Trust layer

Methodology, sources & limitations

MedicareInsights is independent Medicare market intelligence built from named public CMS files. Every figure carries its source, vintage and formula. It is not affiliated with CMS or any carrier. Everything below is the complete source registry — the same records that power the provenance panel on every chart.

Status key

live refreshed automatically from the CMS API · static extract compiled from the named CMS file, automated pipeline pending · curated analyst-maintained editorial content · modeled MedicareInsights estimate — directional, never citable as CMS data.

New to these figures? How to read the data walks through the six that get misread most often — penetration bases, percentage points versus percent, dual capture, SNP types, the star-rating lag and the provenance tags.

Current production methodology — county MA & other health plan share

Source file	CMS Medicare Monthly Enrollment (BENE_GEO_LVL=County) · dataset `d7fabe1e-d19b-4333-9eff-e80e0643f2fd`
Reporting period	April 2026 (prior year April 2025)
Ingested	2026-07-31
Refresh cadence	Monthly, following the CMS Medicare Monthly Enrollment release (CCW-derived, roughly three months in arrears)
Methodology version	2026-08-24
Geographic coverage	All 3,199 US counties and county equivalents, 53 jurisdictions
Numerator	MA_AND_OTH_BENES — beneficiaries in a Medicare Advantage or other health plan (includes cost plans, PACE and demonstrations)
Denominator	TOT_BENES — all Medicare beneficiaries in the county with Part A and/or Part B
Formula	`penetration_pct = MA_AND_OTH_BENES / TOT_BENES * 100, rounded to 1 decimal`
Rounding	Penetration to one decimal. Counts are published integers and are not rounded.
Suppression	CMS suppresses small cells. Counties without both published counts are absent from the file rather than estimated or imputed.
Geography basis	MedicareInsights uses the direct CMS record for each geography level. The CMS National record is canonical for the national figure, the CMS State record is canonical for each state, and the CMS County record is canonical for each county. Summed county values are retained as a reconciliation check and are not substituted for the direct state or national records.
Download	ma_county_penetration_2026-04.csv

Why our MA & other health plan share can read lower than KFF and other published figures

Our denominator is TOT_BENES: everyone in the county with Part A and/or Part B. That includes people with Part A only, who cannot enroll in a Medicare Advantage plan. Sources that use an eligible-only denominator, or that count plan-year rather than monthly enrollment, will report a higher percentage for the same county. Neither is wrong. Match the basis before comparing two numbers, and see how to read the data for a worked example.

Known limitations, stated plainly

MA & other health plan share is computed from two published counts (MA_AND_OTH_BENES ÷ TOT_BENES), so it carries the definition of both.
The denominator includes beneficiaries with Part A only, who cannot enroll in Medicare Advantage. Figures that use an eligible-only denominator will be higher.
CMS suppresses small cells; counties without both counts are absent rather than estimated.
This file does not carry carrier, plan or star-rating detail. Any such figure needs the CPSC ingest, which is not yet built.

Reconciliation, not substitution. Summing county records does not necessarily reproduce the direct CMS state or national record. MedicareInsights reports the direct record for the requested geography and discloses the reconciliation difference rather than silently replacing it with a county-derived value. The three quantities are labelled separately everywhere they appear:

Quantity	National, April 2026	What it is
Direct CMS record	51.1%	Canonical. The CMS National row, 36,012,033 of 70,408,037.
Sum of county records	51.6%	Reconciliation only. Never substituted for the direct record.
Difference	0.5 pt	689,033 beneficiaries CMS reports without a county — mostly the Foreign row plus the suppressed Unknown row, almost all in Original Medicare.

Archived / superseded methodology

Terminology in archived sections. Where the word “penetration” survives below, it refers to the broader CMS MA-and-other-health-plan share (MA_AND_OTH_BENES ÷ TOT_BENES) unless otherwise stated. It never means a pure Medicare Advantage count. Current pages use MA & other health plan share.

Superseded — not used in current production analytics. Kept so figures published before August 24, 2026 stay reproducible. Nothing below feeds a current page, and none of it should be cited as a current MedicareInsights figure.

Basis	Status	Why it was retired
State-level basis state pages, February 2026	superseded	Quoted 57.9% nationally. Not reproducible from `MA_AND_OTH_BENES ÷ TOT_BENES` on the monthly file, which returns 51.1% for the same month. Superseded method. An earlier version of MedicareInsights aggregated county records to produce state and national figures. That approach was retired on August 26, 2026 after reconciliation testing showed county sums did not reproduce the direct CMS geography-level records. Current production uses the direct CMS State record for every state page and the direct CMS National record for the national figure; the county aggregation is still computed by `tools/test_state_reconciliation.py` on every build, as a reconciliation check only. This basis no longer appears on any production page.

See also the withdrawn datasets in the archive.

Withdrawn and archived datasets

Historical datasets are kept so past figures remain reproducible, but they no longer feed any page. Anything below has been removed from production.

Dataset	Status	Why
76-county market extract	withdrawn	Not reproducible from any CMS file. Audited 2026-08-24 against the April 2026 county file: denominators differed by 1.4x to 995x, penetration differed by a mean of 8.5 and up to 35.5 percentage points across 76 counties, the 2023-2026 history was smooth and monotonic in a way observed data is not, and at least one FIPS was wrong (48173 Glasscock was labeled Galveston, which is 48167). It was displayed as a CMS 'static extract', which it was not.
State enrollment, February 2026	superseded	A February 2026 vintage that does not reconcile with the April 2026 county file. Superseded. At the time it was withdrawn, state and national figures were aggregated from the county file. That rule no longer applies: since 2026-08-26 the direct CMS record is canonical for its own geography, and county sums are published only as a reconciliation.

The audit that withdrew these is recorded in docs/data-audit-2026-08-24.md. Archived copies remain at /assets/data/archive/WITHDRAWN_*.

Source registry

Both registries are published as data. dataset_registry.json is the source for every panel’s provenance class, vintage and last-updated date on this page. metric_registry.json names, for each published figure, the one object that owns it, the formula, the jurisdiction set it covers and the pages allowed to render it — and lists the figures we do not publish with the CMS ingest each one waits on.

Provenance standard

Every dataset panel below carries the same eight things, and a dataset that cannot carry them is not published as a live figure.

Exact CMS dataset name, as CMS publishes it.
Source URL, linked from the source-file row.
Source filename or tab where the file has one.
Reporting period (the vintage row), which is the period the data describes, not the day it was fetched.
Retrieval or update date (the last-updated row).
Transformation (the method row): every step between the CMS file and the number on the page.
Provenance classification, shown as a pill on the heading.
Limitations, stated in the dataset’s own terms.

The five classifications. live a CMS file this site ingests and refreshes on a schedule. static extract a CMS file captured once at a stated vintage, refreshed by hand. curated compiled by MedicareInsights from named public disclosures; not regenerable from an ingested file. modeled a MedicareInsights estimate built on assumptions that are stated with it. derived arithmetic performed by MedicareInsights over CMS-published records — accurate to the source, but not a figure CMS publishes. A MedicareInsights calculation is never presented as a CMS-published metric.

MA & other health plan enrollment and share — county, state and national live

The direct CMS record is canonical for its own geography: the national row for the national figure, the state row for each state, the county row for each county. County aggregation is published beside them as a reconciliation check and is never substituted for a value CMS publishes directly.

Source file: Medicare Monthly Enrollment (CCW-derived national, state and county beneficiary counts; data.cms.gov dataset d7fabe1e-d19b-4333-9eff-e80e0643f2fd, public API)
Vintage: April 2026
Last updated: July 31, 2026
Cadence: Monthly. This product is CCW-derived and posts roughly three months after the reporting month; the April 2026 file is the newest published. Cadence differs by product — see the four enrollment sources named below.
Method: Current. Each geography reads its own CMS record from the Medicare Monthly Enrollment dataset (BENE_GEO_LVL = National, State or County) as MA_AND_OTH_BENES / TOT_BENES. Summing the county rows is run as a reconciliation calculation and reported as a residual, not used as the published value. The national residual is 689,033 beneficiaries (0.98%), almost all of it the CMS Foreign row plus a suppressed unknown-geography row; state residuals have a median of 0.065 points and a maximum of 0.30 points. See current production methodology.
Limitations: Health plan share here is measured against all Medicare beneficiaries (Parts A and/or B), not the KFF eligible-only basis, which runs ~4pts higher.

National MA vs Original Medicare, 2010–2026 static extract

Source file: CMS Medicare Monthly Enrollment; CMS MA landscape files (historical)
Vintage: April 2026
Last updated: July 31, 2026
Cadence: Monthly enrollment; annual landscape
Method: Year-end MA and Original Medicare enrollment in millions; premiums/MOOP are enrollment-weighted plan-year averages.
Limitations: Static extract pending automated pipeline. Premium/MOOP averages compiled from CMS landscape releases, not recomputed here.

County-level MA market snapshot (the counties then tracked (dataset withdrawn)) static extract

Source file: Monthly MA Enrollment by State/County/Contract (CPSC family), from CMS Medicare Advantage/Part D Contract and Enrollment Data
Vintage: WITHDRAWN — historical only, not used in current production analytics. See the archive.
Last updated: May 1, 2026
Cadence: Monthly (pipeline pending; currently a curated extract)
Method: Penetration, D-SNP share, carrier count and top carrier per county compiled from CPSC files for the counties then tracked (a dataset since withdrawn — see below).
Limitations: Covers the counties then tracked (dataset withdrawn), not all ~3,100. Velocity = simple YoY penetration point change.

Carrier footprint by state static extract

Source file: Monthly Enrollment by Contract/Plan/State/County (CPSC), March 2026 reporting month, from CMS Medicare Advantage/Part D Contract and Enrollment Data; CMS landscape files. Verified August 26, 2026 against the CMS monthly report file list. The earlier “(OEC)” label was withdrawn: the Monthly Online Enrollment Center (OEC) Report is a different CMS report and is not used here.
Vintage: March 2026
Last updated: May 1, 2026
Cadence: Monthly (pipeline pending; currently a curated extract)
Method: Enrollment in thousands by carrier and state; plan counts, star rating of largest plan, premium/MOOP of flagship product.
Limitations: Curated for major carriers in key states; trend %s are YoY and carrier notes are analyst commentary, not CMS data.

Part D benefit parameters, CY2022–CY2026 static extract

Source file: CMS Part D benefit parameter announcements; IRA statutory provisions
Vintage: CY2026 final parameters
Last updated: April 7, 2026
Cadence: Annual (April announcement)
Method: Published standard deductible and IRA out-of-pocket cap by contract year; carrier deductible-waiver strategies compiled from public formulary filings.
Limitations: Waiver-strategy rows are compiled from public filings and may lag mid-year formulary changes.

Dual-eligible density by state static extract

Source file: CMS Medicare-Medicaid dual enrollment data (MMCO quarterly)
Vintage: 2025 Q4
Last updated: March 15, 2026
Cadence: Quarterly (pipeline pending; currently a curated extract)
Method: Duals as % of state Medicare population; D-SNP coverage % = share of duals in a county with ≥1 D-SNP available.
Limitations: State-level only; FIDE availability flag is binary and does not reflect county variation.

2026 plan exits and member migration curated

Source file: CMS 2026 non-renewal notices; carrier investor disclosures
Vintage: Plan year 2026 (announced fall 2025)
Last updated: January 20, 2026
Cadence: Annual (non-renewals announced each fall)
Method: Field-level provenance, because these five figures do not share a source. Terminated plan count (148) and affected enrollment (2.1M): compiled from 2026 non-renewal announcements and carrier investor disclosures; not reproducible from a CMS file ingested by this site. Migration destinations: January 2026 enrollment reconciliation, directional. Active-shopping share (62%): an estimate derived from that reconciliation, not a CMS-published measure. Carrier attribution: carrier disclosures. The dataset is classified curated until a service-area and plan-year crosswalk ingest can regenerate the headline totals from a named file.
Limitations: Migration percentages are early-year estimates; 'active shopping' share is directional.

Net benefit value (PMPM) vs 2022 baseline modeled

Source file: Modeled from CMS bid/landscape data — MedicareInsights estimate
Vintage: Plan years 2022–2026
Last updated: April 15, 2026
Cadence: Annual
Method: PMPM supplemental-benefit value modeled from landscape benefit files against a 2022 baseline. Directional estimate, not an actuarial figure.
Limitations: Modeled — do not cite as CMS data. Use for prioritization, not filings.

Star rating trajectory by carrier static extract derived

Source file: CMS MA Star Ratings (annual, published each October); CMS prior-auth reporting
Vintage: 2026 Star Ratings (published Oct 2025)
Last updated: February 10, 2026
Cadence: Annual (October)
Method: MedicareInsights enrollment-weighted derived carrier average star rating for each carrier's contracts, one point per rating year from the official October release.
Limitations: Only annual October ratings are shown; CMS does not publish quarterly ratings.

Star measure-level erosion, top carriers static extract

Source file: CMS Star Ratings measure-level data (Part C & D performance files)
Vintage: 2022–2026 rating years
Last updated: February 10, 2026
Cadence: Annual (October)
Method: Five-year measure-score series for UHC, Humana, Aetna on high-weight measures; driver notes are analyst interpretation.
Limitations: Three carriers, selected high-weight measures only. Driver commentary is interpretation, not CMS data.

CMS policy developments curated

Source file: CMS.gov newsroom, Federal Register, CMS fact sheets (curated)
Vintage: Through July 2026
Last updated: July 1, 2026
Cadence: Ad hoc, reviewed monthly
Method: Analyst-curated summaries of CMS announcements with impact ratings.
Limitations: Curated selection, not a complete feed. Impact ratings are editorial.

Regulatory reference library curated

Source file: CMS Medicare Managed Care Manual (Pub. 100-16); IRA 2022; 2027 Advance Notice
Vintage: Current through the CY2027 Advance Notice (published January 26, 2026)
Last updated: February 15, 2026
Cadence: Updated on major rulemaking
Method: Plain-language chapter summaries with analyst TL;DRs.
Limitations: Summaries, not legal text. Always verify against the manual before compliance decisions.

MedicareInsights Market Opportunity Score modeled

What it is: A 0 to 100 ranking of how attractive a county looks as a Medicare Advantage market, computed from published CMS figures with published weights. Deterministic arithmetic, not a prediction, not a recommendation, and not produced by a language model.
Source file: CMS Medicare Monthly Enrollment, county geography level, the same file cited on every county page.
Method: This score measures market attractiveness, not competitive winnability. It says how much opportunity a market holds, not how hard it would be to win — that needs carrier and plan data this site does not yet ingest. A separate Market Winnability Score is reserved for when those inputs exist; it is deliberately not implemented today. Each component is converted to a percentile rank against all 3,199 US counties, then combined with fixed weights: market size 0.30, headroom 0.25, penetration growth 0.25, enrollment growth 0.20. Percentile ranking rather than min-max scaling, so a handful of extreme counties cannot compress the rest of the distribution.
Coverage: Four of eight planned components are available today. Market concentration and carrier diversity require the CMS contract, plan and county enrollment file; product opportunity requires the Landscape and PBP benefit files; quality environment requires contract-level Star Ratings. None of the three is ingested yet. The weights above are renormalized over the four available components, and the four missing ones are shown on every county page as rows reading “not yet available” rather than being quietly dropped. Because all four missing components describe competitive difficulty, the current score measures market attractiveness, not winnability.
Bands: 75 and above HIGH, 50 to 74 MODERATE, 25 to 49 LOW, below 25 MINIMAL.
Limitations: Counties with fewer than 2,000 Medicare beneficiaries are flagged as low reliability, because their percentile ranks move sharply on small absolute changes. The score is a ranking within the county universe, so it is relative by construction: a score of 50 means median, not adequate.
Version: 0.1. Any change to a weight, a component, the normalization method or a band threshold increments the version and is recorded in the changelog. Full method: docs/opportunity-score-methodology.md.

SNP enrollment by plan and state live

Source file: CMS Special Needs Plan Comprehensive Report — plan-level tab SNP_REPORT_PART_17
Vintage: August 2026
Last updated: August 24, 2026
Cadence: Monthly, and it posts close to its reporting month — so this file runs ahead of the CCW-derived Medicare Monthly Enrollment product, which lags about three months. That difference is why the D-SNP capture figure is a cross-vintage proxy. Published only as a zip archive, not through the CMS data API, so the ingest is manual.
Method: 1,803 plan records carrying contract, plan, type (D/C/I-SNP), integration status, qualifying conditions and enrollment. National totals by type and by integration status are taken from the report's own summary tabs rather than re-derived; plan rows sum to the CMS published total of 8,437,333 exactly. Parent-brand grouping is inferred by keyword from the contract name and is labeled as such on the page.
Limitations: State is the finest geography CMS publishes, and only for single-state plans; 44 multi-state plans (146,934 enrollees, 1.7%) are excluded from state rollups rather than split arbitrarily. Plans under 11 enrollees are suppressed into three Under-11 aggregate rows with no state. The D-SNP capture proxy compares August 2026 enrollment to the April 2026 dual population and is suppressed for Puerto Rico.

Growth runway index by state and county modeled

Source file: MedicareInsights index computed from CMS Medicare Monthly Enrollment and CPSC county extracts
Vintage: WITHDRAWN — historical only, not used in current production analytics. Superseded by the April 2026 canonical records
Last updated: July 31, 2026
Cadence: No longer refreshed — withdrawn August 24, 2026
Method: Runway score = penetration gap x velocity x market size, normalized so the top geography scores 100. Penetration gap is the distance to the highest penetration observed in the tracked set (revealed attainable ceiling). Velocity is the YoY penetration point change: actual for tracked counties, enrollment-weighted tracked-county average for their states, and the tracked-set median as a proxy for states with no tracked county. Market size is the Original Medicare base in millions. Geographies under 50% penetration are classed as conversion markets (sell against FFS); 50% and above are switcher markets (sell against incumbents), matching the state-page framing used at the time. That framing is no longer used on any current page.
Limitations: This is a MedicareInsights index, not CMS data, and is never citable as CMS data. Velocity for states without tracked counties inherits the tracked-set median, a rough proxy. Use for territory prioritization, not filings.

2027 disruption pipeline (plan years 2027-2028) curated

Source file: CMS newsroom, Federal Register, CMS rate and star publications (curated forward calendar)
Vintage: Through July 2026, covering plan years 2027-2028
Last updated: July 17, 2026
Cadence: Reviewed monthly; re-dated as CMS publishes each cycle milestone
Method: Analyst-curated calendar of dated CMS actions and statutory milestones that can disrupt members effective January 1, 2027 or during plan year 2027, each mapped to a sales response for AEP 2027 (October 15 to December 7, 2026) and a setup note for plan year 2028. Events marked final cite a published CMS action; expected marks a recurring statutory cycle not yet published for 2027; watch marks an open docket.
Limitations: Curated selection, not a complete feed. The AEP 2027 plays and 2028 setup notes are editorial strategy, not CMS guidance. Expected and watch items are forward-looking and may not occur as described.

Carrier vulnerability battlecards curated

Source file: Assembled from CMS Star Ratings, CMS non-renewal notices, CMS contract enrollment extracts, and a modeled benefit-value series
Vintage: Multiple vintages (shown per fact)
Last updated: July 17, 2026
Cadence: Updates as each underlying dataset refreshes
Method: Each card lists the facts a rep can cite when selling against a carrier: star trajectory (CMS Star Ratings), 2026 plan exits (CMS non-renewal notices), enrollment momentum (CMS contract enrollment extract), and benefit-value erosion (modeled, labeled as such). Every fact carries the vintage of the dataset it comes from; underlying provenance panels are on the page.
Limitations: The selection and phrasing of facts is editorial. The benefit-value series is a MedicareInsights model, labeled modeled, and is never citable as CMS data.

Flagship plan sample by carrier static extract

Source file: CMS MA landscape & CPSC files (selected contracts)
Vintage: Plan year 2026
Last updated: May 1, 2026
Cadence: Annual landscape; monthly enrollment
Method: Largest contracts per carrier with enrollment, stars, premium, MOOP.
Limitations: Sample of flagship plans, not the full ~5,700-plan universe.

County MA & other health plan share — full coverage static extract

Source file: CMS Medicare Monthly Enrollment (county geography level)
Vintage: April 2026
Last updated: July 31, 2026
Cadence: Monthly, CCW-derived, roughly three months after the reporting month (run: refresh_cms_data.py --counties). Not the same rhythm as the CPSC contract files or the SNP Comprehensive Report, which post close to their reporting month.
Method: MA-and-other share of total Medicare beneficiaries per county from the CMS Monthly Enrollment county file; YoY velocity = penetration point change vs the same month one year earlier.
Limitations: Penetration and beneficiary base come from the Monthly Enrollment county file. Optional enrichment layers, added by refresh_cms_data.py only when their source files are supplied, attach to the same records: ACS 65+ population (pop65k; provenance acs_county_65plus), and CPSC-derived carrier count / top carrier / D-SNP share (provenance ma_county_markets) plus landscape $0-premium availability (provenance plan_landscape_zero_premium). Counties under ~20k beneficiaries are flagged low-reliability. Vintage is stamped by the ingest; until first run this dataset is empty and no page depends on it.

County population 65+ (Census ACS 5-year) static extract

Source file: U.S. Census Bureau, American Community Survey 5-year estimates (table B01001)
Vintage: Pending first ACS ingest
Last updated: May 1, 2026
Cadence: Annual (ACS 5-year release; run: refresh_cms_data.py --acs)
Method: Population aged 65 and over per county, summed from ACS table B01001 sex-by-age cells; used as a demographic denominator and market-headroom context, not as the Medicare-eligible base.
Limitations: ACS is a survey with margins of error, largest for small counties. 65+ population is not identical to Medicare eligibility. Optional context layer; no core penetration figure depends on it.

County $0-premium MA-PD availability static extract

Source file: CMS Medicare Advantage / Part D Landscape source files
Vintage: Pending first landscape ingest
Last updated: May 1, 2026
Cadence: Annual (released each fall for the coming plan year; run: refresh_cms_data.py --landscape-file ...)
Method: A county is flagged $0-premium when at least one MA-PD plan in the CMS landscape file lists a $0 Part C monthly premium for that county.
Limitations: Reflects availability, not enrollment or the cheapest plan a given beneficiary qualifies for. Excludes SNP-only or employer plans not in the public landscape file. Optional layer; absent until the landscape file is ingested.

Update cadence

CMS publishes monthly enrollment roughly three months in arrears; star ratings each October; Part D benefit parameters each April; plan non-renewals each fall. The site rebuilds on each ingested release — the freshness badge on every page shows the exact data vintage and the date we last updated.

Penetration basis — read this before citing

Three bases appear on this site. They are all defensible and they do not agree. Every page states which one it uses; cite the basis with the number.

Monthly-file county basis county explorer, April 2026: MA_AND_OTH_BENES ÷ TOT_BENES from the CMS Medicare Monthly Enrollment file at county geography — 51.6% national. This is the definition CMS itself uses for its published county penetration series, so it is the basis for the county explorer and the county-level CSV. The numerator includes cost plans, PACE and other managed care alongside Medicare Advantage.
Pure-MA plan-year basis national trend: Medicare Advantage only, measured at a fixed point in the plan year — 50.8% for 2026. Excludes cost plans and PACE, which is why it reads below the monthly-file county basis.

KFF's widely-cited eligible-only basis restricts the denominator to beneficiaries with both Part A and Part B (64.3M of 70.3M in April 2026) and therefore runs higher than all three.

Get the 5 biggest shifts in Medicare data, monthly

One email per CMS release cycle: what changed in the new data, the five largest market shifts, and links to the updated answers. No product news, no filler.

No spam, unsubscribe anytime. See the privacy policy. Latest issue: Data release briefings →