on paper without a single death prevented, and if states finish building enhanced surveillance, the rate will rise. A program that promises a 50% fall measured by a statistic this unstable has set itself a target that measurement changes can meet or defeat.

What is not an artefact is the Black rate. At 44.8 per 100,000 in 2024, it exceeds every European figure under any method, and the composition of US deaths differs from Europe’s. In the 36-state review committee data for 2017–2019, mental health conditions accounted for 22.7% of pregnancy-related deaths, hemorrhage 13.7%, cardiac conditions 12.8%, infection 9.2%, thrombotic embolism 8.7% and cardiomyopathy 8.5%, with 84% judged preventable and 30% occurring between 43 and 365 days (33). France and the United Kingdom are dominated by cardiac disease, thrombosis and suicide; hemorrhage and infection deaths of the kind that safety bundles address have largely been driven out. Outside the WHO definition altogether, US pregnancy-associated homicide (3.62 per 100,000 live births, exceeding every leading obstetric cause) and overdose (11.85 per 100,000 in 2020, rising to 13.0 by 2023) have no European counterpart at that scale (34,35,36). The US excess is real, but a substantial part of it is not obstetric.

What reduces maternal mortality among Black women

The disparity is not explained by education or income. In the 2007–2016 pregnancy mortality data, Black women with a college degree or higher had a ratio of 40.2 per 100,000 against 7.8 for White women with the same education, a 5.2-fold difference, and higher than White women who had not finished high school (25.0) (37). The ratio widens with age, from 1.5 under 20 to 4.3 at 30–34, consistent with cumulative physiological burden. The leading causes among Black women are cardiac and coronary conditions (15.9%), cardiomyopathy (13.9%), thrombotic embolism (11.9%) and hypertensive disorders (9.9%); more than half of cardiomyopathy deaths in the 32-state review were among non-Hispanic Black women, and more than three quarters were judged preventable (33,41).

A large part of the disparity is located in which hospital a woman delivers in. In New York City, 65% of White deliveries but 23% of Black deliveries occurred in hospitals in the lowest SMM tertile, and Howell estimated that site of delivery accounted for up to 47.7% of the Black-White SMM gap (38). Within the same hospital, the adjusted odds of SMM for Black women remained 1.52, independent of insurance (39). Nationally, among 73.9 million delivery hospitalizations, the failure-to-rescue rate after SMM was 1.79 times higher for Black women (40). The disparity is therefore three things at once: worse hospitals, worse care within hospitals, and worse rescue when things go wrong. Each is a hospital-quality problem, which is the one domain in which we have interventions with measured effects.

The following table ranks interventions by the strength of evidence for reducing mortality or severe morbidity specifically among Black women.

Two cautions belong beside this table. First, California’s overall rate halved to about 7.0 per 100,000 by 2013, on par with Western Europe (44), but its Black pregnancy-related mortality was 56.2 per 100,000 in 2014–2016 and 49.7 in 2019–2021 against 14.0 for White women, and the state’s own reviewers reported that the disparity widened between 2008 and 2016 (45). Hemorrhage bundles narrowed the SMM gap; they did not close the mortality gap, because the deaths of Black women in California are increasingly cardiovascular, and the bundle does not treat cardiomyopathy. Second, a 2026 scoping review found only 16 peer-reviewed intervention studies aimed at maternal disparities in the entire US literature and described them as “limited” in rigor (57). No intervention has yet demonstrated a reduction in Black maternal mortality in a controlled analysis. What has been demonstrated, repeatedly, is that when hospitals measure SMM by race and run standardized hemorrhage and hypertension protocols, Black women benefit disproportionately, because they were disproportionately harmed by the absence of those protocols.

The implication for priorities follows from the cause distribution. Cardiac disease, cardiomyopathy, hypertension and embolism together account for roughly half of pregnancy-related deaths among Black women, and 30% of all deaths occur after six weeks, when the woman has left obstetric care (33). The interventions with the best mechanistic fit are the ones that treat those conditions: universal postpartum blood pressure surveillance with remote monitoring, treatment of chronic hypertension from the first trimester (50), cardio-obstetric pathways for women with cardiomyopathy risk factors, structured postpartum follow-up in the first year with insurance to pay for it, and objective blood-loss measurement with escalation protocols. Review committees also judge more than 90% of preeclampsia and mental-health deaths preventable (60), which argues for the same postpartum year of contact HMHBA proposes, but delivered by clinicians who can prescribe an antihypertensive, not only by a visitor who can recognize a warning sign.

Does the plan address access to physicians and specialists?

No.

The press release does not mention obstetricians, maternal-fetal medicine, cardiologists, anesthesiologists, hospital quality, safety bundles or perinatal quality collaboratives; its workforce language is “growing the maternal health workforce with an emphasis on training midwives” and scaling “integrated care teams, including midwives and doulas” (2).

The Call to Action goes further in the opposite direction. It describes prenatal care as having “traditionally centered on obstetricians as the primary, and often sole, providers of care” and proposes “shifting from siloed OB-GYN care to a broader suite of maternal health providers” so that services “become more accessible.”

It acknowledges that “cardiovascular disease is the leading cause of maternal death, yet prenatal care often focuses narrowly on pregnancy-specific conditions,” but proposes no mechanism for getting a woman with cardiomyopathy risk to a cardiologist.

Physician supply, subspecialist consultation, levels of maternal care, transfer protocols and hospital quality measurement do not appear (3). The plan is, in effect, a substitution plan: fewer encounters with physicians, more with midwives, doulas, community health workers and telehealth.

That is a defensible strategy for low-risk women and a poorly matched one for the women who die.

The evidence on specialist access is thinner than one would like, but what exists points one way. Sullivan and colleagues showed in 2005 that an increase of five maternal-fetal medicine specialists per 10,000 live births was associated with a 27% reduction in state maternal mortality (RR 0.73, 95% CI 0.58–0.93) after adjustment for poverty, education, race and age (61). Kawakita and colleagues replicated this with 2018–2021 data: states in the highest tertile of MFM density had an adjusted incidence rate ratio of 0.70 (0.58–0.85) for maternal mortality and 0.83 (0.71–0.98) for pregnancy-related mortality compared with the lowest, although the middle tertile showed no effect, so the relationship is not a clean gradient (62). An HRSA analysis found the South had the lowest availability of nearly every relevant occupation, obstetrician-gynecologists included, and the highest mortality, while explicitly disclaiming causation (63). These are state-level ecological studies with the usual limits, and none adjusts for hospital quality, but they are the only mortality data we have on any workforce category, and they concern subspecialists, not midwives.

Regionalization is the operational form of specialist access, and here the evidence is mixed in an instructive way. In Massachusetts, women with high-risk conditions who delivered at a hospital without the resources their condition required had an adjusted odds ratio of 3.34 (2.24–4.96) for severe maternal morbidity (64), and in a national sample the excess SMM risk of high-comorbidity women was 9.55-fold at low-acuity hospitals versus 6.50-fold at high-acuity ones (65). But hospital level as such is not protective: in four states, SMM rose with level of care and level I hospitals had marginally lower adjusted risk than level IV (66), and Georgia found no difference in maternal outcomes by service level among high-risk women (67). Risk-appropriate matching helps; a level designation on the door does not. Rural access to the right level is poor: only 27.5% of higher-risk rural women who needed level IV care delivered at such a hospital (68). Twenty-four-hour in-house obstetric coverage has no mortality data, and its SMM signal is one unadjusted survey (69,70). Cardio-obstetric teams reduce 30-day readmission (RR 0.29) but the pooled maternal mortality estimate is null on six small studies with approximated controls (71). MFM telehealth in Arkansas raised consultation rates without outcome data (72).

The most direct evidence that specialist involvement matters for death comes from the confidential enquiries HMHBA says it admires. MBRRACE-UK judged that improvements in care may have made a difference to the outcome in 45% of deaths in 2020–2022 and 2021–2023 and 61% in 2022–2024; in the cardiovascular chapter the figure was 38%, with named failures such as “she was not seen by a cardiologist until 32 weeks’ gestation” and a woman with severe postpartum hypertension who “was not reviewed by a senior doctor prior to her discharge” (73,74). In France, care was suboptimal in 66% of maternal deaths in 2016–2018 and 59.7% were preventable; for cardiovascular deaths the enquiry’s recommendations are multidisciplinary follow-up, repeated risk assessment and “early referral to an expert centre (expert cardiologists, obstetricians, anaesthetists and intensive care)” (75,76). These are expert judgments, not effect estimates, but they come from midwife-led systems, and what they keep finding is not too few midwives. It is delayed senior and specialist review.

For Black women the point is sharper. Their leading causes are cardiac disease, cardiomyopathy, embolism and hypertension (33,41); a substitution model that reduces physician contact in pregnancy and the postpartum year moves in the wrong direction for exactly those conditions.

Nothing in the HMHBA documents proposes to get more Black women to hospitals in the lowest SMM tertile, which is where roughly half of the SMM disparity lives (38), or to a cardiologist before 32 weeks. A plan that names cardiovascular disease as the leading cause and then omits cardiologists has a gap at its center.

Would universal coverage make a big difference?

Coverage is the strongest single policy lever in the evidence base, and it is also the one whose limits are best documented. The two facts need to be held together.

On the first, the only causal-design estimate of a mortality effect is Eliason’s difference-in-differences across 612 state-years, which associated Medicaid expansion with 7.01 fewer maternal deaths per 100,000 live births overall and 16.27 fewer among Black women (53). That is a large effect, about a third of a non-expansion-state rate, and it has not been replicated; it uses death-certificate data across the checkbox transition and the Black subgroup estimate has visible outliers. Delivery-hospitalization SMM studies with stronger designs are null or small: a synthetic-control analysis of 12 million deliveries in 26 states found a 0.02 percentage point change in SMM after expansion (77), a Health Economics event study found “little evidence” of morbidity change (78), and New York’s 2014 expansion produced a 0.42 percentage point relative improvement in the SMM trend for low-income women (79). The coverage-to-utilization chain is robust: expansion reduced early postpartum hospitalizations by 17% (80), the pandemic continuous-enrollment rule raised 12-month postpartum coverage from 59% to 91% (81), and Colorado’s postpartum extension increased treatment of perinatal mood disorders by 20.5 percentage points (82). No study has yet linked the 12-month extension to mortality; the review-committee lag makes that impossible before about 2027 (54).

The reason coverage would help is timing. Thirty percent of pregnancy-related deaths occur 43 to 365 days after delivery, after the historical 60-day Medicaid cutoff, and another 23% between 7 and 42 days (33). In 2015–2018, 26.8% of women with prenatal Medicaid were uninsured before pregnancy and 21.9% were uninsured two to six months after it (83); continuous perinatal coverage was 75.3% among White women and 55.4% among Black women (84). Medicaid pays for roughly two thirds of Black births. Those are the women, and the months, in which cardiomyopathy, postpartum hypertension, overdose and suicide kill. Much of that counterfactual, however, is already spent: 40 states and the District of Columbia have expanded Medicaid and nearly every state now covers twelve months postpartum. The marginal step to universal coverage mainly reaches the roughly eight million uninsured women of reproductive age concentrated in the Southern non-expansion states, which are also the states with the highest Black pregnancy-related mortality.

On the second fact, every universally insured system that has been examined still carries a large Black or African-origin excess, and the table below is the cleanest natural experiment available on what coverage alone does.

Two US findings deserve emphasis. In the TRICARE system, where every woman has the same coverage and most deliver in the same military hospitals, the Black excess in non-transfusion SMM is 68%; the one outcome where the military system appears to have equalized risk is preeclampsia-related SMM, plausibly because military hospitals apply hypertension protocols uniformly (85,89). And in the ACA dependent-coverage study, the reduction in SMM accrued to non-Hispanic White women (aOR 0.89) and not to others, a caution that coverage gains do not flow automatically to the women who need them most (90).

A defensible estimate is that universal coverage would remove on the order of 10% to 25% of the US rate, perhaps two to four deaths per 100,000, concentrated in late-postpartum, cardiovascular, mental-health and chronic-disease deaths in the South, and would narrow the absolute Black-White gap by more than that because Black women are disproportionately uninsured and in non-expansion states. It would not bring the United States to European rates, and the relative Black-White ratio would most likely remain in the 1.5 to 3 range seen in every universal system, because the remainder of the gap lives in which hospital a woman delivers in, what happens to her inside it, her chronic disease burden, and factors that operate identically under the NHS and under TRICARE. Universal coverage is necessary; the countries that have it and still run 12 to 13 per 100,000 show that it is not sufficient. HMHBA’s plan, to its credit, does address coverage: it treats Medicaid as the financing vehicle for 40% of births and supports the postpartum extension (3). It does not address what the coverage buys, which is the specialist and hospital-quality problem above.

Assessment

HMHBA has chosen a target, halving maternal mortality in five years, that is measured by a statistic whose level is disputed by a factor of two and whose year-to-year changes since 2022 have not been statistically significant. It has chosen mechanisms, protecting rural units and expanding midwives and doulas, that improve access and the experience of birth but have never been shown to reduce maternal death in a high-income country, and it has justified them with numbers that do not transfer (a low-income-country model for midwifery, cross-sectional desert associations for closures, an insurer’s propensity-matched claims study for doulas). The comparison with Europe that motivates the program overstates the gap by comparing an inflated numerator with under-counted ones.

The omission that matters most is the one raised by the two new sections. The plan names cardiovascular disease as the leading cause of death and then proposes a workforce with fewer physicians in it; it never mentions maternal-fetal medicine, cardiology, anesthesia, hospital quality or levels of care, which are the only workforce and system variables with any mortality signal in the US literature and the recurring failures named by the European enquiries it cites as models. Universal coverage, which the plan supports through Medicaid, is the single policy with the strongest claim to a mortality effect, but the natural experiments of TRICARE, Kaiser, the NHS and France show that it narrows the Black excess without removing it. None of this means the money will be wasted. The evidence points clearly at where it would do the most: paying every delivering hospital to run hemorrhage and hypertension bundles with race-stratified outcome reporting, funding remote postpartum blood pressure programs and cardio-obstetric follow-up, supporting states to keep women insured for a year, building the enhanced surveillance that HMHBA says it wants (and accepting that the measured rate may rise when the counting improves), and funding rural transfer and readiness systems rather than propping up units delivering fewer than 100 births a year. Doulas and midwives belong in that system for the outcomes they have been shown to change. If the program’s own scorecard is the NCHS rate, it should say now which baseline it is using and how it will separate real change from measurement change; otherwise the result in 2031 will be unfalsifiable, and both success and failure will be claimed.

References

1. Hoyert DL. Maternal mortality rates in the United States, 2024. NCHS Health E-Stats. 2026 Mar;(113):1–7.

2. Healthy Moms Healthy Babies America. Olivia and Tom Walton commit an initial $100 million to cut U.S. maternal deaths in half [press release]. GlobeNewswire; 2026 Sep 10.

3. Healthy Moms Healthy Babies America. Maternal health call to action. Bentonville (AR): Heartland Forward; 2026. Available from: hmhba.org/maternal-health-call-to-action.pdf

4. Kozhimannil KB, Hung P, Henning-Smith C, Casey MM, Prasad S. Association between loss of hospital-based obstetric services and birth outcomes in rural counties in the United States. JAMA. 2018;319(12):1239–47. doi:10.1001/jama.2018.1830

5. Kozhimannil KB, Interrante JD, Carroll C, Sheffield EC, Fritz AH, McGregor AJ, et al. Obstetric care access declined in rural and urban hospitals across US states, 2010–22. Health Aff (Millwood). 2025;44(7):806–11. doi:10.1377/hlthaff.2024.01552

6. Stoneburner A, Chestnut JF, Lucas R, Jones EE, DeMaria AL. Nowhere to go: maternity care deserts across the U.S. Report No. 5. Arlington (VA): March of Dimes; 2026.

7. Fischer S, Royer H, White C. Health care centralization: the health impacts of obstetric unit closures in the United States. Am Econ J Appl Econ. 2024;16(3):113–41. doi:10.1257/app.20220341

8. Chatterji P, Ho CY, Wu X. Obstetric unit closures and racial/ethnic disparity in health. NBER Working Paper 30986. Cambridge (MA): National Bureau of Economic Research; 2023. [Journal publication not verified]

9. Battaglia E. The effect of hospital maternity ward closures on maternal and infant health. Am J Health Econ. 2025;11(2):201–46. doi:10.1086/727738

10. McGregor AJ, Hung P, Garman D, Amutah-Onukagha N, Cooper JA. Obstetrical unit closures and racial and ethnic differences in severe maternal morbidity in the state of New Jersey. Am J Obstet Gynecol MFM. 2021;3(6):100480. doi:10.1016/j.ajogmf.2021.100480

11. Minion SC, Krans EE, Brooks MM, Mendez DD, Haggerty CL. Association of driving distance to maternity hospitals and maternal and perinatal outcomes. Obstet Gynecol. 2022;140(5):812–9. doi:10.1097/AOG.0000000000004960

12. Atwani R, Robbins L, Saade G, Kawakita T. Association of maternity care deserts with maternal and pregnancy-related mortality. Obstet Gynecol. 2025;146(2):181–8. doi:10.1097/AOG.0000000000005976

13. Wallace M, Dyer L, Felker-Kantor E, Benno J, Vilda D, Harville E, et al. Maternity care deserts and pregnancy-associated mortality in Louisiana. Womens Health Issues. 2021;31(2):122–9. doi:10.1016/j.whi.2020.09.004

14. Kozhimannil KB, Leonard SA, Handley SC, Passarella M, Main EK, Lorch SA, et al. Obstetric volume and severe maternal morbidity among low-risk and higher-risk patients giving birth at rural and urban US hospitals. JAMA Health Forum. 2023;4(6):e232110. doi:10.1001/jamahealthforum.2023.2110

15. Sandall J, Fernandez Turienzo C, Devane D, Soltani H, Gillespie P, Gates S, et al. Midwife continuity of care models versus other models of care for childbearing women. Cochrane Database Syst Rev. 2024;4(4):CD004667. doi:10.1002/14651858.CD004667.pub6

16. Bohren MA, Hofmeyr GJ, Sakala C, Fukuzawa RK, Cuthbert A. Continuous support for women during childbirth. Cochrane Database Syst Rev. 2017;7(7):CD003766. doi:10.1002/14651858.CD003766.pub6

17. Kozhimannil KB, Hardeman RR, Attanasio LB, Blauer-Peterson C, O’Brien M. Doula care, birth outcomes, and costs among Medicaid beneficiaries. Am J Public Health. 2013;103(4):e113–21. doi:10.2105/AJPH.2012.301201

18. Falconi AM, Bromfield SG, Tang T, Malloy D, Blanco D, Disciglio S, et al. Doula care across the maternity care continuum and impact on maternal health: evaluation of doula programs across three states using propensity score matching. EClinicalMedicine. 2022;50:101531. doi:10.1016/j.eclinm.2022.101531

19. Falconi AM, Ramirez L, Cobb R, Levin C, Nguyen M, Inglis T. Role of doulas in improving maternal health and health equity among Medicaid enrollees, 2014–2023. Am J Public Health. 2024;114(11):1275–85. doi:10.2105/AJPH.2024.307805 [author list verified via Europe PMC; confirm on PubMed]

20. Nove A, Friberg IK, de Bernis L, McConville F, Moran AC, Najjemba M, et al. Potential impact of midwives in preventing and reducing maternal and neonatal mortality and stillbirths: a Lives Saved Tool modelling study. Lancet Glob Health. 2021;9(1):e24–32. doi:10.1016/S2214-109X(20)30397-1

21. Vedam S, Stoll K, MacDorman M, Declercq E, Cramer R, Cheyney M, et al. Mapping integration of midwives across the United States: impact on access, equity, and outcomes. PLoS One. 2018;13(2):e0192523. doi:10.1371/journal.pone.0192523

22. Grünebaum A, McCullough LB, Orosz B, Chervenak FA. Neonatal mortality in the United States is related to location of birth (hospital versus home) rather than the type of birth attendant. Am J Obstet Gynecol. 2020;223(2):254.e1–8. doi:10.1016/j.ajog.2020.01.045

23. Gunja MZ, Gumas ED, Masitha R, Zephyrin LC. Insights into the U.S. maternal mortality crisis: an international comparison. New York: Commonwealth Fund; 2024 Jun 4.

24. Diguisto C, Saucedo M, Kallianidis A, Bloemenkamp K, Bødker B, Buoncristiano M, et al. Maternal mortality in eight European countries with enhanced surveillance systems: descriptive population based study. BMJ. 2022;379:e070621. doi:10.1136/bmj-2022-070621

25. MBRRACE-UK. Data brief: maternal mortality 2022–24. Oxford: National Perinatal Epidemiology Unit; 2026 Sep. [Author list not verified]

26. Kallianidis AF, Schutte JM, Schuringa LEM, Beenakkers ICM, Bloemenkamp KWM, Braams-Lisman BAM, et al. Confidential enquiry into maternal deaths in the Netherlands, 2006–2018. Acta Obstet Gynecol Scand. 2022;101(4):441–9. doi:10.1111/aogs.14312

27. Callaghan J, Dudenhausen J, Paulson L, Hellmeyer L, Vetter K, Ziegert M, et al. Analysis of maternal mortality in Berlin, Germany – discrepancy between reported maternal mortality and comprehensive death certificate exploration. J Perinat Med. 2024;52(4):375–84. doi:10.1515/jpm-2023-0403

28. Hoyert DL, Miniño AM. Maternal mortality in the United States: changes in coding, publication, and data release, 2018. Natl Vital Stat Rep. 2020;69(2):1–18.

29. Rossen LM, Womack LS, Hoyert DL, Anderson RN, Uddin SFG. The impact of the pregnancy checkbox and misclassification on maternal mortality trends in the United States, 1999–2017. Vital Health Stat 3. 2020;(44). [Page range not verified]

30. Joseph KS, Lisonkova S, Boutin A, Muraca GM, Razaz N, John S, et al. Maternal mortality in the United States: are the high and rising rates due to changes in obstetrical factors, maternal medical conditions, or maternal mortality surveillance? Am J Obstet Gynecol. 2024;230(4):440.e1–13. doi:10.1016/j.ajog.2023.12.038

31. Declercq E, Thoma M. Measuring US maternal mortality. JAMA. 2023;330(18):1731–2. doi:10.1001/jama.2023.19945

32. Janevic T, Declercq E, Howell EA. Data have consequences—centring equity in the maternal mortality surveillance debate. Paediatr Perinat Epidemiol. 2025;39(4):311–3. doi:10.1111/ppe.13138

33. Trost SL, Beauregard J, Njie F, Berry J, Harvey A, Goodman DA, et al. Pregnancy-related deaths: data from maternal mortality review committees in 36 US states, 2017–2019. Atlanta (GA): Centers for Disease Control and Prevention; 2022. [Author list beyond first three not verified]

34. Wallace M, Gillispie-Bell V, Cruz K, Davis K, Vilda D. Homicide during pregnancy and the postpartum period in the United States, 2018–2019. Obstet Gynecol. 2021;138(5):762–9. doi:10.1097/AOG.0000000000004567

35. Bruzelius E, Martins SS. US trends in drug overdose mortality among pregnant and postpartum persons, 2017–2020. JAMA. 2022;328(21):2159–61. doi:10.1001/jama.2022.17045 [issue/pages not verified]

36. Drew LB, Goldman-Mellor S, Joachim G, Gemmill A, Margerison CE. Pregnancy-associated drug overdose mortality in the United States, 2018–2023. Obstet Gynecol. 2026 Jul 30 [Epub ahead of print]. doi:10.1097/AOG.0000000000006389

37. Petersen EE, Davis NL, Goodman D, Cox S, Syverson C, Seed K, et al. Racial/ethnic disparities in pregnancy-related deaths — United States, 2007–2016. MMWR Morb Mortal Wkly Rep. 2019;68(35):762–5. doi:10.15585/mmwr.mm6835a3

38. Howell EA, Egorova NN, Balbierz A, Zeitlin J, Hebert PL. Site of delivery contribution to black-white severe maternal morbidity disparity. Am J Obstet Gynecol. 2016;215(2):143–52. doi:10.1016/j.ajog.2016.05.007

39. Howell EA, Egorova NN, Janevic T, Brodman M, Balbierz A, Zeitlin J, et al. Race and ethnicity, medical insurance, and within-hospital severe maternal morbidity disparities. Obstet Gynecol. 2020;135(2):285–93. doi:10.1097/AOG.0000000000003667

40. Guglielminotti J, Wong CA, Friedman AM, Li G. Racial and ethnic disparities in death associated with severe maternal morbidity in the United States: failure to rescue. Obstet Gynecol. 2021;137(5):791–800. doi:10.1097/AOG.0000000000004362

41. Briller J, Trost SL, Busacker A, Joseph NT, Davis NL, Petersen EE, et al. Pregnancy-related mortality due to cardiovascular conditions: maternal mortality review committees in 32 U.S. states, 2017 to 2019. JACC Adv. 2024;3(12):101382. doi:10.1016/j.jacadv.2024.101382

42. Main EK, Cape V, Abreo A, Vasher J, Woods A, Carpenter A, et al. Reduction of severe maternal morbidity from hemorrhage using a state perinatal quality collaborative. Am J Obstet Gynecol. 2017;216(3):298.e1–11. doi:10.1016/j.ajog.2017.01.017

43. Main EK, Chang SC, Dhurjati R, Cape V, Profit J, Gould JB. Reduction in racial disparities in severe maternal morbidity from hemorrhage in a large-scale quality improvement collaborative. Am J Obstet Gynecol. 2020;223(1):123.e1–14. PMID 31978432. [DOI not captured]

44. Main EK, Markow C, Gould J. Addressing maternal mortality and morbidity in California through public-private partnerships. Health Aff (Millwood). 2018;37(9):1484–93. doi:10.1377/hlthaff.2018.0463

45. California Department of Public Health. California Pregnancy Mortality Surveillance System: pregnancy-related deaths, 2019–2021 [fact sheet]. Sacramento (CA): CDPH; 2025. See also: California Pregnancy-Associated Mortality Review, pregnancy-related deaths 2008–2016.

46. Gillispie-Bell V, Wallace K. Improving and maintaining reductions in severe maternal morbidity through a statewide perinatal quality collaborative. Obstet Gynecol. 2026 [Epub ahead of print]. doi:10.1097/AOG.0000000000006269. PMID 41886758.

47. Borders A, Keenan-Devlin L, Oh EH, Young D, Grobman W, Lee King P. Reducing severe maternal morbidity for birthing persons with severe hypertension through a statewide quality improvement initiative. Am J Obstet Gynecol. 2024 Apr 30 [Epub ahead of print]. PMID 38697335. [Volume/pages not verified]

48. Davidson C, Denning S, Thorp K, Tyer-Viola L, Belfort M, Sangi-Haghpeykar H, et al. Examining the effect of quality improvement initiatives on decreasing racial disparities in maternal morbidity. BMJ Qual Saf. 2022;31(9):670–8. doi:10.1136/bmjqs-2021-014225

49. Kern-Goldberger AR, Hirshberg A, James A, Levine LD, Howell E, Harbuck E, et al. Trends in severe maternal morbidity following an institutional team goal strategy for disparity reduction. Am J Obstet Gynecol MFM. 2024:101529. doi:10.1016/j.ajogmf.2024.101529

50. Tita AT, Szychowski JM, Boggess K, Dugoff L, Sibai B, Lawrence K, et al. Treatment for mild chronic hypertension during pregnancy. N Engl J Med. 2022;386(19):1781–92. doi:10.1056/NEJMoa2201295

51. Arkerson BJ, Finneran MM, Harris SR, Schnorr J, McElwee ER, Demosthenes L, et al. Remote monitoring compared with in-office surveillance of blood pressure in patients with pregnancy-related hypertension: a randomized controlled trial. Obstet Gynecol. 2023;142(4):855–61. doi:10.1097/AOG.0000000000005327 [title as retrieved; confirm]

52. Hirshberg A, Zhu Y, Smith-McLallen A, Srinivas SK. Association of a remote blood pressure monitoring program with postpartum adverse outcomes. Obstet Gynecol. 2023;141(6):1163–70. doi:10.1097/AOG.0000000000005197

53. Eliason EL. Adoption of Medicaid expansion is associated with lower maternal mortality. Womens Health Issues. 2020;30(3):147–52. doi:10.1016/j.whi.2020.01.005

54. Eliason EL, Steenland MW, Gourevitch RA. Extended pregnancy Medicaid during COVID-19 and enrollment and health care use in the postpartum year. Milbank Q. 2026;104(2):488–507. doi:10.1111/1468-0009.70079

55. Greenwood BN, Hardeman RR, Huang L, Sojourner A. Physician–patient racial concordance and disparities in birthing mortality for newborns. Proc Natl Acad Sci U S A. 2020;117(35):21194–200. doi:10.1073/pnas.1913405117

56. Borjas GJ, VerBruggen R. Physician–patient racial concordance and newborn mortality. Proc Natl Acad Sci U S A. 2024;121(39):e2409264121. doi:10.1073/pnas.2409264121

57. Nantwi AK, Janevic T. Intervention research to reduce disparities in severe maternal morbidity and mortality in the United States: a scoping review. Obstet Gynecol. 2026;147(6):817. doi:10.1097/AOG.0000000000006305 [page range not verified]

58. Gray A, Thomas Z, Robinson L, Ashley-McNamee RR, Cornelius A, Louis ER, et al. Midwifery care and doula support for Black women in the United States: a systematic review. J Racial Ethn Health Disparities. 2026 [Epub ahead of print]. doi:10.1007/s40615-026-03121-9

59. McConnell MA, Rokicki S, Ayers S, Allouch F, Perreault N, Gourevitch RA, et al. Effect of an intensive nurse home visiting program on adverse birth outcomes in a Medicaid-eligible population: a randomized clinical trial. JAMA. 2022;328(1):27–37. doi:10.1001/jama.2022.9703

60. Qian J, Wolfson C, Kramer B, Creanga AA. Insights from preventability assessments across 42 state and city maternal mortality reviews in the United States. Am J Obstet Gynecol. 2025;232(4):394.e1–10. doi:10.1016/j.ajog.2024.08.030

61. Sullivan SA, Hill EG, Newman RB, Menard MK. Maternal-fetal medicine specialist density is inversely associated with maternal mortality ratios. Am J Obstet Gynecol. 2005;193(3 Pt 2):1083–8. doi:10.1016/j.ajog.2005.05.085

62. Kawakita T, Atwani R, Hayasaka M, Robbins L, Saade G. The association between maternal-fetal medicine physician density and pregnancy outcomes. Am J Perinatol. 2026;43(9):1121–6. doi:10.1055/a-2717-3951

63. Snyder JE, Stahl AL, Streeter RA, Washko MM. Regional variations in maternal mortality and health workforce availability in the United States. Ann Intern Med. 2020;173(11 Suppl):S45–54. doi:10.7326/M19-3254

64. Osei-Poku GK, Prentice JC, Easter SR, Diop H. Delivery at an inadequate level of maternal care is associated with severe maternal morbidity. Am J Obstet Gynecol. 2024;231(5):546.e1–20. doi:10.1016/j.ajog.2024.02.308

65. Clapp MA, James KE, Kaimal AJ. The effect of hospital acuity on severe maternal morbidity in high-risk patients. Am J Obstet Gynecol. 2018;219(1):111.e1–7. doi:10.1016/j.ajog.2018.04.015

66. Handley SC, Passarella M, Radack J, Amuzie O, Hsu JY, Srinivas SK, et al. Maternal levels of care and association with severe maternal morbidity during birth hospitalizations. PLoS One. 2026;21(7):e0353016. doi:10.1371/journal.pone.0353016

67. Vanderlaan J, Rochat R, Williams B, Dunlop A, Shapiro SE. Associations between hospital maternal service level and delivery outcomes. Womens Health Issues. 2019;29(3):252–8. doi:10.1016/j.whi.2019.02.004

68. Handley SC, Formanowski B, Passarella M, Thorsen ML, Interrante JD, Busse CE, et al. Risk-appropriate childbirth care among higher-risk pregnant rural residents. JAMA Health Forum. 2025;6(11):e254241. PMID 41269701. [DOI not captured]

69. Srinivas SK, Small DS, Macheras M, Hsu JY, Caldwell D, Lorch S. Evaluating the impact of the laborist model of obstetric care on maternal and neonatal outcomes. Am J Obstet Gynecol. 2016;215(6):770.e1–9. doi:10.1016/j.ajog.2016.08.007

70. Torbenson VE, Tatsis V, Bradley SL, Butler J, Kjerulff L, McLaughlin GB, et al. Use of obstetric and gynecologic hospitalists is associated with decreased severe maternal morbidity in the United States. J Patient Saf. 2023;19(3):202–10. doi:10.1097/PTS.0000000000001102

71. Sebastian SA, Atwal H, Ayyalu T, Gulati M. Impact of cardio-obstetrics care on maternal outcomes in pregnant women with heart disease: a systematic review and meta-analysis. Curr Probl Cardiol. 2025;50(12):103190. doi:10.1016/j.cpcardiol.2025.103190

72. Magann EF, Bronstein J, McKelvey SS, Wendel P, Smith DM, Lowery CL. Evolving trends in maternal fetal medicine referrals in a rural state using telemedicine. Arch Gynecol Obstet. 2012;286(6):1383–92. PMID 22821508.

73. MBRRACE-UK. Saving lives, improving mothers’ care: state of the nation report 2025 — lessons learned to inform maternity care from the UK and Ireland confidential enquiries into maternal deaths and morbidity 2021–23. Oxford: National Perinatal Epidemiology Unit; 2025. [Editor list not verified]

74. MBRRACE-UK. Saving lives, improving mothers’ care: state of the nation report 2026 — surveillance of maternal deaths 2022–24. Oxford: National Perinatal Epidemiology Unit; 2026. [Editor list not verified]

75. Deneux-Tharaux C, Saucedo M; Comité National d’Experts sur la Mortalité Maternelle. Mortalité maternelle, adéquation des soins et évitabilité, France 2016–2018. Gynecol Obstet Fertil Senol. 2024;52(4):201–9. doi:10.1016/j.gofs.2024.02.021

76. Bruyère M, Morau E, Verspyck E; Comité National d’Experts sur la Mortalité Maternelle. Mortalité maternelle par maladies cardiovasculaires en France 2016–2018. Gynecol Obstet Fertil Senol. 2024;52(4):221–30. PMID 38373486.

77. Guglielminotti J, Daw JR, Friedman AM, Samari G, Li G. Medicaid expansion and severe maternal morbidity during delivery hospitalizations. Obstet Gynecol. 2026 Aug 27 [Epub ahead of print]. doi:10.1097/AOG.0000000000006415

78. Chatterji P, Glenn H, Markowitz S, Montez JK. Affordable Care Act Medicaid expansions and maternal morbidity. Health Econ. 2023;32(10):2334–52. doi:10.1002/hec.4724

79. Guglielminotti J, Landau R, Li G. The 2014 New York State Medicaid expansion and severe maternal morbidity during delivery hospitalizations. Anesth Analg. 2021;133(2):340–8. doi:10.1213/ANE.0000000000005371

80. Steenland MW, Wherry LR. Medicaid expansion led to reductions in postpartum hospitalizations. Health Aff (Millwood). 2023;42(1). doi:10.1377/hlthaff.2022.00819 [pages not verified]

81. Gordon SH, Chen L, DeLew N, Sommers BD. COVID-19 Medicaid continuous enrollment provision yielded gains in postpartum continuity of coverage. Health Aff (Millwood). 2024;43(3):336–43. doi:10.1377/hlthaff.2023.00580

82. Gordon SH, Lee S, Steenland MW, Deen N, Feinberg E. Extended postpartum Medicaid in Colorado associated with increased treatment for perinatal mood and anxiety disorders. Health Aff (Millwood). 2024;43(4):523–31. doi:10.1377/hlthaff.2023.01441

83. Johnston EM, McMorrow S, Alvarez Caraveo C, Dubay L. Post-ACA, more than one-third of women with prenatal Medicaid remained uninsured before or after pregnancy. Health Aff (Millwood). 2021;40(4):571–8. doi:10.1377/hlthaff.2020.01678

84. Daw JR, Kolenic GE, Dalton VK, Zivin K, Winkelman TNA, Kozhimannil KB, et al. Racial and ethnic disparities in perinatal insurance coverage. Obstet Gynecol. 2020;135(4):917–24. doi:10.1097/AOG.0000000000003728

85. Fuller S, Kuenstler MR, Snipes MA, Miller MJ, Lutgendorf MA. Obstetrical health care inequities in a universally insured health care system. AJOG Glob Rep. 2023;3(3):100256. doi:10.1016/j.xagr.2023.100256

86. Oakley LP, Li X, Tartof SY, Wilkes-Grundy M, Fassett MJ, Lawrence JM. Racial disparities in severe maternal morbidity in an integrated health care system, Southern California, 2008–2017. Womens Health Issues. 2023;33(3):280–8. doi:10.1016/j.whi.2023.01.001

87. Saucedo M, Deneux-Tharaux C; Comité National d’Experts sur la Mortalité Maternelle. Mortalité maternelle en France, 2016–2018, fréquence, causes et profil des femmes. Gynecol Obstet Fertil Senol. 2024;52(4):185–200. doi:10.1016/j.gofs.2024.02.020

88. Urquia ML, Glazier RH, Mortensen L, Nybo-Andersen AM, Small R, Davey MA, et al. Severe maternal morbidity associated with maternal birthplace in three high-immigration settings. Eur J Public Health. 2015;25(4):620–5. doi:10.1093/eurpub/cku230

89. Lundeberg K, Tindal R, Grob PC, Hamilton JL, Gonzalez-Brown VM, Keyser EA. Military healthcare system mitigates racial disparities for severe maternal morbidity from preeclampsia. AJOG Glob Rep. 2023;3(3):100215. doi:10.1016/j.xagr.2023.100215

90. Guglielminotti J, Li G, Daw JR, Friedman AM, Samari G. Reduced odds of severe maternal morbidity associated with the US Affordable Care Act dependent coverage provision. Am J Obstet Gynecol MFM. 2025;7(6):101668. doi:10.1016/j.ajogmf.2025.101668

The nurse remembers one thing.

The obstetrician remembers another.

The resident has moved across the country.

The anesthesiologist has performed thousands of procedures since then.

The patient remembers fear.

The family remembers urgency.

And sitting between all of them is the one witness that does not age:

the medical record.

The Chart Eventually Has to Speak Without You

Clinicians often think of documentation as something done after care.

That is backwards.

Documentation is part of care.

It communicates with the next nurse.

The covering physician.

The consultant.

The postpartum team.

The pediatrician.

The patient’s future obstetrician.

And, sometimes years later, attorneys and expert witnesses.

A useful record should answer two questions:

What happened?

and

Why did we do what we did?

Too many charts answer only the first.

Numbers Without Reasoning Are Not Enough

Read more

168/112.

The blood pressure is repeated.

171/110.

The nurse calls.

The patient has a headache.

Everyone knows she has severe hypertension.

And yet nothing happens.

This is one of the most disturbing forms of preventable obstetric harm because the problem is not hidden.

It has already been measured.

Severe Hypertension Is Different

Many areas of obstetrics contain uncertainty.

Electronic fetal monitoring is imperfect.

Estimated fetal weight is imprecise.

Labor progression varies enormously.

Even experienced clinicians can reasonably disagree about many management decisions.

Persistent acute-onset severe hypertension is different.

Severe hypertension during pregnancy or postpartum is a medical emergency because of the risk of stroke and other severe maternal complications.

SMFM states that expeditious treatment within 30 to 60 minutes is recommended and has created a quality metric specifically to measure whether hospitals achieve timely treatment. [1]

That makes severe hypertension an unusually revealing patient-safety problem.

The threshold is recognized.

The treatment exists.

The urgency is known.

Yet delays still occur.

The Number Is Not the Treatment

Read more

A fetal tracing deteriorates during an oxytocin infusion.

The infusion continues.

Contractions become more frequent.

Someone notices.

Someone else assumes the obstetrician has already been called.

The dose is eventually reduced.

Then comes prolonged bradycardia, an emergency cesarean, and an injured newborn.

Years later, the malpractice case is described in five words:

“They gave too much Pitocin.”

Subscribe now

But that description may miss the more important problem.

The drug did not fail.

The system around the drug may have failed.

Oxytocin Is Not an Ordinary Medication

Oxytocin is one of the most familiar medications in obstetrics. Familiarity can make us forget how consequential it is.

The Institute for Safe Medication Practices includes intravenous oxytocin on its list of high-alert medications in acute care settings. “High alert” does not mean the drug should rarely be used. It means that errors involving it can cause substantial harm and therefore deserve safeguards.

Oxytocin changes uterine activity.

More contractions are not always better contractions.

Excessive uterine activity can reduce the interval during which placental blood flow recovers between contractions. In a fetus with limited reserve, that may matter greatly.

A recent analysis embedded within the randomized STOPOXY trial found that stopping oxytocin in active labor was followed by fewer contractions and fewer fetal heart rate decelerations in the following hour. Importantly, this physiologic finding does not prove that discontinuing oxytocin is always superior or that continuation represents negligence. It demonstrates something simpler: oxytocin exposure can materially affect uterine activity and fetal heart rate patterns.

That is why its administration requires active management.

Read more

A baby is born with severe neurologic injury.

Months or years later, lawyers, physicians, nurses, experts, and jurors gather around strips of paper from the labor room.

The fetal heart rate tracing is reconstructed minute by minute.

A deceleration at 1:42 PM.

Reduced variability at 2:07.

Another deceleration at 2:31.

By 3:15, someone argues, the baby should have been delivered.

Looking backward, the story can appear remarkably clear.

But there is a problem.

The clinicians caring for the patient at 2:07 PM did not know the outcome.

We do.

That difference changes everything.

The Monitor Is Not a Crystal Ball

Electronic fetal monitoring is one of the most ubiquitous technologies in modern obstetrics.

Its basic physiologic premise is reasonable.

Changes in fetal oxygenation can alter fetal heart rate patterns. Monitoring those patterns can therefore provide information about fetal status during labor.

But somewhere along the way, a physiologic surveillance tool acquired a much larger cultural role.

It began to be treated as though it could reliably predict which fetus would suffer permanent neurologic injury.

The evidence does not support that degree of certainty.

In one of the landmark studies, Nelson and colleagues examined fetal monitoring among children with moderate or severe cerebral palsy and controls.

Certain fetal heart rate abnormalities were associated with increased risk.

But association is not prediction.

That distinction is fundamental.

A finding can be associated with an outcome and still perform poorly as a clinical predictor when the outcome itself is uncommon.

What Randomized Trials Tell Us

The Cochrane review of continuous cardiotocography included 13 trials involving more than 37,000 women.

Read more

A postpartum patient calls with a severe headache.

Her blood pressure is elevated, but not yet catastrophic. She has not slept. She had an epidural. She is exhausted.

“It’s probably the epidural.”

Another patient is short of breath.

“She just delivered.”

Another has persistent tachycardia.

“She is anxious.”

Another says repeatedly that something is wrong.

“Everything looks reassuring.”

Most of the time, benign explanations really are benign.

The malpractice problem begins when reassurance becomes a substitute for reassessment.

Normal Pregnancy Creates an Abnormal Diagnostic Environment

Obstetrics is unusual because pregnancy itself changes physiology.

Heart rate rises.

Blood volume increases.

Hemoglobin falls.

Edema is common.

Dyspnea can occur.

Headaches are frequent.

Pain, fatigue, nausea, contractions, bleeding, and dizziness may all be part of otherwise uncomplicated pregnancy, labor, or postpartum recovery.

That makes obstetrics particularly vulnerable to normalization.

A symptom that would trigger concern in a nonpregnant patient can easily be attributed to pregnancy.

Sometimes appropriately.

Sometimes disastrously.

Maternal death reviews helped expose this problem years ago. The National Partnership for Maternal Safety developed maternal early warning criteria after case reviews showed repeated delays in recognizing hemorrhage, hypertensive crisis, sepsis, venous thromboembolism, and heart failure. [1]

The problem was not always that nobody measured the vital signs.

Often, the abnormal information was already there.

It had simply not produced the right response.

The Patient Who “Looks Fine”

Imagine a postpartum woman whose heart rate is 124.

She is pale but talking comfortably.

Her blood pressure remains normal.

Her vaginal bleeding does not look dramatic.

It is easy to be reassured.

Now suppose her heart rate is 130 an hour later.

Her abdomen hurts more.

Her hemoglobin begins to fall.

At what point does “probably normal postpartum physiology” become “possible concealed hemorrhage”?

That transition is one of the hardest tasks in clinical medicine.

There is rarely a bell that rings.

Instead, information accumulates.

The malpractice risk is not that clinicians fail to predict the future.

It is that they continue using yesterday’s explanation after today’s facts no longer fit it.

Reassurance Must Have an Expiration Date

This is an important clinical principle.

Every reassuring diagnosis should have an expiration date.

“The headache is probably related to the epidural” may be reasonable at one moment.

But what if the headache changes?

What if the blood pressure rises?

What if visual symptoms develop?

What if neurologic findings appear?

The original explanation must then be reopened.

Good clinical reasoning is not merely choosing the most likely diagnosis.

It is recognizing when the evidence no longer supports it.

Diagnostic Error Is Often Not a Knowledge Failure

We tend to imagine diagnostic malpractice as ignorance.

The physician did not know the diagnosis.

But many diagnostic failures are more subtle.

The correct diagnosis may actually have been considered.

The problem may have been that it was considered and then dismissed too early.

Or the abnormal finding was attributed to something less dangerous.

Or several different clinicians each saw one piece of the puzzle, while nobody assembled the whole picture.

AHRQ has specifically examined diagnostic error as a contributor to severe maternal morbidity and mortality, including failures involving recognition, communication, escalation, and timely treatment. [2]

This distinction matters.

More medical knowledge alone will not solve a failure of clinical attention.

“Common” Does Not Mean “Safe”

There is another trap.

If something is common, clinicians naturally become less alarmed by it.

But frequency and danger are different questions.

Headache is common postpartum.

Preeclampsia is dangerous.

Tachycardia is common during labor.

Hemorrhage and sepsis are dangerous.

Shortness of breath may occur during normal pregnancy.

Pulmonary embolism and cardiomyopathy can kill.

The clinician’s job is not to assume that every symptom represents catastrophe.

That would create overtreatment, unnecessary testing, anxiety, cost, and harm.

The obligation is different:

Do not use the frequency of a benign explanation to erase evidence of a dangerous one.

Early Warning Systems Were Designed for Exactly This Problem

Maternal early warning systems attempt to make deterioration harder to normalize.

Instead of relying entirely on whether someone “looks sick,” certain physiologic abnormalities trigger evaluation.

That is valuable because humans adapt.

The nurse has been watching the patient’s heart rate rise slowly for four hours.

The physician has heard about several elevated pressures.

Everyone becomes accustomed to the abnormality.

A trigger system interrupts that adaptation.

But early warning systems are not magical. Studies show that their sensitivity varies depending on the condition being detected. [3]

The purpose is not to replace clinical judgment.

It is to force clinical judgment to occur.

That is a very different function.

Listen to the Patient

There is also a human source of early warning that cannot be captured completely in a flowsheet.

The patient.

“I don’t feel right.”

“This headache is different.”

“I can’t breathe lying down.”

“I have never felt pain like this.”

“My heart is racing.”

“I think something is wrong.”

These statements are not diagnoses.

But neither are they noise.

The CDC’s maternal safety work specifically emphasizes listening to pregnant and postpartum patients and acting on urgent maternal warning signs. More than 80% of pregnancy-related deaths reviewed through recent maternal mortality review processes were considered preventable, meaning there was at least some chance of avoiding the death through reasonable changes at one or more levels. [4]

That does not mean 80% involved malpractice.

Preventability and negligence are not synonymous.

That distinction is essential.

But it tells us that opportunities for earlier recognition and intervention are real.

The Medicolegal Problem With “Reassuring”

Medical records frequently contain phrases such as:

“Patient reassured.”

“Exam reassuring.”

“Symptoms likely physiologic.”

There is nothing wrong with those conclusions when clinically justified.

The danger comes when the chart does not show what made the situation reassuring.

What was considered?

What dangerous diagnoses were reasonably excluded?

What findings would require reevaluation?

What return precautions were given?

When will the patient be reassessed?

Reassurance should be the conclusion of reasoning, not a replacement for reasoning.

Preventive Ethics Means Planning for Deterioration

Preventive ethics asks clinicians to anticipate foreseeable problems before they become emergencies.

In this context, that means giving patients clear instructions during pregnancy and before discharge:

Which headaches require attention?

How much bleeding is too much?

When is shortness of breath abnormal?

Which blood pressures require action?

When should a patient call the office, go to triage, or call emergency services?

The best time to teach these warning signs is not while the patient is frightened at 2 AM.

It is before she needs them.

The MedMal Room Question

Bad outcomes do not prove that earlier reassurance was negligent.

Medicine requires judgment under uncertainty.

But every reassuring diagnosis creates a responsibility:

What evidence would make me change my mind?

If there is no answer to that question, reassurance has become something more dangerous.

It has become closure.

And in obstetrics, closure can occur long before the disease is finished evolving.

References

1. Mhyre JM, D’Oria R, Hameed AB, et al. The maternal early warning criteria: a proposal from the National Partnership for Maternal Safety. Obstet Gynecol. 2014;124(4):782-786. doi:10.1097/AOG.0000000000000480.

2. Agency for Healthcare Research and Quality. The contribution of diagnostic error to maternal mortality and severe maternal morbidity. Rockville, MD: AHRQ; 2021.

3. Hannola K, Hoppu S, Mennander S, et al. Obstetric early warning system to predict maternal morbidity of pre-eclampsia, postpartum hemorrhage and infection after birth in high-risk women: a prospective cohort study. Midwifery. 2021;99:103015. doi:10.1016/j.midw.2021.103015.

4. Centers for Disease Control and Prevention. Preventing Pregnancy-Related Deaths. Atlanta, GA: CDC; 2024.

Thirty minutes.

Few numbers in obstetrics have acquired more authority.

A decision is made for an emergency cesarean.

The clock starts.

If the baby is delivered in 24 minutes, the response looks impressive.

If delivery occurs at 34 minutes and the baby is injured, four minutes can suddenly look like negligence.

But there is a problem with treating 30 minutes as a biological boundary:

A fetus does not know what a clock says.

There is no physiologic switch at 29 minutes and 59 seconds.

So where did the 30-minute rule come from, and what does it actually mean?

A Number Became a Rule

For decades, the ability to perform an emergency cesarean within approximately 30 minutes has been used as a benchmark for obstetric readiness.

That makes intuitive sense.

A hospital providing obstetric care should be capable of mobilizing an operating room, obstetrician, anesthesia team, nursing staff, and neonatal personnel rapidly when a mother or fetus is in immediate danger.

But two very different ideas gradually became blurred.

The first is:

An obstetric unit should be capable of responding rapidly to emergencies.

The second is:

Every fetus requiring urgent cesarean delivery must be delivered within 30 minutes or the care was substandard.

Those statements are not equivalent.

Quality Metric Versus Biological Threshold

Read more

Safety Is Not a Passport Stamp

Home birth supporters continuously mention European studies showing that home births are safe. They fail to mention that US studies show the exact opposite.

The recurring claim that planned home birth must be safe in the United States because it is reported to be safe in parts of Europe, Canada, or Australia rests on a basic misunderstanding of what safety is and how it is achieved.

Safety is not an attribute that can be imported wholesale from one country to another. It is not a philosophical position, a cultural value, or a marker of moral enlightenment. It is the product of systems that are deliberately designed to identify risk early, enforce boundaries consistently, escalate care rapidly, and hold professionals accountable when outcomes are poor.

When this claim is paired with accusations that American obstetricians oppose home birth out of ideology, racism, or financial self-interest, the conversation moves further away from evidence and closer to rhetoric. The problem is not disagreement. The problem is that slogans are being substituted for analysis. Birth outcomes are shaped far more by infrastructure than by intent, and without examining that infrastructure, international comparisons are not merely incomplete, they are misleading.

What “Safe Elsewhere” Actually Means

In countries frequently cited as home birth success stories, home birth is not an alternative lifestyle practice operating at the margins of medicine. It is a narrowly defined clinical pathway embedded within a unified national health system. Midwives are educated under standardized national curricula, licensed by a single regulatory authority, and subject to meaningful discipline.

Eligibility criteria for home birth are strict, explicit, and enforceable. Risk stratification is not a matter of individual preference, and deviation from protocol carries professional consequences. Transfer arrangements are formal, expected, and cooperative, not improvised or adversarial. Transport distances are short, obstetric units are centralized, and emergency response times are predictable.

Perhaps most importantly, outcomes are tracked in national registries. Data are mandatory, comprehensive, and transparent. They are used to evaluate performance, refine eligibility criteria, and remove poorly performing providers from practice. When advocates point to these countries, what they are actually pointing to is not home birth itself, but the presence of a dense web of regulation, integration, and accountability that makes limited-risk community birth possible for a small, carefully selected population.

Why the U.S. Is Not Comparable

The United States does not have this system. Here, “home birth” is not a single intervention but a broad and heterogeneous category encompassing multiple credentials with profoundly different training requirements. Some pathways require extensive clinical education and hospital experience. Others require none. Licensure varies by state and is absent or optional in many jurisdictions. Where licensure exists, enforcement is often weak or inconsistent. Outcome reporting is frequently voluntary, incomplete, or nonexistent, making it impossible to reliably identify patterns of harm.

US midwives frequently miss reporting low and bad Apgar scores trying to hide their bad outcomes.

Transfer protocols may be informal, contested, or actively resisted, and relationships with hospitals are often defined by conflict rather than collaboration. In large parts of the country, transport times are long, obstetric units are sparse, and emergency response is unpredictable.

Treating this fragmented landscape as equivalent to integrated European systems is a category error. When adverse outcomes occur in this setting, they follow recognizable patterns: delayed recognition of fetal compromise, prolonged labor without monitoring, hemorrhage managed without timely escalation, neonatal hypoxia compounded by delayed transfer. These are not rare or inexplicable tragedies. They are foreseeable failures of system design.

The Myth of Market Capture

One of the most persistent accusations leveled at obstetricians is that opposition to home birth is driven by a desire to protect market share or professional dominance. This claim collapses under even minimal scrutiny. Physicians do not financially benefit when patients deliver in hospitals rather than at home. Hospitals lose revenue when births occur outside their walls. If self-interest were the primary motive, silence would be easier and safer. What creates professional risk is speaking plainly in a polarized environment where critique is reframed as control. Physicians who raise safety concerns invite harassment, misrepresentation, and accusations of bad faith. They do so anyway because professional responsibility does not end where controversy begins. To frame safety advocacy as economic self-protection is not only inaccurate, it is a way of avoiding engagement with the substantive issues of training, oversight, and accountability.

Racism, History, and the Limits of Moral Substitution

Invoking historical figures associated with racism in medicine is a powerful rhetorical move, but it does not resolve the contemporary safety question.

Structural racism in medicine is real and has caused immense harm. Acknowledging that history is essential.

But it does not follow that every modern safety concern is racist, nor that relocating care to loosely regulated settings advances equity. In practice, fragmented systems with weak oversight often harm marginalized women the most. Delayed transfers, lack of emergency backup, and absence of accountability do not disappear when care moves out of hospitals. They become harder to see and harder to correct. Romanticizing risk under the banner of autonomy or historical redress does not produce justice. It shifts harm onto those with the least margin for error. Ethical analysis requires confronting uncomfortable tradeoffs, not substituting moral accusations for evidence.

Autonomy, Professional Integrity, and Preventable Harm

There is also a fundamental ethical misunderstanding at the heart of this debate. Physicians are not obligated to validate every autonomous choice.

Autonomy does not mean that all options are equally safe or professionally supportable. It means that patients deserve honest information about risk and clear recommendations grounded in evidence and experience. Advising against a setting that lacks minimum safety standards is not paternalism.

It is professional integrity. The same reasoning applies when physicians warn against unlicensed surgical centers, counterfeit medications, or unregulated medical devices. Drawing boundaries around acceptable practice is part of the ethical obligation of medicine. Silence in the face of predictable harm is not respect for autonomy. It is abdication of responsibility.

What Would Make Comparisons Meaningful

If advocates want international comparisons to carry real weight, the path forward is not louder rhetoric or sharper accusations. It is structural reform. A single, enforceable standard of education. One license with meaningful oversight. Mandatory outcome reporting. Narrow, enforced eligibility criteria. Formal integration with hospitals and clear authority for transfer. Transparent consequences when standards are violated and harm recurs. These are the features that make community birth safer elsewhere. Until they exist here, pointing to other countries is not an argument. It is a distraction. Safety is not declared by ideology or achieved by analogy. It is built, maintained, and enforced. When preventable harm continues to occur, physicians have an obligation to say so, even when doing so is unpopular.

Most physicians do not learn about malpractice from medical school.

We learn about it when a colleague is sued.

When a case appears in the news.

When a hospital risk-management committee reviews a bad outcome.

Or, worst of all, when the certified letter arrives with our own name on it.

That is too late.

I have now assembled more than 30 cases for a new series I am calling The MedMal Room.

Check the recent ones HERE.

They come from obstetrics, gynecology, fertility treatment, breast care, gynecologic oncology, surgery, postpartum care, and even conditions that are not technically “ObGyn” at all.

The purpose is not to frighten physicians.

It is not to teach defensive medicine.

And it is certainly not to imply that every bad outcome represents malpractice.

The purpose is much more useful:

to understand how good clinicians get into trouble, where preventable harm actually begins, and what we can change before the next patient is injured.

Malpractice Rarely Begins Where the Lawsuit Begins

A baby develops cerebral palsy.

A woman hemorrhages.

A postoperative patient becomes septic.

A breast cancer diagnosis is delayed.

An abnormal Pap test is never followed.

A suspicious mole seen during a gynecologic examination is referred to dermatology, but the patient never gets there.

An IVF laboratory transfers the wrong embryo.

These cases appear completely different.

But when you study them closely, the same patterns keep appearing.

The problem was recognized, but nobody acted.

The test was ordered, but nobody followed the result.

The referral was placed, but nobody knew whether the patient went.

A reassuring test result ended clinical thinking too early.

Everyone assumed someone else owned the next step.

The patient signed a consent form but never really understood the decision.

The physician knew something was wrong but failed to escalate.

Or the care was entirely appropriate, yet hindsight later made the bad outcome look preventable.

These are not just legal problems.

They are clinical reasoning problems, communication problems, systems problems, and ethical problems.

Check the recent ones HERE.

Some of the Cases Will Be Uncomfortable

One article asks:

Is a 30-minute emergency cesarean really a standard of care, or did a quality benchmark become a medicolegal rule?

Another asks:

Would an expert interpret the same fetal heart tracing differently if the expert did not know the baby developed cerebral palsy?

Another examines a woman with severe hypertension whose dangerous blood pressures were documented repeatedly.

Everyone knew.

Nobody treated them quickly enough.

Then there is the patient with a palpable breast lump and a “normal” mammogram.

The woman with repeated postmenopausal bleeding whose first reassuring evaluation became permanent reassurance.

The patient whose pelvic CT showed normal ovaries but incidentally found another potentially malignant lesion that nobody followed.

And one of my favorites because it captures the entire series:

A gynecologist sees a suspicious mole and refers the patient to dermatology.

She never goes.

Later she develops advanced melanoma.

Was the gynecologist’s responsibility finished when the referral was entered?

That is not really a dermatology question.

It is a question of professional responsibility.

Check the recent ones HERE.

Bad Outcome Does Not Equal Malpractice

This distinction will run through the entire series:

Bad outcome ≠ preventable harm ≠ medical error ≠ negligence.

Good medical care can end badly.

Bad medical care can sometimes end well.

A known complication can occur without negligence.

And a medical error can occur without causing injury.

If we collapse all of these into the word “malpractice,” we learn almost nothing.

The MedMal Room will separate them.

For every case, I want to ask:

What actually happened?

What should have happened?

What information was available at the time?

Was the outcome reasonably foreseeable?

Was there an error?

Was it preventable?

Did the system fail?

Did communication fail?

And most importantly:

What can the rest of us do differently tomorrow?

This Is Preventive Ethics

I have written for years about preventive ethics: the idea that professional responsibility should begin before the crisis.

The best time to explain an emergency cesarean is not always during the emergency.

The best time to teach postpartum warning signs is not after the patient develops severe symptoms at home.

The best time to establish an escalation protocol is not while the patient is hemorrhaging.

The best time to decide who follows an abnormal result is before the result becomes abnormal.

Malpractice prevention, when done properly, is therefore not about protecting physicians from patients.

It is about protecting patients from predictable failures.

And when we do that well, physicians are protected too.

The Question I Keep Coming Back To

Across more than 30 cases, one question appears again and again:

Who owned the next step?

The radiologist found the abnormality.

The laboratory reported the result.

The nurse called the physician.

The referral was entered.

The pathology was finalized.

The blood pressure appeared on the screen.

The patient called with worsening symptoms.

Information existed.

But information without ownership does not protect a patient.

Someone has to recognize it.

Interpret it.

Communicate it.

Act on it.

And make sure the next step actually happened.

That is where The MedMal Room will live.

Why I Am Making This a Series

These cases are too important for a single article.

Over the coming months, paid subscribers to ObGyn Intelligence will receive the full MedMal Room series, one case at a time.

Not sensationalized courtroom stories.

Not generic “risk-management tips.”

Each article will examine the medicine, the evidence, the ethics, the systems failure, and the lesson clinicians can actually use.

Some cases will make you agree immediately.

Others may make you uncomfortable.

A few may make you rethink practices you have considered routine for years.

That is exactly the point.

Because the most valuable malpractice case is not the one we win or lose.

It is the one we learn from before it happens again.

If you are an obstetrician, gynecologist, MFM specialist, fertility physician, midwife, nurse, resident, medical student, risk manager, or simply interested in how medicine succeeds and fails, I hope you will join me in The MedMal Room.

There are already more than 30 cases waiting.

And every one of them asks the same ultimate question:

What could we learn from this case before the next patient walks through the door?

A fetal heart rate suddenly deteriorates.

A patient hemorrhages.

A woman with hypertension has a seizure.

Everyone runs.

The obstetrician arrives. An emergency cesarean is performed. Blood is ordered. Anesthesia responds. The neonatal team begins resuscitation.

Subscribe now

Later, when the case is reviewed, everyone asks the obvious question:

What happened during those final minutes?

Sometimes that is the wrong question.

The more important question may be:

What happened during the hours, days, or even weeks before the emergency?

That is where some of the most important failures in obstetric care occur.

The Last 20 Minutes Are Seductive

Read more

I monitor dozens of communities for clinically meaningful posts: dangerous myths, gaps between belief and evidence, stories guidelines cannot capture. This series -- ObGyn Intelligence on Reddit -- dissects them against the literature, because ObGyns who ignore social media ignore the most unfiltered window into what patients think, fear, and do between appointments.

Summary

A pregnant Redditor asked a simple question and got more than a thousand answers: does anyone still eat sushi, smoked salmon, and deli meat? The overwhelming reply was yes, in moderation, from places they trust.

What stood out was how many women pushed back on the official warnings.

Again and again, commenters noted that bagged salads, pre-cut melon, and cantaloupe cause more listeria outbreaks than any sandwich, yet nobody tells pregnant women to stop eating salad.

A few said they avoid everything, often after a loss or years of trying to conceive, and the thread met that choice with real tenderness. But the question hides a problem, and so does the advice these women were given. It lumps three very different foods together, as if the risk were the same for all of them. It is not, and the body issuing that advice knows it is not.

Three Foods, Three Different Risks

These foods belong in one sentence as risks for Listeria.

Counter-sliced deli meat qualifies: the slicer itself can carry listeria, which is why warming the meat until it steams removes most of the risk.

Cold-smoked salmon qualifies too, and more so; smoking does not reliably kill listeria, and the product is stored cold and eaten as is. One German genomic study estimated that smoked or cured salmon accounted for more than a quarter of that country’s listeriosis cases between 2018 and 2020. Of the three foods named, the smoked salmon is the real listeria concern.

Liusteria does not belong here.

Listeria monocytogenes is a soil and water organism that survives refrigeration and is killed only by cooking or pasteurization, so the genuine danger lives in refrigerated, ready-to-eat foods that sit cold for days with no further heating.

Fresh raw-fish sushi is the outlier. In the federal modeling that traces more than three-quarters of U.S. listeriosis to dairy, leafy greens, and fruit, fresh fish and other seafood account for essentially none of it. Sushi is served quickly, not held cold for days, so listeria has little chance to multiply, and it has not been implicated in U.S. listeria outbreaks. Its real risks are different: the Anisakis parasite, which the FDA’s mandatory deep-freezing of sushi-grade fish all but eliminates; ordinary stomach bugs such as salmonella that make the mother ill but rarely reach the fetus; and methylmercury, which is a matter of fish species, not of raw versus cooked.

What ACOG Says — and Why It Is Wrong

This confusion is not invented by patients. It is printed in the patient materials of the American College of Obstetricians and Gynecologists, so it is worth reading their exact words. On the page ACOG titles “Listeria and Pregnancy,” the instruction is unambiguous:

“Do not eat sushi made with raw fish (cooked sushi is safe).”

— ACOG, “Listeria and Pregnancy” FAQ

The placement is the whole problem. This sentence sits on a page about listeria, and the same page tells the reader that cooking and pasteurization are the only ways to kill the bacterium. Read together, those two statements tell a pregnant woman that raw-fish sushi transmits listeria. It does not.

ACOG has reached for the right bacterium and attached it to the wrong food.

In its general “Having a Baby” guidance, ACOG offers this:

“Raw fish may be harmful during pregnancy.”

— ACOG, “Having a Baby” FAQ

“May be harmful” is the language of liability, not of risk communication. It carries no magnitude, no mechanism, and no comparison. It cannot be acted on and it cannot be proven wrong, which is precisely why it survives in a guideline. A patient deserves a number and a reason, not a phrase engineered to be unfalsifiable.

And in its dedicated sushi FAQ, ACOG explains that raw fish, including sushi and sashimi, are:

“more likely to contain parasites or bacteria than fully cooked fish.”

— ACOG, “Can I eat sushi while I’m pregnant?” FAQ

That single phrase fuses three separate hazards into one alarming blur.

Pull them apart and the case collapses.

The parasite, Anisakis, is largely engineered out by the FDA-mandated freezing of any fish sold for raw consumption.

The bacteria that raw fish can actually carry are salmonella and campylobacter, which cause maternal stomach illness and, in the United States, do not behave like listeria and rarely reach the fetus. Listeria, the one organism that genuinely threatens the pregnancy, is the bug that raw sushi is not known to carry. Lumping these together lets the most frightening member of the group lend menace it has not earned here.

Laid out plainly, the failures are these:

1. Right bug, wrong food. ACOG invokes listeria to justify avoiding sushi, but fresh raw fish is not a documented listeria vehicle and contributes essentially none of the attributed U.S. cases.

2. Hedged, unquantified language. “May be harmful” protects the author from liability while giving the patient nothing she can weigh or act on.

3. Conflated hazards. Parasites, enteric bacteria, and listeria are merged into “parasites or bacteria,” obscuring that the dangerous one is absent and the present ones are mild and freezing-controlled.

4. Inverted risk hierarchy. The warning is loudest on salmon nigiri and quietest, in the same patient breath, on the foods that actually drive listeriosis: cantaloupe, bagged salad, pre-cut melon, soft cheeses, deli meat, and smoked fish.

5. The wrong sorting variable. ACOG sorts seafood on raw-versus-cooked. But mercury tracks fish species, not preparation, and listeria tracks refrigerated ready-to-eat shelf life, not rawness, which is why cooked-but-cold smoked salmon is a real listeria vehicle and raw nigiri is not.

To be fair to the kernel: ACOG is correct that pregnant women are roughly ten times more susceptible to listeriosis, that the consequences can be catastrophic, and that avoiding raw and undercooked seafood does reduce salmonella and parasite exposure.

A simple, conservative rule has value across a large and varied population. The problem is not that the advice is cautious. It is that it is mis-aimed, conflated, and unquantified, and it spends its credibility on the wrong target while staying quiet on the foods that fill the morgue.

What To Actually Avoid: The Listeria List

Here, explicitly, are the foods that transmit Listeria monocytogenes, grouped from highest concern down. The bacterium grows in the refrigerator and dies only with thorough cooking or pasteurization, so the through-line is simple: refrigerated, ready-to-eat, no kill step before the mouth.

1. Dairy (the largest single category)

2. Refrigerated ready-to-eat meats

3. Refrigerated smoked and ready-to-eat seafood

4. Produce (the under-recognized vehicles)

5. Refrigerated prepared and store-deli foods

Here they are:

1. Dairy (the largest single category)

• Unpasteurized (raw) milk and anything made from it

• Soft cheeses unless labeled made with pasteurized milk: queso fresco, queso blanco, panela, cotija, brie, camembert, feta, and blue-veined cheeses

• Any soft cheese under an active recall, even pasteurized (recent brie and camembert outbreaks)

2. Refrigerated ready-to-eat meats

• Deli and luncheon meats and cold cuts, especially counter-sliced, unless heated until steaming (165°F)

• Hot dogs straight from the package, unless heated until steaming

• Refrigerated pâté, meat spreads, and liverwurst

• Refrigerated pre-cooked chicken and ready-to-eat prepared meals, unless reheated

3. Refrigerated smoked and ready-to-eat seafood

• Cold-smoked salmon (lox, nova), gravlax, smoked trout, smoked whitefish, unless cooked or in a fully heated dish

• Refrigerated ready-to-eat crab, surimi, and pre-cooked shrimp, unless heated

4. Produce (the under-recognized vehicles)

• Pre-packaged and bagged salads and pre-cut leafy greens

• Pre-cut and pre-packaged fruit, especially melon; cantaloupe caused the deadliest U.S. listeria outbreak on record

• Raw sprouts of all kinds (alfalfa, bean, clover, radish); rinsing does not decontaminate them

• Enoki and other raw mushrooms implicated in repeated recalls

5. Refrigerated prepared and store-deli foods

• Store-made deli salads: chicken, ham, egg, tuna, and seafood salads

• Refrigerated hummus and dips during active recalls

• Pre-packaged hard-boiled eggs (recall history)

Across every category

• Anything currently under an active FDA or USDA listeria recall — check before assuming a food is safe

• Anything kept past its use-by date in the refrigerator, since listeria multiplies in the cold and time is the real risk multiplier

The rule patients should carry out of the office: refrigeration does not kill listeria. Only cooking or reheating to steaming (165°F) and pasteurization do. Keep the refrigerator at or below 40°F. And note what is absent from this entire list: fresh raw-fish sushi, the very food ACOG warns against on its listeria page.

What It Means

For patients, the risk is real but small, and it is not spread evenly across the menu. The precautions that move the needle are the unglamorous ones on the list above: wash produce, skip the bagged salad and pre-cut melon, heat deli meat until it steams, and choose smoked fish from a trusted source or warm it through. Raw-fish sushi from a reputable restaurant sits far down the list of things worth losing sleep over. For clinicians, the lesson is about credibility. Patients now read the same outbreak data we do. When we hand them a guideline that bans salmon nigiri but stays silent on cantaloupe, they learn that the list is about caution and liability, not about their baby, and that costs us the trust we need for the warnings that matter.

My Take

I have spent fifty years inside this system, and I have read a lot of these handouts. ACOG put sushi on a page about listeria, a bacterium that sushi does not carry, and told mothers that raw fish might be harmful without ever saying how, how often, or compared to what. That is not science. That is a phrase written to survive a deposition. Meanwhile cantaloupe, which caused the deadliest listeria outbreak this country has ever seen, rarely earns a sentence. There is something almost poetic about a woman trusting a sushi chef with twenty years at the same counter over a deli slicer cleaned, if she is lucky, once a day. She is reasoning better than the pamphlet. So tell her the truth, all of it: skip the bagged salad, heat the ham, and enjoy the salmon nigiri. A calmer pregnancy is worth more than a list nobody can defend.

Sources

1. American College of Obstetricians and Gynecologists. Listeria and pregnancy (FAQ). https://www.acog.org/womens-health/faqs/listeria-and-pregnancy

2. American College of Obstetricians and Gynecologists. Having a baby (FAQ). https://www.acog.org/womens-health/faqs/having-a-baby

3. American College of Obstetricians and Gynecologists. Can I eat sushi while I’m pregnant? (Ask ACOG). https://www.acog.org/womens-health/experts-and-stories/ask-acog/can-i-eat-sushi-while-im-pregnant

4. Centers for Disease Control and Prevention. About Listeria infection. https://www.cdc.gov/listeria/about/index.html

5. U.S. Food and Drug Administration. Listeria (Listeriosis). https://www.fda.gov/food/foodborne-pathogens/listeria-listeriosis

6. Interagency Food Safety Analytics Collaboration (CDC, FDA, USDA-FSIS). Foodborne illness source attribution estimates for Salmonella, E. coli O157, and Listeria monocytogenes — United States, 1998–2023. https://www.cdc.gov/ifsac/php/data-research/annual-report-2023.html

7. World Health Organization. Listeriosis fact sheet. https://www.who.int/news-room/fact-sheets/detail/listeriosis

8. Food Standards Agency. The risk to vulnerable consumers from Listeria monocytogenes in ready-to-eat smoked fish. 2024. https://www.food.gov.uk/research/foodborne-pathogens/the-risk-to-vulnerable-consumers-from-listeria-monocytogenes-in-ready-to-eat-smoked-fish

9. Lachmann R, et al. Invasive listeriosis outbreaks linked to salmon products, Germany, 2010–2021. [Full citation and DOI to be verified against PubMed before any manuscript use.]

10. Oster E. How dangerous is sushi during pregnancy, really? ParentData. https://parentdata.org/pregnancy/how-dangerous-sushi-during-pregnancy/

Medicine’s first response was to call this a communication training problem.

My response: we are measuring the wrong thing and naming the wrong standard.

AI has accidentally done us a favor by making the problem visible. Now we have to understand what it can fix and what only we can.

Compassion Is Not a Communication Skill

A patient in active labor at 38 weeks asked her nurse why the fetal heart rate was doing what it was doing. The nurse had three other patients. The answer she gave was accurate, brief, and insufficient. The patient remembered it for years.

That gap is real. It predates AI by decades. A study published in JAMA Internal Medicine made it measurable: patients shown physician responses and AI chatbot responses to medical questions rated the chatbot as more empathic. The finding has been replicated in oncology, in patient portal research, and in fertility care.

Medicine’s reflex has been to treat this as a training problem. If patients prefer AI responses, the argument goes, physicians need better communication skills. Teach them to be warmer. Script the acknowledgments. Train the tone.

That response is wrong, but not because AI has nothing to contribute. AI has a great deal to contribute, and this series takes that seriously. The response is wrong because it misidentifies the problem. The problem is not that physicians are insufficiently empathic. The problem is that we have been measuring empathy when we should have been demanding compassion. Those are not synonyms. Understanding the distinction is what unlocks AI’s real potential in clinical care.

The Distinction That Runs Through Everything

• Empathy is the capacity to perceive and resonate with another person’s emotional state. It is cognitive and affective: you recognize what someone is feeling, and you feel some version of it yourself. An LLM trained on millions of human conversations can produce empathic-sounding language with impressive reliability. It has learned the form. That is what the studies are measuring, and it is genuinely useful as far as it goes.

• Compassion goes further. It is empathy plus the motivation to act, and the act itself. The word comes from the Latin: to suffer with. A compassionate clinician does not just recognize that her patient is frightened. She is moved by that recognition to do something about it. She explains. She stays an extra two minutes. She calls back. She changes her language because this particular patient, in this particular moment, needs a different kind of communication. Compassion is a moral act, not a communicative one.

You can score high on an empathy scale and be a compassionless clinician. You can sound warm and be entirely absent. Patients know the difference even when they cannot name it. What they remember, sometimes for the rest of their lives, is not whether the tone was right. It is whether someone was actually with them.

What the Study Actually Measured

Bioethicist John Lantos argues that most empathy scales capture communicative empathy: warm tone, verbal acknowledgment, scripted validation. Those things matter. They are also reproducible by a language model. What the scales do not capture is what philosopher and psychiatrist Jodi Halpern calls emotional reasoning: a disciplined, medically-informed attunement to what illness means in a specific patient’s life.

The JAMA study measured patients’ perceptions of written responses to medical questions. That is a useful measurement of one thing. It is not a measurement of what happens when a woman has been laboring for 22 hours and the team is discussing cesarean. It is not a measurement of what happens when a sonographer takes too long at 19 weeks and says the doctor will come in. It is not a measurement of what happens when a 49-year-old woman brings the same symptom list to her third appointment and leaves with a pamphlet.

The AI empathy finding is a signal. What it is signaling is not that AI communicates better than physicians. It is that compassion has eroded so far in clinical training and clinical systems that a language model can outscore a physician on its outward form. That erosion is the problem this series addresses. And AI, understood correctly, is part of how we fix it.

ObGyn Intelligence: Safety analysis, the evidence critique, and the verdict are below -- for subscribers who want the full picture.

Seven Clinical Moments. One Standard.

This series examines seven areas of ObGyn medicine where the compassion deficit is most consequential and where AI has a specific and honest role to play.

Labor and delivery. The laboring patient who needs someone present under pressure, not performing presence. AI can reduce documentation burden and surface communication gaps. It cannot be present.

Gynecologic oncology. The woman who asks her oncologist whether she is going to die. AI can synthesize her chart and ensure follow-up. It cannot accompany her through treatment.

Reproductive endocrinology. The patient with two failed transfers who already knows the statistics. AI can model success probabilities and flag protocol deviations. It cannot have the honest conversation about stopping.

Stillbirth. The moment the sonographer cannot find the heartbeat. AI can ensure consistent bereavement protocols. It cannot be present to a loss that medicine cannot fix.

Miscarriage. The 10-minute confirmation appointment. AI tools are filling the midnight gap. That contribution is real and not enough.

Prenatal diagnosis. The anatomy scan that changes everything. AI can prepare patients and generate plain-language summaries. It cannot close the 72-hour window when a patient is alone with her search results.

Menopause. The perimenopausal woman whose symptom burden has been dismissed at three appointments. AI is filling the gap her clinical encounters left. The answer is a clinical encounter that deserves the time and training to be compassionate.

Each post asks the same two questions: what does compassion actually require here, and what can AI genuinely contribute? The answers are different in each clinical territory. The standard is the same throughout.

What AI Can Do — and This Is Substantial

The most important thing AI can do for compassionate clinical care is structural: reduce the documentation burden that steals time from the room. A clinician who spends 40 percent of her working hours on documentation is a clinician with 40 percent less time to be present to her patients. AI tools that automate that burden do not replace compassion. They create the conditions for it.

Beyond documentation, AI tools that synthesize patient histories, flag communication gaps, generate plain-language explanations of complex findings, prompt follow-up after high-acuity events, and connect patients to peer support between appointments are making clinicians more accurate and more consistent. In REI, AI-assisted embryo grading reduces interobserver variability. In oncology, clinical decision support ensures no patient with ovarian cancer goes without genetic counseling. In menopause care, symptom-tracking apps help women identify when their burden warrants treatment and prepare them to ask for it. These are real clinical improvements.

What Only Humans Can Do

The limit is presence. AI cannot be moved by a patient’s suffering. It cannot make the judgment, in real time, that this patient needs silence before she needs information. It cannot carry the weight of having been wrong. It cannot commit to being there through whatever comes next. Those acts define compassionate care in every one of the seven clinical territories this series examines. They require a human being with the training and the moral disposition to stay.

The chatbot learned what warmth sounds like. That is genuinely useful. It did not learn what compassion is. In ObGyn medicine, across labor and delivery, oncology, fertility, loss, and menopause, the gap between those two things can determine how a patient remembers the most significant physical experiences of her life.

Why Getting This Right Matters Now

The AI empathy finding is entering clinical policy conversations at exactly the moment when health systems are under pressure to reduce costs and increase efficiency. If we do not have a clear account of what AI can and cannot do, the path of least resistance is to delegate more clinical communication to AI tools and call it an improvement. That path leads to a system where AI handles the communication layer, clinicians handle the technical layer, and no one is responsible for the compassionate layer because empathy scores have been accepted as a sufficient proxy for it.

The right architecture uses AI to give clinicians the time and support to be compassionate, while holding the profession to a standard that AI cannot meet on its own. Getting that architecture right determines what happens in the room when there is no heartbeat, when the staging is incomplete, when the transfer has failed again, and when the symptom list comes out of the purse for the third time.

Conclusion

Medicine is measuring empathy. Patients need compassion. The two are not the same, and understanding the difference is what unlocks AI’s genuine potential in clinical care. AI can reduce the burden that prevents compassion, prepare the clinician, prompt the follow-up, and fill the midnight gap. It cannot be present, cannot be moved, and cannot stay. In every clinical territory this series examines, the task is the same: build a system where AI does what it does well and clinicians are finally given the conditions to do what only they can. That is what patients in labor, in oncology, in fertility clinics, after losses, and through menopause deserve. It is what this series argues for.

Summary

A user at 9 weeks pregnant noticed on Reddit something unexpected: her spring allergies -- severe enough to require daily antihistamines from March through June -- had essentially vanished. The post landed on r/pregnant and drew 303 comments in under 24 hours. What poured in was not advice. It was testimony. Hundreds of women describing conditions that had improved, disappeared, or transformed during pregnancy: migraines gone, depression lifted, lactose intolerance reversed, anxiety quieted, autoimmune skin conditions cleared. A few dozen noted the opposite -- everything got worse. But the dominant signal was striking: pregnancy, for many women, functions as an uninstructed biological experiment, reshaping the immune system, the nervous system, and the hormonal environment in ways that clinical medicine has not fully mapped. These women are noticing. Their doctors, for the most part, are not asking.

Evidence-Based Commentary

What this thread documents is not surprising to reproductive physiologists. It is, however, largely invisible in obstetric practice -- and that gap is worth examining.

Migraines. This was the most frequently cited improvement, mentioned by more than 30 commenters independently. The mechanism is well established. Migraines in women of reproductive age are predominantly hormonally triggered, driven by estrogen fluctuations across the menstrual cycle. Pregnancy produces stable, progressively rising estrogen levels with no cyclical drops -- exactly the environment migraine-prone brains prefer. Multiple controlled studies have confirmed that migraine frequency decreases significantly in the second and third trimesters in the majority of affected women. The first trimester can be worse due to early hormonal flux, and several commenters noted this accurately. Postpartum return is common and often rapid. Women on migraine preventatives who discontinue them for pregnancy should have a specific plan for postpartum management before delivery -- this conversation is not happening often enough.

Mood disorders, anxiety, and PMDD. The second major theme is more complex and clinically important. Dozens of women described marked improvement in depression, generalized anxiety, and what several identified as PMDD. Multiple commenters noted -- correctly -- that pregnancy puts PMDD in remission because there is no luteal phase. Premenstrual dysphoric disorder is fundamentally a disorder of cyclical hormone sensitivity; remove the cycle, the disorder goes quiet. Several women in this thread reported discovering they had PMDD precisely because pregnancy revealed the contrast. This is a legitimate diagnostic insight, and it should be explicitly discussed: women with mood symptoms that track their cycle who get pregnant and feel dramatically better should be informed of the likely return of those symptoms postpartum, and should have a postpartum mental health plan in place before discharge. The evidence for postpartum mood disorder risk in women with pre-existing PMDD or anxiety is substantial. One commenter observed that progesterone is described as “nature’s Xanax” -- this is a popular simplification of a real phenomenon. Progesterone metabolizes to allopregnanolone, a positive allosteric modulator of GABA-A receptors. The sharp postpartum drop in allopregnanolone is now understood to be a central mechanism in postpartum depression, which led directly to the development of brexanolone (Zulresso), the first FDA-approved treatment specifically for PPD. Women who feel unusually well during pregnancy are not imagining it. They should be told why, and warned what to expect.

Do this postpartum screening and mental health plan.

Immune-mediated conditions: allergies, KP, psoriasis, eczema, IBD. Pregnancy is a state of profound immune reorganization. The fetus is, immunologically speaking, a semi-foreign graft. To prevent rejection, the maternal immune system shifts toward tolerance -- reducing Th1 (cell-mediated, pro-inflammatory) activity and upregulating Th2 (humoral, anti-inflammatory) responses. This shift is well documented. The clinical consequences reported in this thread follow predictably. Autoimmune and inflammatory conditions that are Th1-driven -- psoriasis, rheumatoid arthritis, some eczema variants, Crohn’s disease -- tend to improve during pregnancy. Conditions that are Th2-driven -- atopic dermatitis in some forms, asthma -- may worsen. Multiple commenters reported keratosis pilaris (KP) disappearing during pregnancy; the proposed mechanism involves hormonal effects on keratinocyte differentiation, though the evidence base is limited. The interstitial cystitis improvement noted by one commenter is consistent with case reports and small series in the literature. Ulcerative colitis remission during pregnancy, noted by another user, is well documented -- the pregnancy-associated Th2 shift suppresses the inflammatory cascade driving UC. Women with IBD should be counseled pre-pregnancy that their disease behavior may change and should continue specialist monitoring throughout.

Lactose intolerance. This was reported by 15 or more commenters and is probably the most mechanistically interesting finding in the thread. Lactase -- the enzyme required to digest lactose -- is produced in the small intestinal brush border. There is evidence that estrogen and progesterone influence intestinal lactase expression and GI motility. Pregnancy also significantly slows GI transit time, which may allow more time for incomplete lactase activity to process lactose before symptoms develop. The clinical implication is modest but worth knowing: some women who believe they cannot tolerate dairy may tolerate it better during pregnancy. Blanket dairy restriction in pregnant women who experience lactose intolerance prenatally is not necessarily warranted -- they may be able to tolerate dietary calcium sources they had previously avoided.

What nobody is telling these women. Across 303 comments, the pattern is unmistakable: women are experiencing significant changes in established medical conditions during pregnancy, they are surprised by these changes, and they are sharing them with each other because no one told them to expect them. Prenatal care systematically addresses what pregnancy will add -- nausea, heartburn, back pain, fatigue, edema. It rarely addresses what pregnancy may temporarily resolve and what will return. The postpartum period is when the reckoning arrives: migraines come back, PMDD returns, psoriasis flares, anxiety resurfaces. Women who were not prepared for this are blindsided. The informed consent failure is not dramatic -- no one is harmed acutely -- but it is real. Preparing patients for postpartum biological rebound is part of comprehensive obstetric care, and most practices are not doing it.

What It Means

Pregnancy temporarily reorganizes the immune system, stabilizes estrogen, floods the brain with neurosteroids, and alters GI function. For millions of women, this means conditions they have managed for years -- sometimes their entire adult lives -- go quiet for nine months. The clinical community knows the mechanisms. It is not translating that knowledge into patient counseling. Women are learning from each other that migraines, PMDD, IBS, psoriasis, and anxiety may improve during pregnancy, and they are being surprised when these conditions return postpartum. Obstetric providers should be having explicit conversations about anticipated postpartum rebound for any patient whose pre-existing condition improved during pregnancy. That conversation belongs in the third trimester, not in the postpartum ER.

My Take

I have spent 50 years in obstetrics. I have watched women become different people during pregnancy -- calmer, clearer-headed, migraine-free, pain-free, allergy-free -- and I have watched them come back six weeks postpartum asking what happened to them. We knew what was going to happen. We did not tell them.

The top comment in this thread -- 288 upvotes -- was from a woman who said pregnancy taught her to say no. The second was about migraines. The third was about leg hair. This is what patients think about. This is what they notice. They are running an n-of-1 experiment on themselves every pregnancy, collecting data that medicine should be systematically gathering.

What strikes me most is the PMDD thread.

Multiple women realized they had PMDD only because pregnancy made it disappear.

That is a diagnostic finding. It belongs in a chart.

It should trigger a postpartum screening and mental health plan.

How many women who are “just hormonal” before pregnancy actually have PMDD, go undiagnosed because no one frames the question correctly, and then crash postpartum without any preparation? The answer is: more than we are catching. Reddit figured this out in the comments. We should be doing it in the office.