Disability healthcare access & the insurance value-extraction stack — is insurance a scam? (the actuarial core vs. the operations layer where the extraction actually lives)

Built 2026-06-18 from research/spec-disability-healthcare-insurance-extraction.json.

Two interlocking questions. (A) Is insurance structurally a scam? Graded precisely: no categorically — the actuarial core sells variance-reduction/ruin-avoidance, is solvent without growth, and is not a Ponzi — but yes on a measurable spectrum (loss ratio + denial behavior + insurability), with credit-life / junk-warranty / investment-dressed-life products at the scam end. (B) Where does the real extraction live? Not in the pool math but in the operations layer: PBM middleman rent, algorithmic denial-as-margin, and health-data exploitation — and it falls hardest on patients with disabilities. The scam isn't the risk pool; it's the friction and the middlemen bolted onto it.

1. Is insurance a scam? — the actuarial core

The accounting identity is real. Over any pool/period: claims_paid ≤ premiums + float income − expenses − profit. You cannot have everyone pay in and everyone draw out more than they put in. Insurance is negative-sum in expected dollars by construction — the premium is priced above expected loss, and the gap funds overhead, profit, and the cut. The float (collect now, pay later, keep the investment income between) is a primary profit source.

But it is not a Ponzi — the bright line. A Ponzi requires growth (pays old investors with new investors' money; insolvent by construction). Insurance pre-funds via legally-mandated reserves: a closed book with zero new customers still pays every valid claim if priced right, because claims come from the same cohort's premiums + reserves + float, not new entrants. Pooling (law of large numbers) is mathematically sound for genuinely random, independent, insurable risks.

And negative-EV ≠ scam. Insurance is a bad bet in dollars on purpose — it sells variance reduction / ruin-avoidance, not return. Concave utility makes trading a small certain loss (premium) to avoid a large random ruinous one (house fire, $2M cancer bill) rational: it raises expected utility while lowering expected wealth. A real economic good. The "scam" enters only when a product is mis-sold (whole-life as an investment), insures non-risks, or is engineered to minimize payout. Grade: fact; "almost all insurance is a scam" is false as stated — but points at a real spectrum.

2. The loss-ratio spectrum — which products are scams

The discriminating instrument is the loss ratio (claims ÷ premiums) + denial behavior + insurability:

Verdict: false categorically, but correctly pointing at a spectrum — legitimate at one end (P&C/term-life), scam at the other (credit-life/junk-warranty/investment-dressed-life). Grade: fact.

3. PBM value-extraction — the clearest documented rent

Pharmacy Benefit Managers are the cleanest extraction layer. 6 PBMs manage ~95% of US prescriptions; the Big 3 — CVS Caremark, Express Scripts (Cigna), OptumRx (UnitedHealth) — are vertically integrated with insurers and pharmacies. Documented (FTC interim reports, Jul 2024 / Jan 2025):

• Spread pricing — bill plans more than they pay pharmacies; ~$1.4B from spread on just 51 generic specialty drugs over ~5 years.
• Vertical-integration markups — marked up specialty generics at their own affiliated pharmacies by hundreds-to-thousands of percent, >$7.3B above acquisition cost (2017–2022).
• Rebate-aggregator opacity — PBM-owned aggregators capture rebates, deduct fees first, and the plan sponsor is often never told the original total.
• Inverse incentive — PBMs can profit more when list prices rise (a % of a bigger number), misaligning the middleman from the payer.

Enforcement: FTC suit against the Big 3 + their GPOs over insulin rebating; a Feb 2026 Express Scripts settlement requiring transparency changes. Grade: fact.

4. Denial-as-margin — the operations-layer abuse

The health-insurance extraction isn't the actuarial load; it's denial-as-margin — delay/deny legitimate claims, externalize the cost of contesting onto patients/providers. Two documented, litigated cases:

• Cigna PXDX (ProPublica): an auto-flag system rejected claims not matching a pre-set list; medical directors then denied them without opening patient files — one doctor denying up to 60,000 claims/month at ~1.2 seconds each; >300,000 rejected in two months of 2022. Congressional inquiry + class action (state law requires individual physician review).
• UnitedHealth / nH Predict (naviHealth, Optum, acquired 2020): allegedly overrode physicians to deny elderly post-acute Medicare Advantage care, with a claimed ~90% error rate (9 of 10 denials reversed on appeal). The Senate PSI (Oct 2024) documented UHC's skilled-nursing denial rate rising ~9× and post-acute prior-auth denials going 8.7% → 22.7% (2019–2022), tracking naviHealth's rollout. In 2026 a federal court ordered UnitedHealth to produce broad discovery / disclose the algorithm.

The friction externality: denial-as-margin works because the cost of appealing is dumped on the patient. A 90%-reversed rate means the denials were mostly wrong — but most people never appeal, so the wrongful denial sticks as margin. Grade: fact (ProPublica; PSI figures; court orders); the nH Predict error-rate/intent claims are active-litigation allegations.

5. Why patients with disabilities bear it most

The friction model lands hardest on the disabled — highest utilization, most prior-auth touchpoints, most DME dependence, so every denial/delay/gap compounds:

• Prior-auth denials: Medicaid managed-care denied ~1 in 8 requests (2019); some MCOs >25% — similar concerns in Medicare Advantage.
• DME hurdles: power wheelchairs need face-to-face eval + medical-necessity prescription + approved supplier + prior auth; miss any step → denial; plus annual caps.
• Coverage gaps: Medicare classifies wheelchair ramps as home modifications, not DME — not covered, leaving $1,500–$15,000 out of pocket to make a covered wheelchair usable.
• Admin burden is especially hard to navigate for people with disabilities (Urban Institute).

Grade: fact.

6. Data extraction & surveillance pricing

A third layer: health-data brokering feeding pricing/underwriting. LexisNexis Risk Solutions markets a ~5-decade multi-source dataset — de-identified claims, clinical/consumer data, social-determinants-of-health data — plus "socioeconomic health and readmission risk scores" that "predict health risk independent of traditional healthcare data," and (for life insurance) richer inputs for "more precise pricing and more accurate mortality predictions at scale."

This is the infrastructure for surveillance pricing: scoring individuals from non-clinical/consumer data to price, segment, or steer them — shifting insurance from broad risk-pooling toward individualized prediction, which erodes the cross-subsidy (the healthy funding the sick) that makes pooling a social good. The disabled/chronically-ill are exactly the population such scoring most readily prices up or screens around. Grade: fact (the products are public); "erodes the cross-subsidy" is a documented-direction concern.

7. Synthesis — where the scam actually is

The actuarial core (pooling for ruin-avoidance) is legitimate and not a Ponzi. The extraction lives in the operations layer bolted onto it: PBM rent (spread/markups/opaque rebates), denial-as-margin (algorithmic + friction), and surveillance pricing. The scam isn't the pool — it's the friction and the middlemen — and by design it falls hardest on those who need care most.

So the honest answer to "is insurance a scam?": the math isn't; the loss-ratio spectrum tells you which products are; and the operations layer is where ordinary insurance is turned into extraction — documented, measurable, and reformable (loss-ratio floors, PBM transparency/divestiture, denial-rate accountability, data-broker limits), not inevitable.

8. Limits

Documented: the actuarial economics (identity, float, reserves-not-entrants, variance-reduction, loss-ratio spectrum); the PBM mechanisms + FTC enforcement; Cigna PXDX and the UnitedHealth/nH Predict denial figures (with the 90%-error/intent claims labeled active-litigation allegations); the disability DME/prior-auth barriers and coverage gaps; the LexisNexis data/risk-score products. Graded: "almost all insurance is a scam" = false categorically, true on a spectrum; "surveillance pricing erodes the cross-subsidy" = documented-direction concern. Institutional action documented; incentives — not an asserted unitary corporate mind — explain the behavior; no individual blamed. Overlay edges excluded from the proofs.

Sources: FTC — Pharmacy Benefit Managers report; FTC — second interim PBM staff report (Jan 2025); ProPublica — How Cigna saves millions having its doctors reject claims without reading them; CBS — UnitedHealth AI denial lawsuit; Becker's — court orders UnitedHealth discovery in AI denial case; Urban Institute — Barriers to Accessing Medical Equipment for adults with disabilities (PDF); Medicare.gov — wheelchairs & scooters coverage; LexisNexis Risk Solutions — healthcare market data.

← Research index · structured data: spec-disability-healthcare-insurance-extraction.json · spec-disability-healthcare-insurance-extraction.md