Public evidence library · updated 2026-07-17

A new federal rule opened a window. We made the scattered files usable.

In 2026, covered health plans began publishing their first required prior-authorization metrics for 2025. The disclosures are public, but they are not one clean dataset. We locate them, preserve them, normalize the fields, and show exactly where every number came from.

Important boundary: the new rule covers prior authorization for specified Medicare Advantage, Medicaid, CHIP, and federally facilitated Marketplace plans. It is not a national file of every claim denial or every employer plan.

Why this exists now

The reporting changed. The usability problem did not.

CMS-0057-F required affected payers to post annual, aggregated prior-authorization metrics. The first reports describe calendar year 2025 and were due March 31, 2026. That created a new public accountability trail.

But the records still vary by payer, plan, program, file format, URL, naming convention, and level of detail. Some publish one consolidated page; others publish dozens of contract files. Some give rates without the counts needed to test them. State datasets may use anonymous carrier labels. Those choices make comparison difficult even when a file is technically public.

What is unique here: we preserve both the number and its chain of custody. You can see what a record covers, what it does not cover, what was missing, which version we used, and the original source.

Federal requirements: CMS final-rule fact sheet ↗ · CMS implementation FAQ ↗

From disclosure to evidence

What we had to do to make the data findable

  1. 01

    Locate

    Find payer and government disclosures spread across web pages, PDFs, spreadsheets, and contract-level files.

  2. 02

    Preserve

    Capture the exact public artifact and calculate a digital fingerprint, so a later change does not erase the record we used.

  3. 03

    Extract

    Keep reported counts, rates, labels, and blanks as reported. We do not turn an absent value into a zero.

  4. 04

    Normalize

    Standardize payer names, reporting units, dates, metrics, and plan types without mixing claims data with prior-authorization data.

  5. 05

    Publish

    Attach every record to its citation, original source, captured copy, version, scope, and known limits.

Normalization is not interpretation. We make like fields easier to compare and keep unlike populations separate. A high rate can identify a question worth asking; it cannot prove that an individual denial was wrong.

Collection coverage

What has actually moved through the pipeline

Known
284
files in the defined collection
Located
284
public sources found
Captured
284
copies preserved
Parsed
284
fields extracted
Normalized
284
made comparable
Published
284
records you can inspect

59 published records contain at least one source-reported blank or unavailable field. They are usable with that limitation; they are not stuck in processing. Next scheduled source review: 2026-08-15.

Use the data

Downloads are for analysis, verification, and reuse

The searchable library is the easiest place to start. Download a complete snapshot when you need to sort, calculate, cite, or combine records outside this site.

For patients: use these records as context and as questions for your insurer—not as a substitute for your denial notice, plan documents, or appeal rights.

Search the collection

Browse the records

Showing 284 of 284 normalized records.

The first 150 matches are shown. Narrow the filters or download the complete file.