Hospitals and insurers that cut costs end up cutting out patients too. When “efficiency” becomes the only metric, anything that slows the system (complexity, ambiguity, difference) is branded as waste. In American healthcare, patients pay for that label with repeated tests, missed diagnoses and denied care.
Nobody Is Too Expensive states the general form: once an insurer or hospital labels a group “too expensive,” friction pushes those people out of care without anyone saying no. In medicine the label is written in billing codes, and the friction is a referral loop.
Fifteen-minute visits. Specialists who treat only one body part at a time. Billing codes that reward neat closures over open-ended inquiry. If your illness doesn’t fit a neat checkbox, you simply vanish from the data.
View source
%% title: The diagnostic referral treadmill
%% caption: Single-organ clinics run isolated tests, bill specific codes, and close charts; multi-system illness that fails to fit clean categories is recoded as patient anxiety.
flowchart TD
classDef actor stroke-width:1.6px;
classDef system stroke-width:1.2px;
classDef gate stroke-width:1.6px;
classDef binding stroke-width:3px,font-weight:bold;
classDef status stroke-dasharray:4 3;
classDef load stroke-width:2px;
classDef repair stroke-width:2.4px,font-weight:bold;
classDef failure stroke-width:2.2px,stroke-dasharray:6 3;
P([Complex chronic patient]):::actor --> C1["Clinic 1: Pulmonology"]:::system
C1 -- "labs normal / chart closed" --> C2["Clinic 2: Cardiology"]:::system
C2 -- "labs normal / chart closed" --> C3["Clinic 3: Neurology"]:::system
C3 -- "labs normal / chart closed" --> S{"Categorical cut"}:::gate
S == "no single-organ code" ==> X([Recoded as 'psychosomatic / anxiety']):::failure
X --> D([Care denied / patient abandoned]):::failure
Where the Chart Ends
Take Long COVID. You wake breathless, heart racing, brain clouded, and are shuttled from pulmonologist to cardiologist to neurologist to psychiatrist. Every clinic runs its own tests, bills its own codes, and closes its chart. When each result lands “normal,” your record changes: your plea for help becomes “anxiety.” No one was paid to see the whole story, so the system recodes you as the problem.
That recoding is a blueprint more than a rare glitch. As the section on actuarial medicine below traces, our billing infrastructure sacrifices interpretive labor, pushing it, unpaid, onto those least able to carry it. What can’t be slotted into a clean code is either erased or pathologized.
Rare disorders fare no better. Ehlers–Danlos patients chase a genetic label for years, only to learn the multidisciplinary clinics they need are deemed “experimental” and excluded from coverage. ME/CFS sufferers, lacking a definitive biomarker, are sent into graded-exercise regimens that deepen their collapse, then denied disability. Without quantifiable proof, lived suffering becomes suspect.
Trans bodies face erasure by design. Electronic records force a “male” or “female” checkbox, so a trans man can’t book a Pap smear, and a trans woman can’t schedule a prostate exam. Gender-affirming hormones are coded “cosmetic,” triggering appeals that can outlast the patient. If your anatomy can’t be checked off, your care can’t be guaranteed.
Structural bias deepens the wound. Clinical algorithms (pain scales, kidney-function formulas, fetal-monitoring thresholds) rely on datasets calibrated on white men. When a Black mother’s labs stray, the protocols hold their shape and she disappears into statistical noise, helping explain why Black women die of pregnancy-related complications three to four times more often than comparable white women.
“Universal” care, it turns out, defaults to whiteness.
The Same Logic Outside the Hospital
These failures extend far beyond hospitals. In February 2021, Texas’s power grid, designed for summer profits, froze under Arctic winds. Engineers warned that pipelines needed insulation and turbines heaters, but winterization didn’t pencil out; when pipes burst and hospitals lost water and grid power, dividends still cleared.
In 2023, Hawaiian Electric had no procedure for shutting off power ahead of high winds and kept its lines live through drought-parched brush; Lahaina burned. Lawsuits followed, and in July 2024 the utility adopted its first public safety power shutoff plan.
Even aviation, the supposed gold standard of redundancy, succumbed when Boeing let the 737 MAX’s MCAS rely on a single angle-of-attack sensor and sold the disagreement alert as an extra. After two crashes and 346 deaths, only then did they retrofit safety features, solemnly branded “innovation.”
View source
%% title: The cycle of neglect and retrofit %% caption: Companies skip safety work that does not pay, fix it only after a disaster, then present the fix as innovation. flowchart TD classDef actor stroke-width:1.6px; classDef system stroke-width:1.2px; classDef gate stroke-width:1.6px; classDef binding stroke-width:3px,font-weight:bold; classDef status stroke-dasharray:4 3; classDef load stroke-width:2px; classDef repair stroke-width:2.4px,font-weight:bold; classDef failure stroke-width:2.2px,stroke-dasharray:6 3; N["1. Neglect & extraction<br/>Winterization doesn't pencil out"]:::status --> C["2. Catastrophic collapse<br/>Grid freezes, wildfire, crash"]:::failure C --> R["3. Emergency retrofit<br/>Litigation & public bailouts"]:::repair R --> B(["4. Branded 'innovation'<br/>Earlier warnings drop from the record"]):::binding B -. "cycle repeats" .-> N
This cycle of neglect, collapse, retrofit, rebrand is not a bug; it’s a feature. In these cases the companies acted on criticism only after disaster struck, and the fix arrived as a new plan or a new feature.
Actuarial medicine: exclusion without denial
The same logic runs earlier, before the crash, inside the gatekeeping itself. Hospitals increasingly rely on predictive analytics and risk-scoring software to assess whether a patient is “worth admitting.” The tools scan medical histories, insurance status, age, zip codes, and socioeconomic indicators to assign a risk score. If flagged as high-cost, the patient never receives an outright denial — instead they encounter layers of subtle barriers: prior-authorization friction, re-verification loops, “under review” states that induce attrition without a formal no. The system can report that it denied nothing while the complex patient disappeared.
That is actuarial medicine: the patient’s financial profile becomes the prognosis, and the “efficiency” of the filter is the hospital’s, never the patient’s. Anticipatory denial is the medical cut of “too expensive” — what the chart cannot code, the algorithm pre-emptively refuses to admit.
Remote-work tools championed by neurodivergent and caregiving professionals became “essential infrastructure” only after a pandemic threatened profits. Live-captioning won by Deaf activism reappeared as a “new feature” when gyms and bars discovered they liked muted TVs.
That process of forgetting and rebranding is what I call Institutions Prefer Crises to Early Warnings.
What if, instead of treating these edge solutions as extras, we baked them in from the start?
- Whole-person clinics funded like trauma bays.
- Insulated pipelines and redundant sensors counted as essential.
- Slack in scheduling recognized as life-saving capacity.
- Genuine trust in patient testimony honored as data.
These go beyond boutique R&D: they are seat belts, fire codes, and stable grid design for society.
Yes, that costs more up front. But the true invoice for neglect arrives later: lawsuits, funerals, and emergency interventions. Shuffling those receipts into another quarter and calling it “efficiency” is an accounting trick.
What looks like waste on this year’s budget is often what keeps the grid, the plane and the clinic running later.
Edge-case patients are where the system’s failures show up first. Build a system that works for them, and it will work better for everyone.
Refusing means deciding, knowingly, that sorting patients by cost is the best care they will get.