See the corrections policy for the full process. Reader-submitted corrections route through the per-vertical corrections forms and the Fonteum contact form.
- Entries
- 17
- Maintained by
- Editorial Lead
- Submit a correction
- Contact form
2026-07-19
Correction
Platform counts on the homepage, source registry, API docs, and machine-readable metadata
Cause: Static generation and edge caching outlived the dated database snapshot when source data landed without a deployment
As of July 19, 2026, the affected surfaces continued to serve 83 snapshot source IDs with at least one positive record count while production reported 86. On the same date, the affected surfaces served 71 source IDs whose latest positive-data snapshot fell within the preceding 45 days while production reported 76. The latter is an observation-recency measure; it does not establish a content change or serving state. The count calculation and source crosswalk were not changed. Every published platform count is now qualified by its observation date, and the daily 07:30 UTC platform-stats snapshot job immediately expires the shared data tag, invalidates both production domains, regenerates the affected pages and artifacts, and checks that each regenerated surface carries the newly written date.
2026-07-18
Clarification
Remaining withdrawn research studies
Cause: A second current-production audit found that several July 14 withdrawal reasons were directionally correct but incomplete or too broad
Fonteum completed route-local, read-only re-derivations for LEIE exclusion trends, NPPES anatomy, nursing-home ownership networks, and license-transparency research. None cleared the republication bar. The LEIE cohort has no applicable path to the published HCRIS risk denominator even though production has other CCN-to-NPI bridges; NPPES retains an exact-digest bulk archive, but the historical page conflicts with the matched archive and current serving data and cannot be reproduced by one production SQL statement; nursing-home production has CCN-to-owner and name-keyed system graphs with partial LEI enrichment, but not complete role-safe identity or validated sponsor and landlord classifications needed for the rankings; and the license study still lacks a current national access-and-tier relation. All remain noindexed pending the operator's kill decision, and no narrower replacement result was published under the withdrawn titles.
2026-07-18
Correction
Sitewide source-authority language
Cause: A shared band treated a short healthcare and economic agency list as the authority behind every public figure, including pages backed by other issuers
The global agency list and its every-figure claim were removed. Shared copy now states that source authority is record-specific and directs readers to the issuer and dataset named on the source page or returned record where that metadata is available. Registry size no longer implies that every page comes from one authority, and representative source-family lists are explicitly non-exhaustive.
2026-07-18
Correction
Source-ledger coverage counts
Cause: The prior live-source calculation joined only snapshot IDs already present in the registry crosswalk
The applied source-ledger migration assigns each observed snapshot ID one canonical owner without deleting legacy evidence and labels composite observers and internal ledgers so aliases do not inflate source-family counts. Production now derives three source-lifecycle results through compute_platform_stats: integrated registry rows, snapshot source IDs with positive data, and source IDs whose latest positive-data observation falls within 45 days. State Medicaid jurisdictions are a separate coverage result. No platform-wide complete-source count is published; completeness is evaluated against each named source's expected scope. Public surfaces read the dated database snapshot and do not embed a fixed production tuple.
2026-07-17
Correction
Nursing-home stars versus harm
Cause: The July 14 audit searched for dedicated rating tables and missed current CCN-keyed nursing-home rating claims in the provenance graph
Fonteum re-derived every published figure from current production using one bounded, server-side SQL statement. The corrected headline is 1,464 of 5,832 four- and five-star facilities (25.10%) with at least one CMS G-L citation in the literal 2023-04-01 through 2026-04-01 survey-date window. The prior full-inspection-cycle figures, named table, star bands, sub-rating averages, state and year tables, metadata, structured data, FAQ, and citation were replaced from the same result. The study was republished with its SQL and JSON/CSV results and returned to indexing; the July 14 withdrawal entry remains in the log as history.
2026-07-17
Correction
Home-health quality analysis
Cause: The ownership narrative was reversed and the July 14 audit missed current nursing-home rating claims used by the comparison
Fonteum re-derived every published figure from current production using one server-side SQL statement. The corrected article centers rating availability: 4,431 of 12,392 agencies (35.76%) are unrated, and every unrated row carries the CMS too-few-patient-episodes footnote. Proprietary agencies are 76.22% of rated agencies and 67.01% of unrated agencies, reversing the prior interpretation. The source release is May 7, 2026; May 25 is the ingestion date. The study was republished with current tables, limitations, structured data, SQL and JSON/CSV results, and returned to indexing; the July 14 withdrawal entry remains in the log as history.
2026-07-14
Correction
Indexed research and study figures
Cause: Production re-audit after source re-ingests and date corrections
Fonteum audited 118 indexed URLs covering 197 quantitative claim vectors. Of those vectors, 129 reproduced, 54 drifted, and 14 unsupported vectors could not be derived from the current production data. Surviving drifted figures were replaced with their current production values and stamped with a real 2026-07-14 observation date. Every surviving derivation now points to committed, explicitly bounded server-side SQL; no default-limited PostgREST read was used. The six studies containing unsupported vectors are recorded separately below as pending withdrawal, not as finalized retractions.
2026-07-14
Pending withdrawal
LEIE exclusion trends
Cause: The historical time-series derivation cannot be recovered from the current production data
The historical article remains at its original URL for the audit record, but its findings are no longer presented as current and the page has been removed from active discovery. Withdrawal requires an operator decision; this entry is not a final retraction.
2026-07-14
Pending withdrawal
NPPES anatomy
Cause: One or more published figure derivations cannot be recovered from the current production data
The historical article remains at its original URL for the audit record, but its findings are no longer presented as current and the page has been removed from active discovery. Withdrawal requires an operator decision; this entry is not a final retraction.
2026-07-14
Pending withdrawal
Nursing-home ownership networks
Cause: One or more published network derivations cannot be recovered from the current production data
The historical article remains at its original URL for the audit record, but its findings are no longer presented as current and the page has been removed from active discovery. Withdrawal requires an operator decision; this entry is not a final retraction.
2026-07-14
Pending withdrawal
Nursing-home stars versus harm
Cause: One or more published comparison derivations cannot be recovered from the current production data
The historical article remains at its original URL for the audit record, but its findings are no longer presented as current and the page has been removed from active discovery. Withdrawal requires an operator decision; this entry is not a final retraction.
2026-07-14
Pending withdrawal
Home-health quality analysis
Cause: One or more published quality derivations cannot be recovered from the current production data
The historical article remains at its original URL for the audit record, but its findings are no longer presented as current and the page has been removed from active discovery. Withdrawal requires an operator decision; this entry is not a final retraction.
2026-07-14
Pending withdrawal
License transparency matrix
Cause: The jurisdictional matrix cannot be derived from the current production data with a defensible source lineage
The historical article remains at its original URL for the audit record, but its findings are no longer presented as current and the page has been removed from active discovery. Withdrawal requires an operator decision; this entry is not a final retraction.
2026-07-12
Correction
Public claims across homepage, methodology, sources, pricing, and machine-readable metadata
Cause: Public copy had drifted from production counts and source-specific coverage
Fonteum replaced a catalog-sized platform claim with the production observation of 111 active source-registry rows; reframed the facts headline as 26.2M facts and disclosed 26,211,220 valid-shaped source-file SHA-256 values, 14 missing hashes, and zero deterministic fact-to-signature links; scoped Medicaid coverage to 13 state Medicaid lists; removed unsupported mainnet-validation language; limited historical replay to named sources with retained rows; and corrected the nursing-home penalty description to 6,919 facilities across the full table, $467,344,898 in positive fines, and 2,553 payment-denial records covering 1,950 facilities. Source freshness is now stated from each serving table's observed date rather than as a universal daily claim.
2026-07-12
Correction
Bulk export manifest and gzip CSV downloads
Cause: Format metadata did not match the bytes served by several advertised bulk URLs
Three utilization URLs advertised as gzip CSV downloads returned the same 1,203-byte CMS HTML response instead, and OFAC, UK, and UN URLs carrying a .csv.gz name returned raw XML. The manifest now lists only files whose compression, filename, media type, and CSV columns are delivered and tested. The unsupported forms were withdrawn rather than relabeled. Every currently advertised bulk entry is fetched by an automated test that verifies gzip decompression and its promised CSV header.
2026-07-12
Correction
Developer API, hosted MCP, and @fonteum/mcp contract
Cause: Package metadata and developer documentation drifted from the hosted MCP service
The package still identified itself as 0.2.2 in runtime metadata while npm served 0.3.0, and two tool calls pointed to retired routes. Hosted MCP and its discovery cards now identify the live hosted contract as 1.0.0. The in-repository stdio package and manifests target 0.4.0, which remains unpublished pending operator publication; npm still serves 0.3.0. The seven tool names and schemas use tested REST bridges for the next package release, and documentation now distinguishes anonymous rate-limited hosted MCP access from the package's optional downstream REST Bearer credential.
2026-07-12
Correction
Exclusion screening API, MCP tool, and screening pages
Cause: Serving-table projection lagged a newer signed source artifact
The OIG exclusion serving table remained on its 2026-05-08 release after Fonteum had attested snapshot 906 dated 2026-07-01. As a result, NPI 1689095580 could receive an excluded:false response even though the newer attested artifact contained its 2026-06-18 exclusion. Exclusion screens now compare serving coverage with the latest comparable attestation and return coverage_stale or indeterminate, with nullable negative fields, whenever the serving table is behind or cannot be reconciled. That response is not a clearance. The exact snapshot-906 atomic projection migration is prepared but remains unapplied; production stays fail-safe until the operator applies it and deploys the matching reader.
Updates to the methodology — what the methodology now requires that it didn’t require before — are not corrections. They’re doctrine revisions. Listed separately here so the corrections log isn’t inflated with policy edits.
Recent methodology updates
2026-07-18
Doctrine
Source authority now follows the issuer named on the record.
2026-07-18
Schema
Source lifecycle and state jurisdiction are separate query-reproducible measures.
2026-07-15
Doctrine
Public freshness now requires a proved canonical content transition.
2026-07-14
Research rule
Indexed study figures now require cap-safe SQL and fail closed when a derivation is unsupported.
2026-07-12
Doctrine
Public counts, freshness, history, and cryptographic claims tied to production evidence.
Snapshot publishes — CMS Care Compare extracts, BLS annual pulls, source-pack ingestion runs — are not corrections and not methodology updates. They’re dated data refreshes. Listed separately so the reader can tell at a glance which kind of change a given entry is.
Recent data refreshes
2026-05-03
Data snapshot
CMS Care Compare research bundle (home health + hospice).
Two more Tier-1 research snapshots published from the CMS Care Compare cluster: home-health quality by state and hospice provider availability by state. Supported cited fields displayed available source, date, and limitation metadata; Fonteum did not independently rate, inspect, endorse, or guarantee any agency or hospice.
2026-05-03
Data snapshot
Dialysis facility research snapshot published.
First state-level snapshot of CMS Care Compare dialysis facility quality data. Tier-1 research-only — no facility profile writes; data appears in /research aggregates only.
2026-05-03
Data snapshot
Nursing-home research snapshot published.
First state-level snapshot of CMS Care Compare Nursing Home Provider Information master dataset. Special Focus Facility status reported at state-aggregate level only — never on individual facility profiles. CMS ratings appear as CMS ratings.
- Editorial policy → The full corrections process (7-day window, retractions), independence rules, and the doctrine block.
- Methodology → How figures are sourced, refreshed, and checked — plus the full methodology changelog.
- Source library → Documented entries in the generated source-catalog subset, with tier and limitations.
- Data provenance → Field-by-field source register.