⚠️ The Cost Cascade — What Prior Authorization Actually Produces
The AMA has documented what happens before a denial is even issued: prior authorization requirements introduce delays that themselves produce harm — independent of whether the eventual decision is yes or no. The delay is the first injury. The denial, if it comes, is the second. Most data only counts the second.
What those delays and denials produce — across the patient population — is a predictable cascade. Each outcome represents a cost the insurer chose not to absorb, transferred instead onto someone else's body, budget, or future.
Bankruptcy
When care is delayed long enough that patients seek it out-of-pocket — or when the appeals process consumes years of time, energy, and legal cost — the financial damage outlasts the medical one. Medical debt is the leading cause of personal bankruptcy in the United States. Prior authorization delays are a documented contributor to that pipeline.
Care Abandonment
When the denial blocks care long enough, patients stop trying to access it. The insurer pays nothing. The patient still needs treatment. The condition progresses untreated into the void — where it becomes someone else's emergency, someone else's cost, or simply a life shortened without documentation. Care abandonment is the outcome the system's data never captures because no claim is ever filed.
Disability
Patients who survive the delays often do so at a cost paid in function. The condition the insurer refused to treat in its early, manageable stage becomes the chronic condition that limits what the patient can do, earn, and sustain. The insurance company avoided the treatment cost. The patient absorbed it in their body — and carries it forward, often permanently.
Dismemberment
A patient who is dismembered by a delay is catastrophically more expensive — at every downstream level — than a patient with the same underlying condition who still has all four limbs. Amputation generates cascading costs: revision surgeries, prosthetics, rehabilitation, wound care, infection risk, anticoagulation management, lost function, lost income, and lifetime adaptive equipment. Recovery is always easier with all four limbs intact. Preventing an amputation is always cheaper than managing life without the limb.
The $55,000 stenting procedure UnitedHealthcare denied in 2017 produced over $1.1 million in taxpayer-funded downstream costs and an above-knee amputation. The insurer paid none of it. The patient paid with his leg. The public paid with their taxes.
Death
The terminal outcome of the cascade. Kathleen Valentini waited 41 days for an MRI her surgeon ordered. Her cancer killed her. Deron Wells. Rhett Pascual. Ryan Matlock. The Reason Room documents these cases not as tragedies in isolation, but as the predictable endpoint of a system that treats delay as a financial instrument.
The Externalized Cost Framework — Who Actually Pays
The insurer externalizes every cost — onto the patient's body, the taxpayer, the family forced to absorb the loss of capability, or the void of care abandonment — and records it as a business decision. None of these appear on the insurer's balance sheet as a loss. All of them are.
The Patient's Body
Absorbs the delay in tissue, bone, nerve, and years. The body is the ledger the insurer never has to balance.
The Taxpayer
Picks up the catastrophic care bill the insurer avoided. Medicaid, Medicare, emergency services, public disability programs — the avoided cost doesn't disappear. It transfers.
The Family
Absorbs the loss of capability that policy language never counts. When a parent can no longer run after a child in a parking lot, that loss doesn't appear in any claims database. The family reconfigures around what the denial took — quietly, permanently, without documentation.
The Void
Care abandonment. The insurer pays nothing. The patient still needs treatment. One in three GoFundMe campaigns is medical in nature. That is not a fundraising statistic — it is a measure of system failure expressed in the language of strangers asking strangers to cover bills that insurance was already paid to provide.
The Fiscal Indictment
The government is subsidizing insurer profit through SSDI, Medicare catastrophic costs, and destroyed workforce productivity.
Privatized profit. Socialized loss.
See the human cost of this cascade → Room II
These are not hypothetical outcomes. The Reason Room documents each one by name — the people the mechanism ran through.
The Reason Room — Documented Cases →Legal Challenges & Lawsuits
Ballad Health Sues UnitedHealth Group, Alleging Medicare Advantage Manipulation and Harm to Patients
Ballad Health filed a federal lawsuit against UnitedHealth Group, alleging the insurer systematically denies, delays, or underpays for medically necessary care for Medicare Advantage patients. The first time Ballad Health has taken legal action against an insurer — called a "last resort" after years of failed resolution attempts.
- The lawsuit alleges UnitedHealth improperly uses a "blatantly flawed" AI model to override physicians and prematurely deny post-acute care coverage, leaving patients hospitalized longer than necessary and increasing their risk of hospital-acquired conditions.
- The suit claims UnitedHealth withheld "millions of dollars" it had previously agreed to pay for patient care, and alleges a pattern of overstating how sick patients are to collect higher government payments while simultaneously denying claims.
Estate of Gene B. Lokken et al. v. UnitedHealth Group, Inc. et al.
A federal class-action lawsuit filed by estates of deceased Medicare Advantage patients alleges UnitedHealth Group illegally used an AI model to deny post-acute care to elderly patients. A federal judge allowed the case to proceed on breach of contract and breach of implied covenant of good faith and fair dealing.
- The AI model, "nH Predict," is alleged to be "notoriously inaccurate" with an error rate over 90%. The suit claims UHG knew of these flaws but used the algorithm to override physicians' recommendations, leading to "worsening injury, illness, or death."
- UHG knew only about 0.2% of policyholders would appeal the AI-powered denials. When denials were appealed, over 90% were reversed — demonstrating the model's blatant inaccuracy.
- The suit claims UHG employees were disciplined or terminated if they deviated from the AI's "rigid and unrealistic" predictions, regardless of the patient's actual medical needs.
Judge Rips Insurance Company for 'Immoral, Barbaric' Cancer Denials
A federal judge recused himself from a case where UnitedHealthcare denied coverage for proton beam therapy to a child with cancer — revealing he himself had been diagnosed with cancer and was initially denied the same treatment by his own insurer.
- Judge Robert W. Schroeder III was diagnosed with base-of-tongue cancer in 2018. His insurer denied proton beam therapy initially; it was later approved after appeal.
- The case he recused himself from involved A.F., a child with Ewing sarcoma, whose UnitedHealthcare plan denied the same therapy. The judge found the parallel too close to remain impartial.
Read the CNN Article → Read the National Library of Medicine Article →
Senate Democrats Release Scathing Report on Medicare Advantage Denials
A Senate investigation into the three largest Medicare Advantage insurers (UnitedHealthcare, Humana, and CVS/Aetna) found the companies deny prior authorization requests for post-acute care at far higher rates than other care categories.
- In 2022, Humana denied post-acute care prior authorization requests at a rate more than 16 times higher than its overall denial rate for all PA types.
- UnitedHealthcare's post-acute care PA denial rate more than doubled — from 10.9% in 2020 to 22.7% in 2022.
- CVS and UnitedHealthcare both denied post-acute care requests at rates roughly three times higher than their overall denial rates.
- The number of CVS post-acute care requests requiring prior authorization increased 57.5% from 2019 to 2022.
The enforcement mechanism these lawsuits are missing → Room III
These cases proceed hospital by hospital, estate by estate. S.3829 creates the criminal enforcement framework that addresses the pattern — not just the individual case.
🤖 AI as Tool — Double Featured
The clinical AI and counter-tool section appears here under Damages because it documents the resistance — what the medical community is building in response to the harm. The same technology's weaponization is documented under Scienter as evidence of what UHC knew.
AI as a Tool: What It Actually Does When Used Correctly
The same underlying AI technology, when directed toward patient benefit rather than denial optimization, demonstrates measurably superior medical knowledge and significantly improves physician accuracy. The technology is not the problem. The intent is.
- A peer-reviewed cross-sectional study published in Bioengineering (PMC, 2025) found that AI outperforms physicians in general medical knowledge across most domains — demonstrating that AI applied to diagnosis and clinical support improves patient outcomes rather than obstructing them.
- An AMA survey found that more than 80% of physicians now use AI professionally — with adoption concentrated in administrative tasks, documentation, and clinical decision support. Physicians report improved efficiency and reduced documentation burden.
- Physicians using AI scribes — documented at institutions including Johns Hopkins and Oak Street Health — show measurably greater engagement with patient complexity, spending more time on diagnosis and less time on paperwork. Patients with complex, multi-system conditions report feeling genuinely seen for the first time.
- The critical distinction: AI in clinical support is calibrated toward diagnostic accuracy. AI in denial systems like nH Predict was calibrated toward cost reduction. The programming intention is the difference — and it is auditable.
The Argument: Intent Is Provable in the Programming
The nH Predict case is not a story about AI being dangerous. It is a story about a corporation deliberately configuring AI to exploit the gap between its error rate and its appeal rate — knowing that patients who were wrongfully denied care would not fight back in sufficient numbers to affect profitability.
| Factor | AI as Weapon (nH Predict) | AI as Tool (Clinical Support) |
|---|---|---|
| Optimization Target | Cost reduction / denial volume | Diagnostic accuracy / patient outcomes |
| Known Error Rate | ~90% (deployed anyway) | Outperforms physicians in most domains |
| Physician Role | Overridden by algorithm | Augmented and supported |
| Patient Impact | Denied care, worsening outcomes, death | Earlier diagnosis, reduced administrative burden |
| Appeal/Review Rate | Exploited: only 0.2% of patients appeal | Designed to be reviewed and improved |
| Regulatory Status | Active litigation; Senate investigated | AMA-endorsed with appropriate guardrails |
Legislative Implication
The Clinical Integrity & Patient Safety Amendment and S.3829 do not oppose AI in healthcare — they demand that AI used in coverage determinations be held to the same standard as clinical AI: accuracy-optimized, physician-reviewed, and auditable. If a denial algorithm's error rate exceeds its appeal rate by design, that is not a technology failure. That is fraud by algorithm. The programming intention is the evidence.
The legislative response to AI denial fraud → Room III
S.3829 criminalizes the executive decisions that deploy inaccurate AI systems at scale. § 5 holds the physician who signs the denial personally accountable — regardless of whether an algorithm generated it.
External Health Insurance Watchlists & Resources
UnitedHealth Group Abuse Tracker →
Maintained by: American Economic Liberties Project (AELP)
Compiles news reports, legal filings, and analyses concerning alleged anticompetitive practices, patient care issues, and market power abuses involving UnitedHealth Group and its various subsidiaries across the healthcare industry.
HEALTHCARE — People Over Profit
The POPNYC healthcare page documents multiple "Deaths by Denial," providing specific examples of patients — including Little John Cupp, Nataline Sarkisyan, and Tracy Pike — who died after insurers like UnitedHealthcare, CIGNA, and Blue Cross Blue Shield repeatedly denied their doctors' requests for life-saving or necessary treatments, often deemed "not medically necessary."
- A 2023 class-action lawsuit alleges UHC's "nH Predict" AI algorithm has a 90% error rate but is used anyway — the company allegedly knows only a small percentage of patients will appeal.
- One "medical benefits management company" (EviCore) that insurers hire to deny claims advertises a 3-to-1 return on investment for those insurers.
- A 2022 survey of oncologists found 42% of prior authorizations were delayed, with doctors attributing "loss of life" to these delays 36% of the time.