Browse the evidence · sourced insurer records

Search, compare, then inspect the source.

Two kinds of denials, ranked separately. Insurers report two different kinds of denials. A claim denial is a no after care: you received the care, then the insurer refused to pay. A prior-authorization denial is a no before care: the insurer refused permission for the service. We keep them separate because they happen at different moments and use different records.

The comparable claims table currently covers 2024 activity. The newer 2025 prior-authorization postings are on the insurer report pages; 2026 activity is still being collected.

The appeal gap · Medicare Advantage · 2024

Most prior-authorization denials were not appealed. Most appealed denials were reversed.

885 never appealed93 appeals won by patients22 denials held up
Source · KFF analysis of CMS Medicare Advantage data (2024)
Published
Jan. 28, 2026
Capture note
Insurer-reported to CMS; unaudited.
Scope
rates applied per 1,000 denials
Public lineage
Original source linked below. A public immutable archive and provenance ID are not yet published.
Open original source ↗

Each dot represents 4,100 prior-authorization requests that Medicare Advantage insurers denied in 2024.

885 of every 1,000 denials did not enter the formal appeal process. The published data does not say why.

115 were appealed. Insurers reversed 93 and kept 22 denials in place.

Claim denials · marketplace plans (HealthCare.gov) · 2024

Marketplace plans: who denied the most claims after care

Share of in-network claims denied in 2024 among parent companies with more than five million HealthCare.gov claims. Nineteen companies qualify; the five highest and five lowest are shown.

Highest in-network claim denial rates among large marketplace parent companies, 2024
Oscar Health25%
Molina Healthcare22%
GuideWell (Florida Blue)22%
Harris Health21%
Cigna21%
and the fewest
Lowest in-network claim denial rates among large marketplace parent companies, 2024
BCBS of Michigan17%
Arkansas BCBS16%
BCBS of South Carolina15%
Scott & White15%
Elevance Health8%

Marketplace average: 19 percent of in-network claims were denied — roughly 85 million claims. This is a 2024 Marketplace comparison, not a measure of your personal claim or an employer plan. View source → · All 19 companies →

Prior-authorization denials · Medicare Advantage · 2024

Medicare Advantage: who denied the most permission-before-care requests

Share of prior-authorization requests denied in 2024. This is a different measure, in a different market, from the Marketplace claims table.

Medicare Advantage prior-authorization denial rates by insurer, 2024
UnitedHealth12.8%
Centene12.3%
Humana5.8%
Elevance4.2%
where the denials concentrate
65%
long-term care hospitals
54%
inpatient rehabilitation
<8%
all services overall

No federal file currently gives the same insurer-by-insurer claim-denial view for Medicare Advantage. That missing comparison is itself a finding. Source: KFF analyses of CMS data, Jan. 28 and Jul. 6, 2026. View source →

From Kaiser’s posting · CY2025

One Kaiser contract: 27,992 denials, 310 appeals.

Its flagship Southern California Medicare contract denied 19.5 percent of 143,690 standard requests. One denial in ninety was appealed. A quarter of those appeals succeeded.

Source · Kaiser Permanente CMS prior-authorization reports (Southern California PDF)
Published
report dated Mar. 27, 2026
Capture note
Captured Jul. 8, 2026.
Public lineage
Original source linked below. A public immutable archive and provenance ID are not yet published.
Open original source ↗
From Cigna’s posting · CY2025

Cigna’s entire disclosure is one row.

Standard requests denied: 27 percent. Appeals won by patients: 16 percent. Number of requests: not disclosed anywhere.

Source · Cigna CMS annual prior-authorization statistics, CY2025
Published
created Mar. 2026
Capture note
Captured Jul. 8, 2026.
Public lineage
Original source linked below. A public immutable archive and provenance ID are not yet published.
Open original source ↗
From UnitedHealthcare’s posting · 2025

Same company, 10× spread.

Across 75 Medicare contracts, denial rates run from 2.3 percent to 22.8 percent. Which contract you’re in decides what gets approved.

Source · UnitedHealthcare CMS interoperability pages (full CY2025 summary metrics)
Published
2026
Capture note
Rates carry broad exclusions (drugs, post-acute transitions, delegated providers); contract-level PDF blocks archival tools.
Public lineage
Original source linked below. A public immutable archive and provenance ID are not yet published.
Open original source ↗

Documented findings about automated decisions

The machine, in four documented facts

UnitedHealthcare’s post-acute denials rose from 10.9% to 22.7% in two years after automation was introduced. Senate investigators reported the change; it does not prove that every automated decision was wrong.
Source · U.S. Senate Permanent Subcommittee on Investigations, majority staff report
Published
Oct. 17, 2024
Public lineage
Original source linked below. A public immutable archive and provenance ID are not yet published.
Open original source ↗
Allegation
A lawsuit alleges a 90% error rate in UnitedHealth’s nH Predict algorithm. This remains an allegation, not a court finding.
Source · Estate of Gene B. Lokken v. UnitedHealth Group, class-action complaint (allegations)
Published
Nov. 14, 2023
Capture note
Allegations, not findings.
Public lineage
Original source linked below. A public immutable archive and provenance ID are not yet published.
Open original source ↗
Denials are concentrated in harder-to-navigate care. Medicare Advantage post-acute requests were denied far more often than the overall prior-authorization average in 2024.
Source · KFF, MA prior authorization for post-acute care
Published
Jul. 6, 2026
Public lineage
Original source linked below. A public immutable archive and provenance ID are not yet published.
Open original source ↗
Automation changes the balance of effort. A company can issue a denial at scale; a patient still has to spend time challenging it. The public record gives that challenge more evidence.
Source · CMS WISeR model; WP Intelligence (Washington Post) Texas analysis via PYMNTS
Published
Jun. 25, 2026
Public lineage
Original source linked below. A public immutable archive and provenance ID are not yet published.
Open original source ↗

Public-report check · ten insurer organizations

Did insurers publish numbers people can actually check?

We checked whether each insurer’s 2025 prior-authorization report was easy to find, included the counts the rule asks for, covered the required scope, could be read, and could be saved. This is a check of the public report — not a score for how the insurer treats patients.

InsurerPublic-record gradeWhat the insurer postedWhat the report still leaves outOriginal file
Cigna0/5One line of percentages. No counts to show how large the problem was.
  • Number of requests
  • Number of denials
  • Number of appeals
  • Enough detail to check or compare
open payer source
CVS Health / Aetna1/5Percentages only; some figures are cut off.
  • The counts behind the percentages
  • A complete table for every contract
  • A data format that software can read
open payer source
Molina Healthcare1/5A public announcement; the actual tables require a login.
  • The tables themselves
  • The counts behind the percentages
  • A data format that software can read
search log on file
UnitedHealthcare2/5A summary is online, but detailed counts are hard to save or check.
  • Readable counts for each contract
  • A headline rate that covers all relevant care
  • A file that can be preserved
open payer source
Centene2/5Counts exist, but they are spread across many brand websites.
  • One place to search
  • A data format that software can read
open payer source
Elevance Health2/5Files exist state by state, with no single place to search.
  • One place to find every state
  • Counts in the files we located
  • A data format that software can read
search log on file
Highmark3/5A complete report, but two contracts repeat the same table.
  • Separate, accurate tables for each contract
  • A data format that software can read
open payer source
Humana4/5Complete counts and appeal results in downloadable reports.
  • A data format that software can read
  • An easy-to-find directory
open payer source
BCBS Michigan4/5Complete counts, appeal results, and decision times on a public page.
  • A data format that software can read
open payer source
Kaiser Permanente5/5Complete reports with counts in PDF and a data format.
  • Nothing material found in this pilot
open payer source

The grade is 0–5, one point each: findable without expert help · raw counts disclosed · complete and accurate · machine-readable format · open to archival tools. Full rubric and correction path on the methodology page.