On March 5, 2026, HHS Secretary Robert F. Kennedy Jr. and Education Secretary Linda McMahon announced that 53 medical schools across 31 states have committed to require at least 40 hours of nutrition education for students starting in fall 2026. I agree with this direction. In fact, I have been saying it for years. The average US medical student currently receives about 1.2 hours of formal nutrition education per year. Fewer than one in four US medical schools required a nutrition course as of 2024. That is not a curriculum gap. That is a curriculum failure.

This report is not about the politics of the announcement. RFK Jr. made some claims that go well beyond what the evidence supports. I will leave the carnivore diet commentary to others. What I want to do here is more useful: tell OBGYNs specifically what they need to know about nutrition, what they should be counseling at every prenatal visit, and where the evidence is solid versus still evolving.

But I want to start with an argument the HHS announcement barely touches. Forty hours in medical school is better than 1.2 hours per year. It is still too late.

The Women's Health Tech Report: Safety analysis, the evidence critique, and the verdict are below -- for subscribers who want the full picture.

The Problem Starts Long Before Medical School

“Dum differtur vita transcurrit.”

Seneca, Epistulae Morales, 1.1 | “While we delay, life passes.”

A future physician arrives at medical school having already formed her dietary habits. She has eaten through 22 years of childhood, adolescence, and college on whatever her family, school cafeteria, and college dining hall offered. She has spent late nights studying on pizza and energy drinks. She has never had a single structured lesson on what food does to the human body. Now, in year three of a four-year medical school program, she will receive 40 hours of nutrition education and be expected to counsel patients for the rest of her career.

We are building the house after the foundation has already set.

The HHS initiative recognizes medical schools as the intervention point because they are the easiest institutional lever to pull. But the evidence on nutrition education is clear: habit formation occurs early, and the period from adolescence through early adulthood is when dietary patterns are most malleable and most predictive of adult chronic disease risk. A 2021 systematic review in the International Journal of Behavioral Nutrition and Physical Activity found that school-based health education interventions in adolescents aged 10-19 years produced measurable improvements in fruit and vegetable intake and reductions in sugar-sweetened beverage consumption. The effect sizes were modest, but they were real. These interventions cost a fraction of what downstream chronic disease management costs. (6)

Two things need to happen alongside the medical school push. First, high schools should require nutrition education as part of a health curriculum, not as an elective, not as a unit inside a physical education class, but as a standalone, evidence-based course. Fewer than 1% of American children and adolescents currently meet recommended dietary guidelines. That number does not improve by waiting until those children are in medical school. Second, undergraduate colleges, especially those with pre-health tracks, should include applied nutrition as a core science requirement. A student who cannot name the function of folate or explain why protein intake matters for fetal brain development has no business counseling pregnant patients on those topics, no matter how many clinical rotations she completes.

The HHS initiative is a floor, not a ceiling. Forty hours in medical school matters. It also arrives 20 years too late to shape the physician’s own relationship with food. If Secretary Kennedy is serious about reversing the chronic disease epidemic through nutrition, the education reform needs to reach much further down the pipeline. The Department of Education is a co-signatory on this initiative. That opens the door. Use it.

The rest of this post is for paid subscribers.

Subscribe now

Read more

The Organ We Are Ignoring

Every prenatal checklist in the country tells pregnant women to see a dentist. Good advice. Pregnancy hormones affect gum tissue, and periodontal disease has been linked to preterm birth.

But pregnancy hormones also affect the eyes. They change tear production, corneal thickness, intraocular pressure, and the blood vessels of the retina. For women with diabetes or hypertension, pregnancy can accelerate retinal disease. For women taking fertility medications like clomiphene, visual disturbances are a recognized side effect that can, in rare cases, be permanent.

Nobody tells women any of this.

Here is what makes the gap even more striking. The retina is the only place in the human body where you can directly observe blood vessels without cutting anything open. It is a live window into vascular health, neurological health, and metabolic function. And in the last three years, artificial intelligence has turned that window into one of the most powerful diagnostic tools in medicine.

AI can now look at a retinal photograph and predict cardiovascular risk, detect undiagnosed diabetes, estimate biological age, identify kidney disease, and, most relevant to obstetrics, predict preeclampsia months before a woman develops symptoms.

We tell pregnant women to visit the dentist. We do not tell them about the organ that may soon screen for the most dangerous complication in obstetrics.

This report covers the technology that is about to change that.

The Women's Health Tech Report: Safety analysis, the evidence critique, and the verdict are below -- for subscribers who want the full picture.

The Retina: Medicine’s Best-Kept Diagnostic Secret

The field is called oculomics: using the eye to diagnose and monitor diseases that have nothing to do with vision. The concept is simple. The retina is a thin layer of neural tissue at the back of the eye, supplied by a dense network of tiny blood vessels.

The retina is the only place in the human body that allows a view directly into your brain.

These vessels share anatomical and physiological properties with the vasculature of the heart, brain, and kidneys. When systemic disease damages small blood vessels anywhere in the body, the retina often shows it first.

Ophthalmologists have known this for decades. Hypertensive retinopathy, diabetic retinopathy, and papilledema are not eye diseases. They are systemic diseases that happen to be visible through the eye.

What is new is the scale and precision that AI brings to this observation.

In a landmark 2018 study published in Nature Biomedical Engineering, Google researchers trained deep learning models on retinal photographs from over 284,000 patients. The AI predicted cardiovascular risk factors that ophthalmologists did not know were visible in retinal images: age (within 3.26 years), sex (AUC 0.97), smoking status (AUC 0.71), and systolic blood pressure (within 11.23 mmHg). It also predicted major adverse cardiac events (1).

Baiju VP, Subash R, Venkatesan N. Integrated artificial intelligence and omics for prediction and monitoring of pre-eclampsia. Int J Gynaecol Obstet. 2026 Jan 30. doi: 10.1002/ijgo.70820. Online ahead of print. PMID: 41614371.

Pre-eclampsia is a difficult pregnancy condition that causes high blood pressure and can lead to health complications in both mother and newborn, resulting in a higher fatality rate. It presents with a wide range of symptoms and lacks specific indicators, as the contemporary diagnostic techniques, including proteinuria testing and blood pressure measurements, are not reliable.

The current evolution in artificial intelligence (AI) technology tends to show a promising transformation of pre-eclampsia management. AI algorithms are applied to process larger sets of clinical, biochemical, and image data that facilitate timely medical interventions by bringing up the early-onset and severity of pre-eclampsia. By analyzing the red cell distribution width (blood test indicators for pre-eclampsia), it is recognized as a cost-effective way of detecting inflammation.

The application of AI technology on non-invasive diagnostic (wearable) devices enables continuous monitoring with imaging techniques for the placenta and retina via cloud-based systems. These developments are not only applied for early detection of pre-eclampsia, but also assist decision making capabilities in both high- and low-resource environments.

The retina, it turns out, encodes far more information than the human eye can extract. AI can read what we cannot see.

What follows is the part most readers never find: the practical, evidence-based guidance that separates knowing a technology exists from knowing how to use it. For clinicians and patients navigating women's healthcare, that is not a convenience; it is a competitive and clinical advantage worth far more than the cost of a subscription.

Keep reading with a 7-day free trial

Subscribe now

Read more

There is a persistent myth in obstetrics: that electronic fetal monitoring is a failed technology.

That cardiotocography (CTG) [aka “fetal monitoring”] has increased the cesarean rate without reducing cerebral palsy.

That it was, as some have called it, the worst test ever introduced into clinical practice.

This is wrong.

The technology is not the failure. The monitor records two channels: the fetal heart rate and uterine contractions. It does this reliably. The failure is in how we teach clinicians to interpret what the monitor shows, and specifically, how we are being taught interpretation.

A new paper in Obstetrics & Gynecology makes this painfully clear. The authors compared ACOG (The American College of Obstetricians and Gynecologists) and AWHONN (The Association of Women's Health, Obstetric and Neonatal Nurses) guidelines across fetal heart rate interpretation, uterine activity, oxytocin management, and labor management. They found 14 areas of disagreement against 8 of agreement. But the real story is not the count.

It is what the disagreements reveal about whose interests each organization serves.

The Monitor Is Not the Problem. The Interpretation Is.

Cardiotocography records two signals that must be read together. A fetal heart rate deceleration means one thing in the context of normal contractions and something entirely different during tachysystole. The contraction channel is not decoration. It is half the diagnostic information.

Here is where we have failed. ACOG’s interpretation framework, the three-tier category system, focuses almost entirely on the fetal heart rate. Uterine activity gets a passing mention. ACOG does not define what constitutes excessive contraction duration, elevated resting tone, or inadequate relaxation time between contractions. The only uterine activity parameter ACOG recognizes is the NICHD definition of tachysystole: more than five contractions in 10 minutes.

That is like reading an EKG and ignoring the rhythm strip.

AWHONN, by contrast, defines specific uterine activity parameters that clinicians can measure and act on: contractions lasting two minutes or longer, resting tone above 25 mmHg by IUPC, relaxation time under 60 seconds in the first stage and under 45 seconds in the second stage. These are concrete, measurable criteria. You can teach them. You can program them. You can build a clinical decision support tool around them.

ACOG’s silence on uterine activity parameters is not restraint. It is an interpretation framework with a blind spot in one eye.

The Women's Health Tech Report: Safety analysis, the evidence critique, and the verdict are below -- for subscribers who want the full picture. The following includes an interactive tool to help you understand the differences among interpretations.

Read more

The Technology

Issue 4 of this series established why prompt engineering determines the quality of clinical AI output and identified the six elements of a well-constructed clinical prompt. This issue translates those principles into specific, usable frameworks for the clinical questions ObGyn practitioners encounter most frequently.

The frameworks below are not hypothetical. They are constructed from the evidence on what prompt structures consistently improve AI output quality in medical question-answering, from my own clinical experience testing AI systems on obstetric questions, and from the documented failure modes that produce misleading responses. Each template can be copied, modified for your specific clinical situation, and used directly in any major AI platform. [CITATION NEEDED -- systematic review of prompt engineering in clinical medical question-answering]

The Clinical Application

The case for structured clinical prompting rests on a straightforward empirical observation: the same AI system asked the same clinical question in different ways produces substantially different responses. In a study testing AI performance on clinical reasoning tasks, structured prompts incorporating role definition, clinical context, and explicit uncertainty requests outperformed unstructured queries by a measurable margin on accuracy, completeness, and appropriate acknowledgment of uncertainty.

For a busy ObGyn practice, this creates a practical challenge. Nobody has time to construct a perfectly engineered prompt from scratch for every clinical question. The solution is not to spend more time on each prompt -- it is to develop a small library of prompt structures that can be quickly adapted to specific clinical situations. That is what this issue provides.

The templates below are organized by the type of clinical question: management decisions, patient education, literature synthesis, and guideline interpretation. Each includes a worked example and notes on common errors to avoid.

The Women's Health Tech Report: Safety analysis, the evidence critique, and the verdict are below -- for subscribers who want the full picture.

Read more

Every year, approximately 21,000 stillbirths occur in the United States. Most happen in pregnancies that looked normal. No warning. No second chance.

For decades, we have asked pregnant women to count how many times their baby kicks. The idea is simple: if the baby moves less, something might be wrong. But here is the problem. Women miss most fetal movements entirely. In controlled studies, maternal perception of fetal movements is shockingly low. We have known this for years. We have not had a better option.

Subscribe now

A new study published in Obstetrics and Gynecology suggests that a smartphone placed on the abdomen might do what mothers cannot: reliably detect and classify fetal movement using artificial intelligence. The claim is bold. The results are interesting. But there is a long road between a proof-of-concept study and a tool that saves lives.

What the Study Did

Moise et al. conducted a prospective study at the University of Texas at Austin. They enrolled 136 pregnant women between 24 and 38 weeks of gestation. The women had uncomplicated singleton pregnancies. Two cohorts were studied: 30 women followed repeatedly over time, and 106 women seen once.

Each participant had an iPhone placed on her abdomen for 30-minute sessions while a simultaneous ultrasound recorded actual fetal movements. Women pressed a key on a laptop whenever they felt the baby move. That created three simultaneous data streams: what ultrasound showed, what the phone heard, and what the mother felt.

The audio recordings were processed using machine learning. The system converted sound into numerical features called Mel-frequency cepstral coefficients (MFCCs) and trained a model to recognize different types of fetal movement from those acoustic patterns. The algorithm was adjusted for gestational age and maternal body weight, both of which affect how sound travels through tissue.

The Women's Health Tech Report: Safety analysis, the evidence critique, and the verdict are below -- for subscribers who want the full picture.

Read more

The quality of what AI gives you is almost entirely determined by how you ask. In women’s health, that gap between a good prompt and a bad one is not an inconvenience -- it is a clinical risk.*

The Technology

Prompt engineering is the practice of structuring questions and instructions to AI systems in ways that produce more accurate, more useful, and more appropriate responses. It is not a programming skill. It does not require technical training. It is, at its core, the art of communicating precisely with a system that interprets language literally and has no access to the context you have not provided.

Every clinician and every patient who has typed a question into ChatGPT, Claude, Gemini, or any other AI chatbot has been doing prompt engineering -- almost always without knowing it, and almost always doing it poorly. The difference between a well-constructed prompt and a poorly constructed one is not subtle. It routinely determines whether the AI gives you a useful clinical framework or a confident-sounding answer that is wrong in ways that are not immediately obvious.

The Clinical Application

The appeal is straightforward. AI systems contain an enormous amount of medical knowledge. They can synthesize literature, explain mechanisms, draft patient education materials, summarize guidelines, and help clinicians think through differential diagnoses. For a field like obstetrics and gynecology, where the evidence base is vast, the guidelines are frequently contested, and the clinical questions are often complex and time-pressured, a well-functioning AI assistant could be genuinely valuable.

The problem is that most clinicians and patients interact with AI the way they use a search engine -- typing a short, vague question and expecting the system to infer what they actually need. Search engines are designed to handle that kind of query. AI language models are not. They respond to exactly what you ask, filling gaps with plausible-sounding content that may or may not reflect your actual clinical situation.

A clinician who types “what is the management of preeclampsia” will get a generic, textbook-level answer that may not reflect current evidence, may not account for gestational age, and will not ask whether the patient has comorbidities, what her blood pressure trend looks like, or whether she is already on magnesium. A clinician who types “I have a 32-week patient with new-onset blood pressure of 158/102, proteinuria of 2+ on dipstick, and a headache that started this morning. She has no prior hypertension. What does current ACOG guidance say about inpatient versus outpatient management, and what are the criteria for delivery at this gestational age?” will get a response that is specific, actionable, and worth reading critically.

The same question. Completely different outputs. The difference is the prompt.

This is not a trivial distinction in women’s health. Prenatal care, labor management, gynecologic oncology, and reproductive medicine all involve clinical questions where the wrong answer -- confidently delivered -- can delay appropriate intervention, falsely reassure a patient, or lead a clinician down a management path the evidence does not support. The stakes of bad prompting in a clinical context are not the same as the stakes of bad prompting when asking AI to plan a vacation.

The Women's Health Tech Report: Safety analysis, the evidence critique, and the verdict are below -- for subscribers who want the full picture.

Read more

In 2024, the Society for Maternal-Fetal Medicine published an updated emergency checklist for placenta accreta spectrum in the American Journal of Obstetrics and Gynecology. The recommendation for how to deploy it? Laminate it and place copies in the OR, the code book, the anesthesia cart, and the hemorrhage cart.

Laminate it.

We have large language models that can process an entire patient chart in seconds, cross-reference drug interactions, adjust risk calculations in real time, and deliver context-specific guidance at the point of care. And the leading journal in our field is telling us to laminate a piece of paper.

I do not say this to be glib. I say it because there is a fundamental mismatch between the tools we publish and the tools that exist, and that mismatch has consequences for patient safety.

Checklists Work. The Problem Is How We Deliver Them.

Let me be clear about what I am not saying. I am not saying checklists are useless. The evidence is strong. The WHO Surgical Safety Checklist reduced surgical mortality by 47% and complications by 36% in its original pilot across eight countries. In obstetrics, oxytocin checklists have improved neonatal outcomes. Cesarean delivery safety checklists, published by SMFM in AJOG in 2021, address the unique complexity of two patients with separate care teams. The concept is sound.

What is not sound is the delivery mechanism. A paper checklist, however well designed, has structural limitations that no amount of lamination can fix.

Become a paying subscriber of ObGyn Intelligence to get access to this post and other subscriber-only content. It also includes an link to the interactive checklist.

Subscribe now

Read more

The Algorithm That Missed Preeclampsia

A documented failure in maternal risk prediction is a lesson every clinician and every pregnant patient needs to understand.*

The Technology

Commercial maternal early warning systems use algorithms to generate real-time risk scores for serious pregnancy complications including preeclampsia, hemorrhage, and sepsis. They aggregate vital signs, laboratory values, and clinical variables to produce a number that is supposed to tell the bedside team how worried to be. Many hospitals have adopted them enthusiastically, often on the strength of vendor claims and regulatory clearance rather than independent evidence.

What Happened

In 2019, researchers auditing a commercial maternal early warning algorithm found something that the hospital that deployed it did not know and the vendor had not disclosed. Risk scores for preeclampsia were systematically lower for Black patients than for white patients with identical vital signs and lab values -- the same blood pressure, the same proteinuria, the same clinical picture, but a lower algorithmic risk score if the patient was Black.

The algorithm had been trained on a dataset that underrepresented severe preeclampsia outcomes in Black women. The irony is brutal: Black women in the United States experience preeclampsia at higher rates and with more severe outcomes than white women. They are the population for whom accurate risk stratification matters most. They are the population the algorithm systematically underscored.

The algorithm was not malfunctioning. It was doing exactly what it had been designed to do. It had learned a pattern from its training data, and that pattern reflected the existing disparities in the care and documentation of preeclampsia across the hospital systems that contributed to the dataset. It reproduced those disparities at scale, automatically, at every hospital that deployed it, 24 hours a day.

The Women's Health Tech Report: Safety analysis, the evidence critique, and the verdict are below -- for subscribers who want the full picture.

Read more

Large language models -- AI tools that read, write, and reason with text -- are reshaping how clinicians and patients engage with medical information. Some are being marketed directly to women’s healthcare providers and patients with claims that deserve the same scrutiny we apply to any new clinical technology.

Before WHTR evaluates these tools, you need to know how to use the most widely known one. ChatGPT, made by OpenAI, is the most recognized AI tool in the world -- and like any clinical tool, the results depend entirely on how well you set it up. This guide walks you through exactly how to do it.

These steps apply whether you are an OB, MFM specialist, midwife, L&D nurse, gynecology nurse practitioner, or a patient trying to understand your own care.

Step 1. Create Your Account

Go to chat.openai.com and click Sign up. You can register with your email address or sign in with Google or Apple.

Once you are in, you land on the main chat screen. The text box at the bottom is where you type. Before you start chatting, spend ten minutes on setup. It will change how useful this tool is for you.

Step 2. Choose the Right Plan

ChatGPT offers a free plan, a Plus plan at $20 per month, and a Pro plan at $200 per month. Here is what matters for clinical and patient use:

Free plan: You can ask questions and have short conversations using GPT-4o, OpenAI’s main model. Memory and file uploads have limits. Workable for quick questions, but not reliable for document-heavy clinical work.

ChatGPT Plus ($20/month): This is the plan WHTR readers should use. You get priority access to GPT-4o, the ability to upload PDFs, web search, longer conversations, and full memory features. For reading research, reviewing guidelines, and drafting clinical content, Plus is the right choice.

ChatGPT Pro ($200/month): Designed for very heavy users. Includes access to more advanced reasoning models and higher usage limits. Most clinicians and patients will not need it.

For clinicians: Uploading a full ACOG Practice Bulletin, SMFM Consult, or NICE guideline and asking ChatGPT to analyze it against another document requires Plus. Free handles brief questions but not multi-page document review.

For patients: If you want to upload the consent form your doctor gave you and ask ChatGPT to explain what it means in plain language, you need Plus. Free can handle short, direct questions only.

The Women's Health Tech Report: Safety analysis, the evidence critique, and the verdict are below -- for subscribers who want the full picture.

The next six steps — Customize, Settings, Projects, Memory, document upload, and knowing Claude’s limits — are where I explain (for $ 0.16/day) how an AI tool becomes a clinical-grade assistant that saves you hours every week, which means this guide alone is worth more than your annual WHTR subscription.

Read more

Large language models or LLMs — AI tools that read, write, and reason with text — are now being used across medicine, including women's healthcare.

Some are being marketed directly to clinicians and patients with claims that deserve the same scrutiny we apply to any new technology.

Before WHTR evaluates these tools, you need to know how to use the most capable one available right now. Claude, made by Anthropic, is currently the best general-purpose LLM for clinical reading, evidence analysis, and medical writing — and like any clinical tool, it performs in direct proportion to how well you set it up. This guide shows you exactly how.

Claude is an AI assistant made by Anthropic. It is one of the most capable AI tools available for reading medical literature, analyzing guidelines, drafting documents, and thinking through complex clinical problems.

But like any clinical tool, the results depend heavily on setup. A few minutes of configuration makes an enormous difference. This guide walks you through exactly how to do it -- no technical background needed.

These steps apply whether you are an OB, MFM specialist, midwife, L&D nurse, gynecology nurse practitioner, or a patient trying to understand your own care.

Step 1. Create Your Account

Go to claude.ai and click Sign up. You can register with your email address or sign in with Google.

Once you are in, you will see a text box at the bottom of the screen. That is where you type. Everything else builds from there.

What follows is the part most users never find: the six setup steps that separate a generic chatbot from a tool that knows your specialty, reads your guidelines, challenges weak evidence, and drafts documents at a level that would take you an hour to produce yourself — and for clinicians and patients navigating women’s healthcare, that is not a convenience, it is a competitive and clinical advantage worth far more than the cost of membership.

Subscribe now

Step 2. Choose the Right Plan

Claude offers a free plan and a paid plan called Claude Pro, which costs $20 per month. Here is what matters for clinical and patient use:

Free plan: Good for trying it out. You can ask questions and have short conversations. However, you cannot reliably upload PDFs, and longer conversations get cut off. Not ideal for clinical work.

Claude Pro ($20/month): This is the plan WHTR readers should use. You get priority access, much longer conversations, and the ability to upload full PDF documents -- journal articles, clinical guidelines, patient information sheets, discharge summaries. If you plan to read research or draft clinical content, Pro is worth every dollar.

For clinicians: Uploading a full ACOG Practice Bulletin, SMFM Consult, or NICE guideline and asking Claude to compare it against another guideline is only possible on Pro. Free will not handle documents of that size.

For patients: If you want to upload the consent form your doctor gave you and ask Claude to explain what it means in plain language, you need Pro. Free can handle short questions but not document review.

The next six steps — Customize, Settings, Projects, Memory, document upload, and knowing Claude's limits — are where I explain (for $ 0.16/day) how an AI tool becomes a clinical-grade assistant that saves you hours every week, which means this guide alone is worth more than your annual WHTR subscription.

The Women's Health Tech Report: Safety analysis, the evidence critique, and the verdict are below -- for subscribers who want the full picture.

Read more

Summary

Multi-cancer early detection (MCED) tests are blood tests that look for signs of cancer in the bloodstream. Right now, about 40 of these tests are either for sale or being developed.

They work by detecting tiny fragments of DNA that cancer cells shed into the blood. The Galleri test, made by a company called Grail, is one of the most studied. It just completed the largest randomized trial of any MCED test ever run, enrolling about 142,000 people through the UK’s National Health Service.

The trial tested whether annual Galleri screening could reduce the number of late-stage cancer diagnoses. It did not meet its main goal.

For gynecologic cancers specifically, existing tests already screen for cervical and breast cancer, but there is no good screening test for ovarian cancer, one of the deadliest cancers in women. MCED tests are being watched closely for this reason.

What It Means

For women’s health, the promise of MCED tests is most meaningful for cancers we currently cannot screen for well. Ovarian cancer is the clearest example. It is usually found late, when it is much harder to treat, because there is no reliable, widely recommended screening test. MCED tests could, in theory, change that. But the current data are sobering.

These tests correctly identify cancer in only 30 to 80 percent of people who actually have it. And in one Galleri study, nearly half of people who got a positive result did not actually have cancer. False positives in gynecologic cancer screening are not a minor inconvenience. They lead to imaging, surgery, and serious anxiety. For cervical cancer, where we already have a highly effective screening system, MCED tests are not more sensitive than existing tools and could give false reassurance if women skip their Pap smears. Whether catching cancer earlier with these tests actually saves lives has not been proven yet.

My Take

I understand why a Super Bowl ad showing a man relieved by a negative cancer test is appealing. Fear of cancer is real and universal. But the science is not there yet, and in gynecologic oncology, the stakes of getting this wrong are high. The Galleri trial just failed its primary endpoint. In the previous large Galleri study, 38 out of 92 positive tests were real cancers. That means 59 women out of every 92 with a positive result did not have cancer. In a gynecologic context, those women may have gone on to have unnecessary imaging, biopsies, or surgery. Ovarian cancer is the one area where MCED tests hold genuine promise because we have nothing better. But that promise requires rigorous trial data showing actual survival benefit, not just earlier detection. Early detection of a cancer that still kills the patient is not a win. I will be watching the full NHS-Galleri results closely. Until we see them, I would not recommend these tests outside a clinical trial.

Share

Introductory Issue — Free to All Readers

Why I'm Writing This

For a year, ObGyn Intelligence has been cutting through the noise in women's health. Now it's time to do the same for the technology that is changing it.

About a year ago, I started ObGyn Intelligence -- a newsletter built on a simple premise: evidence matters in women's health, and there is far too much noise drowning it out. Since then, the response has been more than I expected. Clinicians, researchers, patients, and educators have found it and kept coming back. The newsletter has grown into something I am genuinely proud of, and the conversations it has started -- in comments, in emails, in clinical hallways -- have confirmed what I suspected when I launched it: there is a real hunger for honest, evidence-based analysis of women's health that does not talk down to its readers and does not pull its punches.

Over the past several months, many of you have asked the same question in different forms.

• What do I think about AI in obstetrics and in gynecology. In Women’s Health?

• What about the fertility apps patients keep bringing up in appointments?

• What about the fetal monitoring algorithms, the maternal early warning systems, the chatbots patients are using to interpret their own lab results?

• What about patient safety issues?

Is this technology actually helping, or is it moving faster than the evidence can keep up with?

Those are exactly the right questions. And they deserve their own dedicated space.

This is that space.

What The Women's Health Tech Report Is

The Women's Health Tech Report publishes two to three times per week as a dedicated series within ObGyn Intelligence. Each issue covers technology in depth -- one honest assessment of what the evidence actually shows, what the tool actually does, and where it fails. No press releases. No vendor language. No cheerleading for technology that has not earned it.

The technologies covered here span the full landscape: artificial intelligence, clinical algorithms, diagnostic software, consumer health apps, wearable devices, fertility technology, and anything else that touches women's health in the clinic or at home. Some of it is genuinely transformative. Some of it is harmful. Most of it sits somewhere in between, deployed with confidence that the evidence does not yet support.

That gap is the reason this series exists.

Why This Matters Now

Health technology is entering women's care at a pace the regulatory and clinical evaluation systems were not built to handle. The FDA clears AI-enabled medical devices through pathways designed for traditional hardware. Peer review of clinical algorithms is inconsistent and often proprietary. Consumer health apps operate almost entirely outside any clinical evidence standard. And patients are encountering all of it -- in their clinics, on their phones, in their hospital rooms -- without anyone explaining what it does and does not do.

The result is a landscape where a hospital can deploy a maternal early warning algorithm without a published bias audit. Where a patient can receive an AI-generated fetal risk score without knowing how that score was calculated or validated. Where a fertility app can claim clinical accuracy without a single peer-reviewed study.

This is not theoretical. These things are happening now, in clinical settings, to real patients. The consequences range from missed diagnoses to unnecessary interventions to the perpetuation of existing disparities in care. Women deserve better than that.

Who I Am and Why I'm the Right Person to Write This

I have practiced obstetrics and maternal-fetal medicine for more than 50 years. I have watched technology enter clinical medicine in waves -- the introduction of electronic fetal monitoring, the ultrasound revolution, the genomics era, and now the AI moment.

I am the first ObGyn to publish on ChatGPT

Each wave brought genuine advances and genuine harms, often arriving together and often indistinguishable from each other until the evidence caught up.

I am an active researcher using the CDC natality database, one of the most comprehensive sources of population-level birth outcome data available. I am a peer reviewer for major obstetric journals. I co-developed the preventive ethics framework for obstetric practice. I have spent decades studying the gap between what clinical trials show and what actually happens to patients at scale.

I am not neutral about bad evidence. I am not impressed by a well-designed press release. I am not persuaded by the argument that a technology must be good because it is new. And I have seen enough catastrophic outcomes in obstetric care to know that a tool that performs well in a controlled validation study can fail the patient in room 4 in ways the study never measured and the developer never anticipated.

—→ My job in this series is to tell you what the technology actually does -- not what its developers say it does, not what the hospital administrator who approved the purchase believes it does, not what the algorithm says about itself.

Who This Is For

For ObGyns and other Clinicians

This series is for clinicians who want to evaluate health technology with the same rigor they apply to drug trials. A new medication requires demonstrated efficacy, safety data, dose-response characterization, and post-market surveillance. A clinical algorithm that drives the same decisions should meet the same standard. Most do not. This series will tell you which ones fall short and what to look for.

For Patients and anyone else interested in Tech

It is for patients who are encountering AI-generated health information, fertility apps, wearable monitors, and chatbot medical advice and deserve an honest assessment of what to trust and what to question. The information asymmetry between what technology developers know about their products and what patients know is not acceptable when the stakes are maternal and fetal health.

It is for anyone who believes that evidence matters in women's health -- because in this space, the gap between what technology promises and what it delivers is not a minor inconvenience. It is a patient safety issue.

What Free Readers Get, What Paid Subscribers Get

Every issue covers one technology. Free readers receive the technology description and the clinical context -- what the tool is, what it is supposed to do, and how it is being used. Paid subscribers receive the safety analysis, the evidence critique, and the bottom line. That is where the work is, and it is what the subscription supports.

If you are already a paying subscriber to ObGyn Intelligence, you are already in. This series is included.

What Comes Next

The first issues focus on artificial intelligence because AI is the dominant story in health technology right now and because the stakes are highest where adoption is fastest. Issue 1 examines AI interpretation of fetal heart rate tracings in labor. Issue 2 examines a documented algorithmic failure in maternal risk prediction. Issue 3 examines how patients are using AI chatbots in pregnancy and what clinicians need to know about it.

From there the series expands to cover the full landscape: consumer fertility technology, wearable monitoring devices, genetic testing platforms, clinical decision support tools, digital therapeutics, and whatever else arrives that is being deployed in women's health faster than the evidence can evaluate it.

If you are a clinician using a technology in your practice that deserves scrutiny, write to me. If you are a patient who has encountered a health technology that raised questions about accuracy, safety, or transparency, write to me. The most important evidence in this field often starts in the clinic, not the journal.

I am glad you are here.

Amos Grunebaum, MD
The Women's Health Tech Report | ObGyn Intelligence | obmd.com

Subscribe now

Why I am here

Women’s health technology is moving fast. Too fast for most clinicians to evaluate, and too fast for most patients to navigate alone. This report publishes two to three times per week -- not to celebrate every new tool, but to ask the question nobody else in this space is asking: does the evidence actually support what this technology claims to do?

If something in this issue changed how you think about a tool, a system, or a conversation with a patient -- forward it to one colleague who needs to read it. That is how this reaches the people it is meant for.

Already a subscriber? Thank you. If someone sent this to you, you can subscribe free at obmd.com.

Subscribe now

Share

Patients are already using chatbots and LLMS to understand their pregnancies. Here is how to guide them -- and exactly where the line is.

The Technology

Large language model chatbots are being used by pregnant patients to look up symptoms, interpret lab results, understand ultrasound findings, research medications, and prepare questions for appointments. A 2024 survey found that more than 40% of pregnant respondents had used a chatbot for health information at least once during their pregnancy. [CITATION NEEDED] The number is almost certainly higher among younger patients and will continue to rise. These tools are part of your patients’ prenatal experience whether you have addressed them in your practice or not.

The simplest way to remember

• AI = the entire field of intelligent machines (medicine as a profession)

• LLM = the language-reasoning engine inside AI (a physician’s clinical reasoning ability)

• Chatbot = how you interact with the LLM (the clinic visit where you interact)

• Claude = one specific LLM (and its chatbot product: one particular doctor)

This distinction matters because when an answer is wrong, the failure may come from the reasoning model, the interface behavior, or the way the question was asked.

The Clinical Application

Used well, AI chatbots can genuinely improve the prenatal experience for patients who have access to them. The information gap between what pregnant patients want to understand and what clinical visits have time to cover is real and well documented. A 15-minute prenatal appointment cannot answer every question a patient has accumulated in the three weeks since her last visit. Patients who leave appointments with unanswered questions turn to Google, to social media, to parenting forums, and increasingly to AI chatbots. Given those alternatives, a well-used chatbot is not the worst option.

A patient who does not understand why her doctor ordered a glucose tolerance test at 24 weeks can ask an AI to explain it in plain language before her next appointment. A patient who receives a report showing an NT measurement of 2.8mm can ask what that number means, what additional testing might be recommended, and what questions to raise at her next visit. A patient who is confused about why she is being recommended for a repeat cesarean can use an AI to understand the clinical reasoning before she decides whether to ask about alternatives.

This is the use case that works: AI as a tool for preparation and comprehension, helping patients arrive at clinical encounters more informed and better equipped to participate in shared decision-making. That is a good outcome from an imperfect tool.

The Women's Health Tech Report : Safety analysis, the evidence critique, and the verdict are below -- for subscribers who want the full picture.

Read more