Every figure on this site traces to one of the 37 public sources below. The list is generated at build time from the dataset's source registry (170 registered sources in all; the ones the Atlas renders from are listed here), so a source cannot be added to a page without appearing on this one. Each entry names the publisher, the date the file was taken, the vintage, and the link.
Medicare cost reports
8 sources
The filings behind every financial statement, ratio, trend, peer pool and CSV download. Published values are as filed with CMS; calculations are labeled and their formulas shown on the methodology page.
Centers for Medicare & Medicaid Services (CMS), U.S. Department of Health and Human Services
Accessed
2026-06-23
Vintage
One zip per fiscal year (FYE calendar year). Each download is dated and frozen. Federal restates are handled by re-downloading as new dated vintages (AD-016).
On the site
FY20 to FY25 as filed, 5,254 providers with a multi-year series
License
Public domain — CMS public use file. No restrictions on use.
Scope
FY = fiscal year end calendar year (per AD-008). Each zip contains RPT, NMRC, and ALPHA CSV files; ALPHA not used for this build. NMRC is long-form (one row per worksheet/line/column/value). Report status codes: 1=As Submitted, 2=Settled No Audit, 3=Settled With Audit, 4=Amended. Deduplication rule: settled (2 or 3) wins, else latest PROC_DT (AD-010).
Renders
Hospital profiles (all subtypes) · Hospital state rollups · Peer pools and percentiles · CSV downloads · Uncompensated care (Worksheet S-10)
FY19 to FY25 as filed, 8,555 providers with a multi-year series
License
Public domain (CMS public use file)
Scope
Dual form version. CMS-1728-20 effective for periods beginning ≥2020-01-01 (the backbone, ~10k/yr); CMS-1728-94 a 670-record FY2020 tail. Worksheet F = balance sheet, F-1 = income statement, S-3 = visits. Freestanding cohort (hospital-based HHAs on parent 2552-10).
Renders
Home health profiles · Home health state rollups · Home health benchmark gaps
FY20 to FY25 as filed, 4,505 providers with a multi-year series
License
Public domain (CMS public use file)
Scope
Single form (no -94/-20 style transition; legacy -99 has only FY1999-2014). F000000 balance sheet, F200000 income (F-2), S100000 days by level of care. Freestanding cohort. Chain from S-1 L9/L14.
Renders
Hospice profiles · Hospice state rollups · Hospice benchmark gaps
FY20 to FY25 as filed, 7,948 providers with a multi-year series
License
Public domain (CMS public use file)
Scope
Stable single form (no version transition). F000000 balance sheet, F100000 income statement, C000000 treatments, S000002 chain info. Independent/freestanding cohort. 95% chain-affiliated. Legacy RNL-REPORTS.zip (ends ~2012) NOT used.
FY21 to FY25 as filed, 904 profiles with a single-year exhibit
License
Public domain, CMS public use file (as the entry above)
Scope
CMS-222-17 (current). Legacy 222-92 (HCLINIC, RHC+FQHC mix) ends FY2018 — not in window. RHC-only (FQHCs on 224-14 since 2014). No financial statements (cost-per-visit only). No Care Compare. Cost/visit median $140; 57% capped above AIR.
Renders
Rural health clinic profiles (cost per visit, visits)
FY19 to FY25 as filed, 3,308 providers with a multi-year series
License
Public domain (CMS public use file)
Scope
Confirmed 224-14 FQHC (NOT 222-92 RHC). Income statement only (F-1); NO balance sheet. Medicare cost-report view, not whole-org (grants off-form). Worksheet B cost/visit. NO Care Compare for FQHCs (AD-086).
Renders
Health center profiles (Medicare cost-report appendix)
CAH department-line-level Medicare economic model (analysis build)SRC-CAH-001
Publisher
Astrelis analysis build on CMS HCRIS filings
Accessed
2026-06-23 (analysis build)
Built from
REUSES frozen 05_Raw_Frozen/HCRIS/FY{2020-2023}/2026-06-23/HOSP10_* (SRC-HOSP, no re-download)
Scope
CAH-only (cost-reimbursed). Cost-finding core (A col7, C cols, RCC) + data-quality flags = verified/shipped. Department Medicare apportionment = PROVISIONAL (partial D-series; ties ~29% of E Part B L21). Non-Medicare net rev is MODELED (estimated), facility total ties to G-3. 2552-10 instructions (Pub 15-2 Ch 40) for column meanings only.
Renders
Critical access hospital department teasers (contribution margins, Medicare utilization)
Identity, classification and geography
4 sources
Who a facility is, where it is, and which peer pool it belongs to.
Two segments unioned via pos_segment (zero CCN overlap, verified): QIES = Hospital(01)/RHC(12)/FQHC(21)/cat06/cat19; iQIES = SNF(20)/HHA(3)/Hospice(12)/Dialysis-ESRD(7)/+others. Provider-category & iQIES-type decodes confirmed empirically vs the spine; QIES cat 06/19 and iQIES type 5/6/8/10/11/13 left as RAW codes (no guess). GNRL_FAC_TYPE_CD gives hospital subtype incl. CAH(11)/REH(12)/LTCH(02)/P
Renders
Facility identity of record: name, state, certification status, hospital subtype · Related facilities
CMS updates this dataset on a quarterly/rolling basis. This pull is the 2026-06-23 snapshot, frozen as downloaded. Future refreshes write to a new dated subfolder.
License
Public domain — CMS public use data
Scope
facility_id = CCN (matches HCRIS PRVDR_NUM). hospital_overall_rating is the CMS 5-star Overall Quality Rating; footnote field carries suppression codes. Dataset represents the current active facility list — does not include historical closed facilities (closed facilities may still appear in HCRIS).
Renders
Hospital names, type and overall star rating where the cost report carries none
public domain (U.S. Government work); free, no API key
Scope
A *service*, not a frozen file: the authoritative frozen artifacts are (a) the input addresses (already frozen upstream) and (b) the deterministic result table. SNF rows retain their direct Care Compare lat/long; the geocoder supplies county FIPS uniformly. Batch match 89% on 42,029 geocodable addresses; overall coords 85% of all rows after the GEO-2b FQHC/RHC fix (0/5%→87/85%).
Renders
Facility geocoding: county, coordinates for related facilities
Long format (FIPS × Attribute × Value); RUCC_2023 attribute pivoted per county. County-level standard for rural-health analysis. RUCA (tract/ZIP-level) deferred — no single clean county file; RUCC is the canonical county metric.
Renders
Rural and metro classification on every profile and rollup · Rural PPS peer pool membership
Quality and operations
10 sources
The federal quality and operations files behind the How it operates panels. A measure CMS does not compute for a provider type never renders for it.
per-dataset CMS modified (Jan–Apr 2026); recorded per file
License
public domain (U.S. Government work)
Scope
~1.10M raw rows. Match 94–99.9% of dataset CCNs in the 6,233-CCN HOSP10 cohort; broad measures cover 72.7% of cohort, IQR/HVBP/HAC/HRRP only 39–49% (CAHs program-exempt — the small-rural finding, AD-114). No dataset 404'd or changed ID. Excluded (separate universes under same theme): ASC, VHA, PCH, IPF, DRG payment tables, National/State rollups, Birthing-Friendly (no CCN), CJR. Evidence-layer/dat
per-dataset CMS modified date (Nov 2025 – May 2026); recorded per file
License
public domain (U.S. Government work)
Scope
Full published field set (1,093 cols inventoried). Match: SNF/Dialysis ~94%, HHA ~84%, Hospice ~68% (current snapshot vs cohort fiscal years). SNF QRP Swing Beds joins CAH cohort (0% vs SNF, by design). Evidence-layer/data-level labeling deferred to attended atlas pass.
Renders
Skilled nursing, home health, hospice and dialysis quality measures · Dialysis station capacity and modalities
Current snapshot — 14,700 providers as of 2026-06-23
License
Public domain (CMS provider data catalog)
Scope
Contains Five-Star ratings (overall + 4 components), nurse staffing hours/resident/day, turnover rates, 3-cycle health inspection scores, fines and penalties. Join to HCRIS via CCN.
CY2025 Q1–Q4 (latest full year; series 2017-2025), ~5.29M facility-day rows
License
public domain (CMS)
Link
Not recorded in the registry; dataset: data.cms.gov "Payroll Based Journal Daily Nurse Staffing"; quarterly CSVs PBJ_dailynursestaffing_CY2025Q{1,2,3,4}.csv (also API-enabled).
Scope
Per facility per day: MDScensus + nurse hours (RN DON/admin/RN, LPN admin/LPN, CNA, NAtrn, MedAide), emp/ctr split. HPRD = hours ÷ resident-days, annualized across the 4 quarters (RN=DON+admin+RN; LPN=admin+LPN; CNA/aide=CNA+NAtrn+MedAide; total=sum). Raw daily hours/census = verified_fact; the annualized HPRD = inference (derived). Weekend HPRD = Sat/Sun WorkDates only. ⚠ SNF-only — PBJ does not
Renders
Skilled nursing nurse hours per resident day
CMS Nursing Home Care Compare — Provider Information (staff turnover/tenure fields, PBJ-derived)SRC-CMS-PBJ-002
Publisher
Centers for Medicare & Medicaid Services (CMS)
Accessed
2026-06-27
Vintage
Turnover fields: Total nursing staff turnover %, Registered Nurse turnover %, # administrators who left
License
public domain (CMS)
Link
Not recorded in the registry; dataset: provider-data catalog 4pq5-n9py; NH_ProviderInfo_Jun2026.csv.
Scope
CMS-published turnover (evidence_layer=verified_fact), itself derived from PBJ. Blank / footnote-only = not_reported (never zeroed). Turnover available for 13,365 of 14,685 PBJ facilities. Median total nursing turnover 45.3% (P10 27.6 / P90 65.2). SNF-only. Loaded as the turnover columns of the same quality_detail/snf_pbj_staffing.csv table.
Data year 2023 (RY2025 release; files modified 2026-05-29)
License
Public domain (U.S. federal government work)
Scope
The post-acute complement to the hospital INP/OUT PUFs (SRC-CMS-INP/OUT-001). Five type files unioned on an identical 29-col core via provider_type. Provider-level rows only (STATE/NATION aggregates excluded). Headline measures: distinct benes, episodes/stays/elections, service days, charges/allowed/Medicare/standardized/outlier payment, dual %, rural %, avg age, avg HCC risk. Validated vs nationa
Renders
Home health ZIP codes served and visits or episodes · Post-acute utilization context
HCI (Hospice Care Index H_012_00), HVLDL (H_011_01), HIS composite (H_008_01), avg daily census, CAHPS family rating. Provider data is long-format (pivot by CCN). Join by CCN.
Renders
Hospice quality measures and CAHPS caregiver rating
Health Resources and Services Administration (HRSA)
Accessed
2026-06-24
Vintage
UDS 2021-2024 (financials built on 2024); sites file 2026-06-24 pull
License
public (HRSA)
Scope
FQHCs only (RHCs not in UDS). Income statement only — no balance sheet → margins only. 967 awardees (2024); financials 50.5% (Table 9D consent-gated), total margin median +1.1%. Cell map verified vs the workbook UIIDInfo codebook. Crosswalk: 4,395 FQHC CCNs, 97.7% matched → 1,221 awardees. data_level=awardee; shared context never summed across a multi-site awardee's CCNs.
Renders
Health center organization panel: patients, visits, payer mix, clinical quality
Health Resources and Services Administration (HRSA)
Accessed
2026-06-24
Vintage
UDS 2021-2024 (financials built on 2024); sites file 2026-06-24 pull
License
public (HRSA)
Scope
FQHCs only (RHCs not in UDS). Income statement only — no balance sheet → margins only. 967 awardees (2024); financials 50.5% (Table 9D consent-gated), total margin median +1.1%. Cell map verified vs the workbook UIIDInfo codebook. Crosswalk: 4,395 FQHC CCNs, 97.7% matched → 1,221 awardees. data_level=awardee; shared context never summed across a multi-site awardee's CCNs.
Renders
Health center awardee to CCN crosswalk (match status printed on each profile)
U.S. General Services Administration, Federal Audit Clearinghouse (https://www.fac.gov); filings submitted by the auditees under 2 CFR 200 Subpart F
Accessed
FAC pulls 2026-10-09 to 2026-10-11 (fac-audits FAC-01 to FAC
Vintage
FAC audit years 2019 to 2025; the three newest public filings per organization render
License
Public record (FAC dissemination is public for is_public = true filings; suppressed filings are counted and labeled, never fetched by other means)
Scope
Whole-organization audited statements, not the Medicare cost report. A government, tribe or university auditee's audit is the parent's: no balance sheet renders on the health center's page. The site computes nothing from the cells; the payload is built by scripts/build-fqhc-audit.mjs (quarterly stage 6n) from the frozen snapshot through the registered export path once this entry lands.
Entry read from the site's registration patch (docs/FAC_SOURCE_REGISTRATION.patch); the dataset's source registry lands it by its own order.
Renders
Health center audited financial statements (Federal Audit Clearinghouse filings read by the FAC-11 template; every status column shown; rows attached by name labeled matched by name)
HRSA Bureau of Primary Health Care (Health Center Program H80 award data)
Accessed
2026-06-24 (frozen)
Vintage
Reporting years 2021, 2022, 2023, 2024 (4 H80 workbooks)
License
Public domain (U.S. federal — HRSA)
Scope
Thin vs full UDS — operational CONTEXT, not benchmarking-grade financials. The financial depth Capital Link sells is NOT in these public profiles (those come from the 990 layer). Awardee-level, not site-level (AD-062). county_fips derived (proxy) from the center ZIP via Census ZCTA crosswalk SRC-ZCTA-001. CANONICAL broad center-level UDS source. Distinct from SRC-UDS-PAYER-001 (the single-year tri
Indian Health Service (official basemaps_IHS ArcGIS account)
Accessed
2026-06-30
Vintage
1,140 facilities; 12 IHS Areas. ITU_CODE: T (Tribal) 901 · I (IHS-operated) 133 · U (Urban) 89 · N (Non-IHS) 17. LOCATION_TYPE: Title 5 Tribal 638 (699), Title 1 Tribal (199), IHS (132), Urban (89), Self-Governance (3).
License
US Government public-domain work (17 U.S.C. § 105) — owner determination 2026-07-22
Scope
The authoritative IHS/Tribal/Urban anchor. Source carries NO CCN (ASUFAC is the IHS facility id) → crosswalked to spine CCNs by name + address + geo (haversine on LAT/LONG vs provider_geo_master; name+state via POS) — the MN/WA address-anchor approach. Operation type from ITU_CODE (I=IHS-operated / T=Tribal-638 / U=Urban). Most facilities are tribal clinics/health centers that are NOT Medicare-cos
Renders
Indian Health facility profiles: governance, I/T/U classification, services
County context
4 sources
The county tiles beside every profile. Only the fields named here are read from the county file; no third-party county rankings are mapped.
The Census data API now requires a key; the bulk Summary File is keyless. DP profile tables not in the bulk tree → concepts computed from detailed B-tables (cell maps in Area-Context spec §2). data_level=county.
Renders
County tiles: median household income, poverty, uninsured, age 65 and over
3,144 counties × 40 measures; suppressed cells preserved. Discontinued CKD measure excluded (confirmed absent from current release). CDC WONDER county mortality DEFERRED — API is national-only (county grouping confidentiality-gated); recorded as a gap with method, not fabricated.
County rows, data year 2022 (matched num/denom) + 2021 trend
License
public domain (BEA)
Scope
⚠ BEA discontinued selected county tables 2024-11-14 (budget); CAINC35 ARCHIVED, final data year 2022 (CAINC1 still updated to 2024) → both taken at 2022. Line codes verified: 1000 total · 2100 retirement/disability (2110 Social Security) · 2200 medical (2210 Medicare, 2220 Medicaid/public-assistance medical, 2230 military medical) · 2300 income maintenance · 2400 unemployment · 2500 veterans. Val
Renders
County tiles: transfer share of income, medical benefits share
2026-07-03 freeze; study data 2012–2016 (91 CAHs, 18 states); IMPLAN-derived.
License
Public (federally-funded research study).
Renders
Economic footprint box (jobs, wages, county contribution), labeled an illustrative estimate
Physicians
4 sources
The four files behind the physician page: who a clinician is and where they practice, what Medicare paid them, what industry paid them, and the enrollment and NPI records that resolve the rest. Two entries are read from the site's registration patch until the dataset's registry lands them; two are addenda to entries the registry already carries.
Centers for Medicare & Medicaid Services (CMS), Provider Data Catalog (data.cms.gov/provider-data)
Accessed
2026-10-20 (PHYS-01)
Vintage
Catalog modified 2026-10-02 for the two roster files; 2026-08-18 for the PY2024 quality file. Care Compare refreshes these files on a rolling schedule; a refresh is a new dated vintage folder.
License
Public domain (U.S. federal government work, CMS)
Scope
Individuals only. 1,637,594 NPIs in the roster, 83,666 groups, 947,716 NPIs with a facility affiliation across 41,034 CCNs, 72,481 NPIs with quality rows over 314 measures. The server advertises no size for these files (no Content-Length, no Range support), so sizes were measured on receipt.
Registry
Entry read from the site's registration patch (docs/PHYSICIAN_SOURCE_REGISTRATION.patch); the dataset's source registry lands it by its own order.
Centers for Medicare & Medicaid Services (CMS), Open Payments (openpaymentsdata.cms.gov; files served from download.cms.gov/openpayments)
Accessed
2026-10-20 (PHYS-01)
Vintage
Program year 2025, first publication 2026-06-30 (P06302026_06032026). Open Payments republishes every program year each June with disputes resolved; a republication is a new vintage.
License
Public domain (U.S. federal government work, CMS Open Payments)
Registry
Entry read from the site's registration patch (docs/PHYSICIAN_SOURCE_REGISTRATION.patch); the dataset's source registry lands it by its own order.
Renders
Physician page: 2025 general payments total and records, research payments, the companies paying most
FFS Medicare only (excludes MA) — undercounts where MA penetration high. Fields: totals (HCPCS codes, benes, services, submitted charges, allowed, payment), bene avg age, dual count, avg HCC risk score. CMS blanks counts <11 → suppressed (never zeroed).
Physician page use
The physician page (PHYS-01 panel): beneficiaries, services, allowed and paid totals, the ten HCPCS with the most services (services additive across place of service; dollar totals DERIVED as services × published average), drug/medical sub-totals carried as SUPPRESSED (*) or COUNTER_SUPPRESSED (#) when CMS blanks them. 1,235,757 individual rendering NPIs; 89,266 have no published HCPCS row (every code under 11 beneficiaries); 23,767,417 (NPI, HCPCS) pairs are unpublished in total.
Centers for Medicare & Medicaid Services (CMS) / NPPES
Accessed
2026-06-27
Vintage
June 2026 V2 (main member npidata_pfile_20050523-20260607.csv; data through 2026-06-07)
License
Public domain (U.S. federal government work)
Physician page use
The physician page (PHYS-01 panel): NPPES is the fallback for name, credential, primary taxonomy code (text absent: no NUCC crosswalk is frozen) and practice location, and names the reason an S1 NPI is absent from Care Compare (deactivated; not PECOS-enrolled; enrolled but not rostered). PECOS supplies PAC ids, enrollment counts, provider types and states per NPI.
Files read
05_Raw_Frozen/NPPES_2026_08/NPPES_Data_Dissemination_August_2026_V2.zip (SHA-256 re-verified 2026-10-20, 8242b80b1f54ad5b…; member npidata_pfile_20050523-20260809.csv streamed, 9,726,865 NPIs: 7,415,294 individuals, 1,959,660 organizations, 351,911 deactivated without entity type) and sources/CMS_PECOS_Enrollment/2026-06-28/pecos_enrollment_2026_04.csv (SHA-256 re-verified, 315b6d49141cece7…; 2,98
Registry
Addendum read from the site's registration patch (docs/PHYSICIAN_SOURCE_REGISTRATION.patch); the dataset's source registry lands it by its own order.
Renders
Physician page: fallback name, credential, taxonomy and location; the reason an NPI is absent from Care Compare; Medicare enrollment (PECOS, SRC-CMS-PECOS-001)
This release
Reporting period
cost-report fiscal years FY2019 to FY2025 as filed · by provider type: hospital FY2020–FY2025, snf FY2020–FY2025, home health FY2019–FY2025, hospice FY2020–FY2025, dialysis FY2020–FY2025, rhc FY2021–FY2025, health center FY2019–FY2025
Source posting date
CMS cost-report postings of 2026-04-30 and 2026-07-14 and 2026-10-05, released to the site.
CMS reposts the cost-report files through the year. Each posting Astrelis takes is frozen and fingerprinted on the day it is taken, and the site reads only the postings that have been released to it. A posting that is frozen but not yet released is listed so the gap is visible.
Frozen
CMS posting
Status
What it carries
2026-06-23
2026-04-30
Released to the site
the first freeze: hospital, SNF and home health file years 2020 to 2023; hospice, dialysis, RHC and FQHC all-years
2026-07-07
2026-04-30
Released to the site
hospital, SNF and home health file years 2024 and 2025
2026-07-21
2026-07-14
Released to the site
CMS-2540-24 skilled nursing, file years 2025 and 2026
2026-10-10
2026-10-05
Released to the site
every family (AD-1355, DATASET-Q); held beside the archive until AD-1356 taught the gate this state; released 2026-10-21 by the owner's typed `go release 2026-10-10` (REL-01), the release chain at trunk 599e7dc1 on candidate c5's calculation (b6b455c5; RT-01 pass 8)
Frozen, not refreshed
Economic impact (jobs, wages, county contribution). Economic-impact figures come from a frozen 2026 snapshot. Its producer chain was retired after two of its three stages were found to read an intermediate that no longer exists; the published values are the independently verified artifact, checksummed and unchanged. These figures will not move between quarterly refreshes until that chain is rebuilt.
Industry studies
The studies draw on the cost reports and quality files above and, where an exhibit uses a published report, cite it on the exhibit. The sources cited across each study's exhibits:
The 2026 National Critical Access Hospital Study
Cited
cost reports and quality files above
The 2026 National Home Health Study
Cited
cost reports and quality files above
The 2026 National Hospice Study
Cited
CMS Hospice Monitoring Report, April 2026, Exhibit 1 · CMS Hospice Monitoring Report, April 2026, Exhibit 4a · CMS Provider Data Catalog, May 2026 refresh (2023 to 2024 claims); distribution computed by Astrelis · MedPAC, March 2025 Report to the Congress, Chapter 9 · MedPAC, March 2026 Report to the Congress, Table 10-5 · MedPAC, March 2026 Report to the Congress, Table 10-9 · MedPAC, March 2026 Report to the Congress, p. 298
Where Public SNF Benchmarks Break
Cited
CMS-2540 Worksheet S-3, FY2024 anchor. All values real; none suppressed or imputed. · Cohort staffing ladder at reported wages, exact scorer reproduction only (13,075 facilities). · FY2024 anchor, n=13,365. Material flip: crosses the cohort median and shifts at least 10 percentile points. · FY2024 cost report anchor. Care Compare and PBJ through June 2026. n=13,970. · HCRIS Worksheet S-3, FY2024 anchor. n=13,970. · Hold floor preserves each facility's current overall rating. Priced at reported wages. · PBJ contract hours against cost report purchased nursing services. Period alignment partial; see appendix.
Methodology and Full Findings
Cited
800 permutations, seed 12345, parent-clustering controls. n = 14,006 facility pairs. · Benchmark universe, fiscal year of each facility's anchor cost report. · CMS-2540 Worksheet S-3, FY2024 anchor. All values real_value; none suppressed or imputed. · Cohort staffing ladder priced at reported wages. Composition rules applied against each facility's own survey base. · Exact reproduction 13,075 of 13,943 (93.8%). 1,434 facilities excluded as not scored. · FY2024 anchor, n = 13,365. Material flip: crosses the cohort median and shifts at least 10 percentile points. Median-tier rates are in the materiality tiers table. · FY2024 anchor. Metric-facility readings, all six core metrics. · FY2024 anchor. National rate 9.86 percent. · FY2024 anchor. One line at or above the cohort P90, its pair at or below P25. n=13,911. · Five-Star scored from the frozen CMS Technical Users' Guide. Hold floor preserves the current overall rating. · HCRIS FY2024 anchor, Care Compare 2026-06-23, PBJ 2026-06-27. Medians. 75 percent filer-coverage gate applied. · HCRIS Worksheet S-3, FY2024 anchor. n=13,970. · PBJ 2026-06-27 against HCRIS FY2024 anchor. n=13,098 with both sources present. · Period-aligned subset: FY2025-anchored facilities against CY2025 PBJ (vintage 2026-06-27). · Universe n = 13,365 facilities with both benchmarks computable.
