Sources & methodology
Know what is behind the numbers.
Definitions, coverage and limits for MedRes analysis of CMS Federal IDR Public Use Files.
Source & coverage
Reporting period: October 1–December 31, 2025, based on closure in the federal IDR portal. The reporting quarter is not the date of service. CMS published this release on Jul 22, 2026; MedRes refreshed this analysis on Sep 17, 2026.
- CPT codes in the data
- 3715
- Anonymous dollar source rows
- 1,419,980
- Named emergency/non-emergency source rows
- 1,407,684
- Included dollar line items
- 1,197,949
- Usable dollar awards
- 1,107,531
- Included named line items
- 1,197,834
The directory includes every explicitly reported CPT code in either the anonymous dollar file or the named emergency/non-emergency file, with no volume quota. Codes only present on bundled or component records remain discoverable but have no single/batched benchmarks. Codes with fewer than 11 usable dollar awards open in payer view. Where neither view supports a published award or selection metric, the page remains searchable but is excluded from search indexing and the sitemap. Labels use reviewed MedRes wording where available, otherwise the most frequent service description reported in the CMS named file; missing descriptions are identified explicitly.
Only CPT single and batched line items are included. Bundled primary services, bundled components and air ambulance are excluded from benchmark calculations. The source dollar file includes air ambulance, so its full row count differs from the emergency/non-emergency named source.
Two separate data views
Dollar benchmarks
The QPA-and-offers file provides prevailing dollar offers, reported initial payments, QPA and geographic region. It does not provide payer names or dispute/DLI identifiers. Geographic regions are reported metropolitan or other QPA regions; they may span state boundaries or be unreported.
Payer outcomes
The named outcomes file provides payer names, service state, selected party, dispute identifiers and QPA-relative offers. It does not provide corresponding dollar awards. Payer names are normalized only for case and whitespace. We do not combine similarly named insurers, subsidiaries, plans or administrators. Distinct names can refer to the same organization.
These files are never joined by row order, CPT, geography, IDR entity or inferred matches. A payer-specific dollar award cannot be reliably derived from the published files.
How metrics are calculated
Median and middle 50%. The median is the 50th percentile of usable observations in the selected cohort. P25 and P75 use linear interpolation between ordered observations. We aggregate original observations, never averages of subgroup medians. Dollar display is rounded to whole dollars; exports retain source-derived aggregate cents.
Metric-specific denominators. Each statistic uses its own valid observations. Initial-payment and award sample sizes can differ. We do not subtract their medians and call the result recovery. Zero is a valid reported numeric value; missing and suppressed values are not zero.
Reported award / QPA. We verified numeric cells in the source workbook use percentage formatting, while the CSV stores the underlying ratio. A CSV value of 3.68 is 3.68× QPA (368%). We publish the median of reported line-level ratios, not a ratio of medians. CMS notes QPAs can be nominal, incorrect or unit-based, so especially large ratios require caution.
Provider offer selected. This is the number of line items selecting the provider/facility offer divided by all known provider-or-plan selections in the cohort. It is not the dispute-level majority outcome, a MedRes win rate or a probability for a future claim.
Defaults. The default-inclusive view includes reported default decisions. The non-default view includes only records explicitly marked “No.” Records with unknown default status remain in the inclusive view only.
Decision time. Median business days from initiation to closure, using one observation per unique dispute within each cohort. A dispute may appear in several code cohorts and should not be summed across codes. The 2023–2024 definition began at IDR entity assignment; it should not be compared directly with the 2025 measure.
Duplicate handling. Named records use dispute number plus DLI number. Exact duplicates are removed; conflicting duplicate identities stop the build for review. The anonymous dollar file has no stable identifiers, so identical-looking rows are retained. This single-quarter release avoids pooling multiple release snapshots; reopened/corrected-record uncertainty remains.
Suppression. Source statuses such as ^, N/R, N/A, + and REDACTED remain distinct in source processing. No suppressed amount is reconstructed. MedRes additionally withholds a metric when fewer than 11 valid observations support it, and omits table cohorts with fewer than 11 line items. Totals may exceed the sum of visible rows.
What the data cannot show
- All OON claims. These are payment determinations from a selected dispute population, not all claims, settlements, withdrawn matters or ineligible disputes.
- Cash received or incremental recovery. Prevailing offers are selected total OON payment amounts. Initial payment is not QPA, the insurer's submitted offer is not initial payment, and the award is not proof of collection.
- Perfect clinical comparability. All reported modifiers are pooled. Modifier collection began late in 2025, and a missing modifier is unknown—not proof of an unmodified service. Professional and technical components can therefore be mixed. Anesthesia time/unit differences also limit comparability.
- Eligibility from a CPT code. Claim facts, plan type, network/facility context, jurisdiction and timing control eligibility. Scheduled OON care at an OON facility is not made eligible by appearing under a listed CPT.
- Current negotiated fee schedules. Q4 2025 outcomes are historical; the 2026 publication date does not make them 2026 services or contracted prices.
Updates & reproducibility
Methodology 1.1.0 · Analysis updated Sep 17, 2026. Derived statistics are MedRes analysis of the source, not statistics calculated or endorsed by CMS.
The public coverage report records the source URL, checksum, counts, exclusions and percentage-format evidence. Raw files are stored outside the website. New source releases require schema validation and rebuilding; this directory is not a live CMS feed.
Source SHA-256: 60cac5a2c8af73418312974be14f7e225d560b0d2faed2bf3e20688dc317ada9
Explore the benchmarks