Every year, health insurers deny tens of millions of claims. Some of those denials are appropriate — claims for services not covered under a given plan, or for care that was not medically necessary by any reasonable clinical standard. But a significant and documented share of denials are overturned on appeal, which means they were wrong in the first place. An insurer that denies a claim and then reverses that denial on appeal has not made a clinical judgment. It has made a financial one, and been caught.
The question this article examines is not whether denials happen — they are a structural feature of any claims-based insurance system. The question is how the denial system is structured, who operates it, what the denial rates look like by company, what the appeal system actually produces, and what the named accountability record looks like when investigators, whistleblowers, and courts have examined specific denial practices.
How Claims Are Reviewed and Denied
When a provider submits a claim to an insurer, it enters an adjudication process that operates at multiple levels. The first level is automated: claims processing systems check for coding errors, coverage eligibility, plan limitations, and prior authorization requirements. A large share of denials at this stage are administrative — the claim was submitted incorrectly, the patient was not enrolled, or the service required prior authorization that was not obtained. These denials are often correctable and providers routinely resubmit.
The second level involves clinical review — a determination that the care provided was or was not medically necessary, or that it was provided in an inappropriate setting. Clinical reviews are conducted by nurse reviewers, medical directors, and, increasingly, algorithmic systems that evaluate claims against clinical criteria without direct physician involvement.
The role of algorithmic denial tools has expanded significantly in the past decade and has produced the most documented cases of systematic over-denial. UnitedHealth Group’s use of an AI tool called nH Predict — developed by its subsidiary NaviHealth — became the subject of a federal class action lawsuit filed in 2023. The complaint, drawing on internal data and whistleblower testimony, alleged that the algorithm denied post-acute care claims for Medicare Advantage patients at a rate exceeding 90 percent, overriding the clinical judgments of treating physicians and hospital discharge planners. Internal documents cited in the complaint showed that NaviHealth employees were aware the algorithm’s predictions diverged substantially from actual patient outcomes but that denial targets were maintained regardless.
The lawsuit alleged that UnitedHealth used nH Predict not as a clinical appropriateness tool but as a claims reduction mechanism — denying coverage for skilled nursing facility and inpatient rehabilitation care that physicians had determined was medically necessary, on the basis of an algorithm that the company’s own staff documented as inaccurate. UnitedHealth denied the allegations. The case proceeded through federal court and drew significant investigative attention, including a ProPublica investigation published in 2023 that independently corroborated the core findings of the complaint.
The nH Predict case is the most thoroughly documented example of algorithmic denial at scale, but it is not an isolated one. Cigna’s use of an automated system to deny claims without individual physician review was the subject of a 2023 ProPublica investigation that found Cigna medical reviewers were rejecting claims at a rate of 1.2 seconds per claim — a pace incompatible with any meaningful clinical review. Cigna disputed the characterization.
Denial Rates by Insurer
Comprehensive, insurer-level denial rate data for commercial plans does not exist in a publicly accessible form because federal law does not require its disclosure. What is available comes from three sources: ACA marketplace plan data that HHS collects and publishes, state insurance commissioner reports in states that require disclosure, and investigative journalism drawing on internal documents and whistleblower accounts.
The Kaiser Family Foundation’s analysis of HHS marketplace data provides the most systematic available comparison. In 2021, UnitedHealthcare denied approximately 32 percent of in-network claims submitted through marketplace plans. Cigna denied 25 percent. Ambetter — a marketplace-focused insurer operating in multiple states — denied 49 percent of claims. The average denial rate across all marketplace insurers was approximately 17 percent.
These figures are for marketplace plans only and likely understate denial rates in commercial and employer-sponsored markets, where disclosure requirements are weaker or nonexistent. They also do not capture the full scope of effective denial — claims that providers do not submit because prior authorization was denied, or claims that patients do not pursue because the appeal process is too burdensome to navigate.
The variation in denial rates across insurers — from 17 percent average to 49 percent for the highest-denying marketplace insurer — is not explained by differences in the patient populations covered or the clinical appropriateness of care sought. It reflects differences in denial practice and the aggressiveness with which individual insurers use the denial system as a financial management tool.
What Happens on Appeal
Federal law gives patients the right to appeal denied claims through an internal insurer appeal process and, if that fails, through an external review by an independent organization. The existence of these rights is not in dispute. Their practical accessibility is.
Across all ACA marketplace plans in 2021, fewer than two-tenths of one percent of denied claims were appealed internally, according to KFF analysis of HHS data. Of those that were appealed, insurers reversed their denials in approximately 41 percent of cases — meaning that more than four in ten appealed denials were found to be wrong by the insurer’s own internal review process.
That 41 percent reversal rate on internal appeal is a documented measure of systematic over-denial. When an insurer reverses more than four in ten appealed denials through its own process, it is not making close calls that reasonable reviewers might decide differently. It is denying claims that it subsequently acknowledges should have been paid — at a scale that the near-zero appeal rate means most affected patients never recover.
The external review system provides a further check. Independent Review Organizations — IROs — conduct external reviews of denied claims when internal appeals are exhausted. External review decisions are binding on the insurer in most states for fully insured plans. Overturn rates in external review consistently exceed 40 percent nationally, corroborating the internal appeal data: a substantial fraction of denied claims that reach any form of independent review are found to have been wrongly denied.
The practical barrier to appeal is substantial. The process requires documentation, persistence, and time — resources that are unevenly distributed across the insured population. Patients who are ill, elderly, or without administrative support are least likely to navigate a multi-stage appeal process. The deterrence effect of appeal complexity is not an accident of system design. It is a structural feature that insulates a high proportion of wrongful denials from reversal.
The ERISA Preemption Problem
For the roughly two-thirds of covered workers in self-insured employer plans — plans governed by ERISA rather than state insurance law — the external review protections available in fully insured markets do not apply uniformly. ERISA preempts state insurance regulations, including state external review requirements, for self-insured plans. Federal external review rules apply, but enforcement mechanisms and remedies under ERISA are more limited than under state law.
A patient in a fully insured marketplace plan who wins an external review appeal can obtain binding reversal of a denial and, in some states, pursue additional remedies for bad-faith denial practices. A patient in a self-insured employer plan operating under ERISA has more limited recourse — the plan is not subject to state insurance commissioner oversight, state external review mandates, or state bad-faith insurance statutes. The insurer administering the plan — acting as a third-party administrator rather than as the risk-bearing entity — has additional layers of insulation from accountability.
The Named Accountability Record
Beyond the nH Predict and Cigna automated review cases, the documented accountability record for denial practices includes the following:
Humana paid $90 million to settle a class action lawsuit in 2023 alleging that the company improperly denied Medicare Advantage claims for post-acute care using the same NaviHealth algorithm at the center of the UnitedHealth litigation. Humana denied wrongdoing as part of the settlement.
Aetna — now CVS Health — settled a lawsuit in California in 2018 brought by the state insurance commissioner after an investigation found that a medical director had denied claims without reviewing patient medical records. The medical director testified in a deposition that he had never reviewed medical records before making denial determinations. Aetna paid $1 million and agreed to corrective action.
Cigna faced a class action lawsuit in 2023 following the ProPublica investigation, alleging that the automated mass-denial system described in that investigation violated California insurance law. The litigation was pending as of the close of this article’s research period.
The enforcement record — fines and settlements that represent a fraction of the revenue generated by the denial practices they address — illustrates a structural accountability gap. When the financial benefit of systematic over-denial substantially exceeds the expected cost of enforcement, the rational institutional response is to continue the practice and treat enforcement costs as a cost of doing business.
Health Insurance Hub
00 — Hub: Health Insurance Industry
01 — How the Health Insurance Industry Works — and Who It Works For
02 — How Health Insurers Make Money
03 — Designed to Discourage: How Benefit Structures Reduce Claims
04 — The Denial System: How Insurers Decide What Not to Pay
05 — Prior Authorization: What Patients Experience
06 — The Administrative Burden and What It Costs
07 — Narrow Networks and What They Cost You
08 — The Employer-Sponsored Insurance Trap
09 — The Broker and Consultant Layer
10 — Billed for Diseases They Never Treated: How Medicare Advantage Fraud Works
11 — What Single-Payer Resolves: The Evidence From This Hub
12 — Health Care Forum: Join the conversation here
This article was researched and drafted with AI assistance under human review. See our full AI and editorial practices.