A critique of the leading study of American homebirth

American homebirth is one of those topics that can turn a calm dinner party into a full-blown debate club meeting faster than you can say “birth plan.” Some people hear homebirth and picture candlelight, quiet, and a midwife with the patience of a saint. Others picture an ambulance siren and a relative who suddenly “read an article” and is now an expert. The truth is less cinematic and more complicated: homebirth outcomes depend on who is giving birth, who is attending, what risks are present, and how well the local system handles the moment when birth decides to freestyle.

This article takes a close, practical look at the most influential U.S. study that is frequently cited as the “largest” or “leading” analysis of planned American homebirth: the Midwives Alliance of North America Statistics Project (often shortened to MANA Stats), published in 2014 and covering births from 2004–2009. It is widely referenced because it’s big, detailed, and built around intended place of birthan important point that many datasets still struggle to capture.

But “big” doesn’t automatically mean “settled.” The MANA Stats paper contains valuable informationand also has limitations that matter a lot if you’re using it to make broad claims about safety, risk, and policy. Let’s unpack what the study actually shows, what it does not show, and how to read homebirth research without falling into the trap of thinking one paper gets to be the Supreme Court of Birth.

Meet the study: what MANA Stats actually examined

The 2014 MANA Stats outcomes paper describes 16,924 people who planned a home birth at the onset of labor, with data drawn from a national, web-based registry created by midwives. In the study, about 89.1% of participants ultimately gave birth at home; the rest experienced intrapartum transfer, postpartum transfer, or other changes in setting.

The reported intervention rates were low (for example, a small fraction required oxytocin augmentation and/or epidural analgesia after transfer). The cesarean rate in the overall cohort was reported as 5.2%, and the vaginal birth after cesarean (VBAC) success rate among those attempting it was high. Newborn outcomes were also presented, including low Apgar scores and perinatal mortality figures.

In plain English: MANA Stats describes what happened to a large group of people who planned homebirth with midwife-led careand it documents high rates of physiologic birth and low rates of certain interventions. That is meaningful, especially in a U.S. system where intervention rates in hospitals can be high and patient experience can be uneven.

What the study does well (and why it became “leading”)

1) It focuses on planned homebirth, not “baby arrived in the bathtub, surprise!”

One of the biggest data problems in U.S. homebirth research is the confusion between planned and unplanned home births. Planned homebirth tends to involve prenatal care and an attendant; unplanned homebirth is more likely to be linked with limited prenatal care, social risk factors, or precipitous labor. Mixing these together can make any analysis look like it was stirred with a broom.

MANA Stats explicitly analyzes births that were planned as homebirths at labor onset, which is closer to the real question most families are asking: “If I plan a homebirth, what tends to happen?”

2) It captures outcomes across settings (intended vs actual place of birth)

Many population datasets only record where the baby was born (home vs hospital) and do not reliably record where the person intended to give birth. That matters because transfers are a featurenot a failureof community birth systems. If someone plans homebirth but transfers during labor, a simple “hospital birth” label can erase the story you’re trying to study.

A registry design can, in theory, handle this better than standard birth certificate data, and MANA Stats aimed to do exactly that.

3) It offers granular, practice-relevant details

Beyond “alive/not alive,” the paper describes reasons for transfer, intervention patterns, VBAC attempts, and breastfeeding at six weeks. Those are the kinds of details clinicians and families actually care aboutand they can inform system improvements (especially around transfer).

Now the critique: where the paper’s power ends

Here’s the big idea: the MANA Stats study is best read as a descriptive portrait of outcomes in a particular registry, during a particular era, among participants captured by a voluntary reporting system. It is not a population-wide verdict on American homebirth safety. That’s not an insult; that’s simply being honest about research design.

1) The sample is not population-based (translation: it may not represent U.S. homebirth as a whole)

The authors themselves describe a central limitation: the registry is a voluntary sample and not a mandatory, comprehensive system capturing all planned homebirths in the United States. That means we cannot confidently say the outcomes represent all planned homebirths nationwideor even all planned homebirths attended by similar providers.

Why does this matter? Because selection bias can creep in through several doors at once:

• Who participates: If certain midwives, practices, or regions are more likely to contribute, outcomes can tilt.
• Which births get entered: Even with good intentions, incomplete case capture can distort rates.
• Who chooses homebirth: Planned homebirth populations often differ from hospital populations in education, parity, baseline health, and preferences.

A population-based dataset has its own flaws (we’ll get to those), but it at least includes everyone. A registry offers depth, but can sacrifice representativeness.

2) Credentialing and practice variation are hard to standardize in U.S. community birth

The U.S. doesn’t have one unified, nationwide midwifery system. State laws vary, credential pathways vary, collaboration agreements vary, and hospital reception to transfers ranges from “welcome, let’s help” to “why are we yelling already?”

MANA Stats includes midwife-led births, but the paper notes it could not precisely quantify what proportion of midwives of various credentials contributed data in that era. If you’re trying to make a policy claim like “planned homebirth is safe if attended by X,” you need strong clarity on who “X” is across the dataset.

In other words: the U.S. is not one homebirth environment. It’s fifty different regulatory realities wearing a trench coat.

3) Rare outcomes are statistically tricky (and homebirth debates hinge on rare outcomes)

Perinatal death and serious neonatal neurologic injury are, thankfully, rare in high-resource settings. The downside is that rare outcomes are hard to estimate preciselyeven with large samples. A handful of cases one way or another can shift rates, confidence intervals, and conclusions.

The MANA Stats paper reports perinatal mortality figures, and it also acknowledges challenges like limited statistical power in subgroups and inconsistent definitions across the broader literature. If you’re trying to compare “home vs hospital” as if this were a simple product review (“five stars, would birth again”), the math will humble you quickly.

4) Handling congenital anomalies and cause-of-death classification can change the story

Mortality reporting can look different depending on whether analyses include lethal congenital anomalies, how deaths are categorized (intrapartum vs early neonatal vs late neonatal), and how consistently causes are confirmed (for example, autopsy rates and documentation can vary).

The MANA Stats discussion notes that unknown causes of death and congenital anomaly-related deaths can complicate comparisons, especially when confirmatory testing is declined or unavailable. This isn’t unique to MANA Stats; it’s a recurring problem in birth setting research. But it becomes especially important when the public debate treats “the mortality number” as if it’s a single, universal truth.

5) Data quality: “medical record–based” is good, but it’s not magic

Registry data drawn from medical records can be stronger than self-report, but it still depends on consistent measurement and documentation. The MANA Stats paper itself highlights a classic example: postpartum hemorrhage estimates can be unreliable when based on visual estimation of blood loss, which is known to be imprecise across settings.

That doesn’t mean outcomes were “made up.” It means some clinical variables are inherently hard to measure consistentlyand that matters when you compare settings or attempt to generalize.

The “missing comparison group” problem: what MANA Stats can’t prove

MANA Stats is not a randomized controlled trial. It does not include a contemporaneous, matched hospital cohort inside the same paper, with identical definitions, follow-up windows, and risk adjustment. So it cannot, on its own, prove that homebirth is safer, as safe, or less safe than hospital birth in the U.S.

It can strongly suggest that many planned homebirths in the registry had low intervention rates and generally favorable outcomes. But comparison claims require comparison designs.

This is where the U.S. literature splits into two big approaches:

• Registry and intention-to-treat studies (like MANA Stats) that track planned births and can capture transfers well, but may not be population-based.
• Vital records and linked birth–infant death studies that are population-based, but often struggle to correctly identify intended place of birth and planned transfers.

The National Academies have emphasized this tension: birth certificate data are comprehensive but can misclassify planned community births as hospital births when transfers occur, because many systems record only the actual place of birth at delivery rather than intent at labor onset. That misclassification can bias comparisons in either direction depending on how analyses are constructed.

So what do other U.S. studies showand why do they disagree?

If you’ve ever wondered why one headline says “Homebirth is safe!” and another says “Homebirth doubles risk!” the answer is not that researchers can’t do math. It’s that they’re using different datasets, definitions, and inclusion criteriasometimes with profoundly different abilities to capture intention and transfers.

Linked birth–infant death analyses: powerful, but vulnerable to exposure misclassification

Studies using linked birth and infant death records have reported higher neonatal mortality rates associated with planned homebirth compared with certain hospital comparison groups. These analyses have the advantage of large numbers and population coverage. They also face a core methodological challenge: accurately identifying which births were planned homebirths (and which were unplanned) and ensuring that transfers are handled correctly.

Even small misclassification can matter when outcomes are rare. And because transfer cases may be recorded as “hospital births” in vital statistics, the “planned homebirth” group can be left looking artificially low-risk while the hospital group absorbs higher-acuity transferred casesunless the dataset has strong intent markers.

State-level integrated community midwifery studies: context matters

Some U.S. state-level research in settings with more established community midwifery integration has found low adverse outcome rates and similar outcomes between planned home and planned freestanding birth center births for low-risk pregnancies. These studies tend to emphasize that transfer systems and credentialing standards are not minor details; they are central to safety.

Transfer data illustrate why parity matters: first births have substantially higher intrapartum transfer rates than subsequent births in multiple studies and reviews. A critique of any “homebirth safety” paper that ignores parity is like reviewing a restaurant without mentioning whether you ate the food.

How to read homebirth research without getting tricked by your own brain

Your brain loves a simple story: “Home is safe” or “Hospital is safe.” Reality offers a less dramatic but more useful question: Which births, under which conditions, in which systems?

Here is a practical checklist for evaluating any “planned homebirth” study, including MANA Stats:

1) Does it separate planned vs unplanned homebirth?

Planned and unplanned homebirth populations differ substantially in both social and clinical risk profiles. If a study blends them, treat conclusions with caution.

2) Does it measure intent at labor onset?

Intent earlier in pregnancy can change; intent at labor onset is often more relevant to intrapartum outcomes. Studies that identify “planned homebirth” only by actual place of birth can misclassify transfers.

3) How does it handle transfers?

A well-functioning system expects transfers and handles them smoothly. A study should clarify the rate, reasons, urgency, and outcomes of transfersnot treat transfer like an embarrassing secret to be hidden under the rug.

4) Who attended the birth, and what standards applied?

The U.S. has variation in credentialing, scope, and legal frameworks. “Midwife-attended” can mean different things in different places. Without clarity, comparisons become shaky.

5) How are outcomes defined and verified?

Neonatal mortality (within 7 days? within 28 days?), perinatal mortality (does it include intrapartum fetal death?), and neurologic outcomes can be defined and recorded differently. If definitions differ, rates may not be comparable.

6) Are results presented with absolute risk and uncertainty?

A twofold increase sounds terrifying until you learn it may represent a change from “very rare” to “still very rare.” Absolute risk, confidence intervals, and context should always be part of the conversation.

What a fair critique of MANA Stats sounds like

A fair critique isn’t “MANA Stats proves homebirth is perfect” or “MANA Stats is useless.” A fair critique is:

• MANA Stats is valuable as a large descriptive cohort that captures planned homebirth and transfer-related outcomes in a way many population datasets cannot.
• MANA Stats is limited because it is not population-based, relies on voluntary reporting, and has constraints around credential quantification, subgroup statistical power, and standardized measurement for certain outcomes.
• MANA Stats cannot, by itself, settle safety comparisons between home and hospital births nationally. That requires multiple complementary data sources and careful methods to handle intent and transfers.

If you want one sentence to tape to the fridge, make it this: Homebirth research is not a single-number contest; it’s a systems question. And systems are messyespecially in a country where you can cross a state line and suddenly the rules, provider types, and hospital relationships change.

What this means for policy and practice

The National Academies have argued that no birth setting is risk-free and that integration across settings, continuous risk assessment, workforce investment, and better data infrastructure can improve safety. If policymakers use MANA Stats to argue “there’s no problem here,” they risk ignoring real gaps (like transfer hostility, uneven credentialing, and data limitations). If policymakers use other datasets to argue “ban it,” they risk pushing births underground, worsening outcomes, and ignoring why people seek community birth in the first place.

The most constructive direction is not ideological. It’s practical:

• Improve the ability of vital records to capture intended birth setting and planned attendance.
• Strengthen interprofessional transfer pathways so emergencies are handled quickly and respectfully.
• Support risk-appropriate care, including access to hospital-based options for VBAC, breech, and twins where clinically appropriate.
• Invest in the maternity workforce and reduce structural barriers that leave many communities with limited choices.

If the U.S. treated transfers the way airports treat connecting flightsexpected, planned, and supportedthis entire debate would look different. Nobody gets mad at you for “failing” to fly direct when there wasn’t a direct flight. They get mad when the connection is chaotic, delayed, or hostile.

Conclusion

The leading American homebirth registry study (MANA Stats) offers a detailed look at outcomes among thousands of people who planned a midwife-led homebirth, documenting low intervention rates and generally favorable outcomes in that cohort. But its design also limits the claims it can support: it is not a nationwide census, it cannot single-handedly prove comparative safety against hospital birth, and it highlights the broader research problem that U.S. data systems still struggle to capture intent and transfers consistently.

A smart critique doesn’t throw the paper away; it puts it in its proper place: a major contribution, not a final verdict. When you synthesize MANA Stats with population-based data, state-level studies, professional guidance, and the National Academies’ emphasis on system integration, the picture becomes clearer: outcomes are shaped not only by location, but by risk selection, provider preparation, and the speed and quality of transfer when needed.

For families making decisions, the most evidence-aligned question is not “Is homebirth safe, yes or no?” It’s: “Am I low risk? Who is attending? What are their credentials and emergency skills? How far is the hospital? What is the transfer plan? And does the local system collaborateor fight?”

Experiences from the real world (the part research tables don’t fully capture)

Numbers are essential, but they don’t tell you what it feels like to be the person in labor, the midwife watching subtle signs, or the hospital team receiving a transfer at 2:00 a.m. And since this article is a critique of research, it’s worth naming the “experience layer” that shapes outcomesand shapes what gets measured.

Experience #1: The decision is often about trust, not décor. A lot of people assume homebirth is chosen for “vibes.” Sometimes it is (no judgmentfluorescent lights are nobody’s soulmate). But many families describe a deeper motive: they want to feel heard, unhurried, and respected. They may have prior trauma from a difficult hospital birth, fear unnecessary interventions, or simply want continuity with a known provider. Ironically, this means the “homebirth population” can include highly motivated, highly informed people who plan intenselysometimes more intensely than the average hospital birth, where the system is designed to catch problems even if you don’t read three books and color-code your snacks.

Experience #2: Transfers can be medically routine and emotionally intense at the same time. In many planned community births, transfers happen for non-emergent reasons: slow labor progress, exhaustion, pain relief, rising blood pressure, or a fetal heart pattern that needs a closer look. Clinically, that can be straightforward. Emotionally, it can feel like whiplashespecially for first-time parents who expected “home all the way.” This is where the quality of the system matters. When transfer is treated as collaboration, families report feeling safe. When transfer is treated as a courtroom cross-examination (“So… why did you do that at home?”), families report feeling shamed, stressed, and less willing to seek care quickly next time. That experience can influence outcomes indirectly by changing how soon people agree to transfer.

Experience #3: The “safest place” is partly a logistics question. A well-prepared homebirth team may have oxygen, IV supplies, uterotonics for hemorrhage prevention/management, neonatal resuscitation skills, and a plan for rapid transport. But there are limitsespecially for time-sensitive emergencies requiring surgery or advanced neonatal support. Families living 10 minutes from a hospital with a known receiving pathway face a different risk reality than families living 45 minutes away on winter roads while the nearest facility is on diversion. People don’t always plan for the “what if the hospital is slammed tonight?” scenario, but clinicians worry about it constantly, because the baby and placenta do not care about traffic.

Experience #4: Provider relationships can make or break safety. Homebirth midwives and hospital clinicians can share the same goalhealthy parent, healthy babyyet still clash over philosophy, liability fears, and communication norms. In places where midwives and hospitals regularly collaborate, safety practices tend to look more “system-like”: prenatal risk screening is consistent, transfers are smoother, and everyone knows the protocols. In places where the relationship is adversarial, everyone loses: midwives may delay transfer to avoid confrontation, hospitals may overreact to transfers because they arrive without context, and families are caught in the middle like the world’s least fun group project.

Experience #5: What people remember is not the statisticit’s the story. One family will tell you homebirth was “the most empowering thing ever,” because everything went smoothly and they felt supported. Another will tell you it was terrifying because they transferred urgently and felt judged, or because the baby needed NICU care and the transition was chaotic. Both stories can be true without cancelling each other out. This is why the best research doesn’t just compare settings; it studies systems, transfer quality, candidate selection, and the lived experience of care. The birth setting debate is not only about outcomesit’s about how we deliver care in a way that is both safe and humane.

If that sounds like a lot to ask, welcome to healthcare. It’s always a lot. But it’s also the point: the real opportunity in U.S. homebirth research isn’t to win an argument; it’s to build conditions where fewer people have to choose between feeling safe medically and feeling safe emotionally.

SEO Tags