SERIES: THE BILL NO ONE WANTS TO PAY • PART 2 OF 5: Follow the Money

Administrative bloat, the AI false promise, nonprofit fictions, and why physicians who left are not the problem: The Modern Heretic

In Part One, I established the foundation: a reimbursement system that has structurally undervalued care delivered to women’s bodies for thirty years, a burnout crisis disproportionately breaking the clinicians who serve them, and a physician workforce actively hemorrhaging at the exact moment demand for specialized women’s health care is peaking.

Now let’s follow the money. Because somewhere between the patient who needs care and the physician who trained to provide it, an extraordinary amount of resources disappears. And it is not going where you think.

About those administrative costs. Let’s talk.

In 2023, hospital administrative costs reached $687 billion compared to $346 billion in direct patient care. This is a ratio of approximately 2:1. Administrative expenditures grew 87.2% between 2011 and 2023, outpacing direct patient care spending growth of 75.4%.

For every dollar that reaches a patient, nearly two dollars disappears into administration.

And the trend is worsening, not improving. Administrative costs now account for more than 40% of total hospital expenses in delivering care to patients. On average, about 6.5% of total employment in a hospital is devoted to negotiating with and attempting to collect reimbursement from insurance companies on care it has already delivered. That is an army of people whose entire job is fighting with insurance companies. In my previous practice, we had approximately 40 fulltime non clinician employees; at least 4 of them were devoted to these exact tasks. In 2025 alone, hospitals spent nearly $18 billion overturning claims denials. They spent $43 billion total trying to collect payments.

And what is the patient getting from all of this spending? In traditional care settings, most primary care physicians manage over 2,000 patients and see 20 to 30 per day - leaving just minutes per visit. Minutes. For a menopausal woman with vasomotor symptoms, sleep disruption, cognitive changes, mood dysregulation, sexual dysfunction, joint pain, and a question about whether HRT is right for her specific history. You cannot deliver that care in eight minutes. You cannot build a therapeutic relationship in eight minutes. You cannot practice medicine that changes the trajectory of a woman’s health in her most consequential decade in eight minutes. The system does not care.

A word about the algorithm that’s coming to save us all.

Because right now, somewhere in a boardroom adjacent to every broken system I’ve just described, someone is giving a presentation about artificial intelligence. The slides are beautiful. The ROI projections are compelling. And the pitch is seductive: AI will close the physician gap. AI will streamline care. AI will solve the access problem that burnout and reimbursement failures created.

Let’s be precise about what AI is actually doing in American healthcare right now, because the evidence is already in, and it is not a story about salvation.

A class action lawsuit against UnitedHealth Group alleged that its AI algorithm, nH Predict, was used to deny post-acute care coverage to Medicare Advantage patients, overriding the judgments of treating physicians. When those denials were appealed before federal administrative law judges, approximately 90% were reversed. Ninety percent. Between 2019 and 2022, UnitedHealthcare’s denial rate for skilled nursing facility admissions jumped ninefold, from 1.4% to 12.6% in the first full year after the AI system began managing those decisions. The algorithm wasn’t saving anyone. It was saving the insurer money, at scale, faster than any human could manage.

And this is not one company’s aberration. A Senate Permanent Subcommittee on Investigations report documented that UnitedHealthcare’s prior authorization denial rate for post-acute care more than doubled between 2020 and 2022 after it implemented algorithms to automate its review process. Cigna and Humana faced similar lawsuits in the same period. Newer AI models don’t just flag claims for denial, they learn and then predict which patients are likely to fight back, creating a feedback loop that rewards algorithmic aggression: the patient gives up, the insurer saves money, and the model learns that this type of denial is “successful” and replicates it.

This is AI as it is currently deployed in the business of American medicine. Not to heal. To deny. At scale. Invisibly.

But here is where it gets specifically dangerous for the women we are trying to serve. AI clinical decision support tools rely on pattern recognition drawn from historical patient data. This data is profoundly skewed toward male presentations of disease. In cardiology, for example, AI systems trained predominantly on male datasets may not weight the symptoms more common in women - fatigue, nausea, jaw pain - appropriately, leading to missed or delayed diagnoses. Without deliberate action to address these existing gaps, AI will not correct them. It will encode them into its models and automate their reproduction at scale.

A 2025 study in Oxford Open Digital Health analyzing approximately 400 published clinical AI and machine learning models found that the majority failed to adequately disclose the racial and gender composition of their training data. This is the same pattern of exclusion that has defined drug and device development for decades. We already know what happens when you train a medical system on male bodies and apply it to women. We have years of evidence. We called it standard of care.

Male-only clinical trials already outnumber female-only trials by 67%. Those evidence gaps ripple through clinical guidance and directly into AI development. An algorithm trained on that literature doesn’t just reflect the bias, it launders it. It transforms decades of neglect into a number, a score, a recommendation. And it does so at a speed and scale that makes human correction increasingly difficult.

The menopause space is particularly vulnerable. There is no condition in medicine more shaped by the historical dismissal of women’s self-reported experience. Vasomotor symptoms were called psychosomatic for decades. Cognitive changes were attributed to anxiety. Sexual dysfunction was treated as a relationship problem. When we build AI clinical decision tools that draw on that history, learning from charts written by physicians who were trained to minimize, normalize, and under-treat, we are not building intelligence. We are building a very fast, very expensive amplifier for everything that was already wrong.

I want to be clear: I am not anti-technology. I use it. I believe in its potential. There are real applications of AI in medicine that are genuinely promising: in radiology, in genomics, in reducing administrative burden for clinicians, in expanding access for underserved populations who currently have no care at all. These are worth pursuing with rigor and intentionality.

But AI deployed as a cost-cutting mechanism by insurers, or as a workforce substitution strategy by health systems trying to paper over a physician shortage they helped create, is not innovation. And for women whose symptoms have been dismissed, whose data was excluded, whose care has been systematically undervalued, handing the decision-making to a model trained on that history is not a solution. It is an acceleration of the original injury.

The question is not whether AI will transform healthcare. It will. The question is: who is it being built for, who controls it, and whose bodies does it know?

Until we can answer those questions with something better than “mostly men, mostly white, mostly insurers, and not enough women”, we should be very cautious about calling it progress.

The elephant in the room: “nonprofit” hospitals that pay their executives millions.

I want to name something else rarely gets named in polite spaces: the concept of the “nonprofit” health system that nonetheless generates massive executive compensation while simultaneously squeezing physician reimbursement and eliminating services that don’t profit.

The top-heavy administrative structure of American health systems is not a secret. It is a choice. And it is a choice made over and over again at the expense of the clinicians delivering care and the patients receiving it. When women’s health services - less remunerative by design, as we’ve just documented - are the first to be cut, reduced, or consolidated in the interest of system efficiency, it is not an accident. It is a predictable outcome of a payment system that has never valued care delivered to women’s bodies.

The “nonprofit” designation was intended to serve communities. When it instead serves to obscure profit extraction while defunding the lowest-margin services - the ones disproportionately used by women - that is not a market failure. That is a values failure. And it deserves to be called what it is

And what about those of us who left.

There is a persistent narrative - at times quietly implied, increasingly wielded like a blunt instrument - that physicians who have moved to direct pay, membership, or cash-pay models are abandoning patients. That we are in it for the money. That we are choosing affluence over access.

I reject that framing entirely.

Growing numbers of physicians are turning to concierge and membership-based medicine not out of greed but out of the recognition that the traditional insurance-driven system is unsustainable. Endless administrative tasks, prior authorizations, shrinking reimbursement, and EHR documentation have made running a financially viable medical practice impossible and taken priority over meaningful patient care.

The proportion of physicians in private practice fell to 42.2% in 2024. Only 36% of practices are physician owned, down from 70% in 2018. The independent physician is nearly extinct. This is not because doctors don’t want autonomy, but because the system has made it more and more challenging to provide excellent, time-intensive care under standard fee-for-service while keeping the lights on.

Those of us who have built practices outside that system are not the problem. We are a symptom of the problem. But perhaps more importantly, we are a proof of concept. Patients will pay for time, relationship, and expertise when the alternative is a rushed visit with a burned-out clinician who has forty more patients to see before 5pm.

Patients are choosing direct care because it offers what the conventional system cannot: same-day access, longer visits, and direct communication - transparency they value, especially when weighed against high-deductible plans that don’t deliver meaningful access.

The criticism should not be directed at the physicians who left. It should be directed at the system that made leaving the ethical choice.

So what do we actually need?

Those fighting for menopause visibility - and I am proudly among them - must be willing to expand the frame. Yes, train the clinicians. Yes, fund the research. Yes, mandate menopause education in medical schools. Do all of it.

And then fight like hell for the structural conditions that allow that training to translate into care:

• Reimbursement parity. CPT codes for gynecologic services must be revalued. The 30-year disparity in work RVUs between female-specific and male-specific procedures is not a rounding error. It is sex-based discrimination embedded in the financial architecture of American medicine.

• Payment models that reward time and relationship. A prevention-focused, relationship-based model of menopause care cannot exist in a volume-driven, diagnosis-code-dependent system. The move toward value-based care is the right direction. It must accelerate.

• Transparency about administrative waste. Nearly two dollars of administration for every dollar of direct patient care is not an inevitable feature of a complex system. It is a policy failure, and it is eating the system alive.

• Oversight of AI in clinical and coverage decision-making. Any algorithm that makes or informs decisions about women’s health must be trained on representative data, audited for gender and racial bias, and subject to meaningful human review. Speed and scale are not substitutes for accuracy and equity.

• Accountability for nonprofit health systems that defund women’s health services. The tax-exempt status that subsidizes these organizations should come with genuine obligations to the communities - and the patients - they claim to serve.

• Respect for clinicians who are voting with their feet. When experienced, trained physicians leave high-volume employment to practice the kind of medicine they entered the field to deliver, that is information. Policy makers should be asking what drove them out, not blaming them for going.

The women in the middle of perimenopause and menopause right now - the ones suffering, the ones being dismissed, the ones googling symptoms at 3am - deserve a healthcare system rebuilt around their needs. That means confronting not just the knowledge gap, but the economic architecture that determines who gets care, for how long, at what depth, and by whom.

We can train all the doctors we want. If we don’t fix the money, we fix nothing.

Dr. Suzanne Gilberg is a board-certified OB/GYN, integrative medicine physician, Chief Clinical Officer at MonarchMD, and founder of the Menopause Boot Camp. She is the author of Menopause Bootcamp and the forthcoming Plant Medicine. Find her at @askdrsuzanne and @themenopausebootcamp.Paid subscriptions are coming. Reply to this email if you want founding member pricing.