Available now
Hospital, nursing home and home health agency financial data from Medicare cost reports
The annual Medicare cost reports that hospitals, skilled nursing facilities and home health agencies file with CMS, in one typed schema: beds, days, costs, revenue, margins and balance sheets for 314,687 reports from 34,244 providers. 7 tables, 349,552 rows.
Coverage: Hospitals (form CMS-2552-10): 80,077 reports, fiscal years 2011 to 2023. Skilled nursing facilities (CMS-2540-10): 193,420 reports, 2011 to 2023. Home health agencies (CMS-1728-20): 41,190 reports, 2020 to 2023. 56 states and territories. The data ends with fiscal year 2023, the latest year CMS has published as a public use file.
Why pay for public data
CMS publishes a separate public use file for each provider type and each year, with its own column labels. This package types every column and maps the three forms to one set of names, so a margin or a Medicaid share of days means the same thing in every table, and adds a provider table, decodes and a dictionary that gives the worksheet line and the fill rate by year for each column.
How it is built
- 1
Pull the Hospital Provider, Skilled Nursing Facility and Home Health Agency Cost Report public use files from data.cms.gov: 30 files.
- 2
Type every column, rename to shared names with the CMS field kept in the dictionary, and compute operating margin, total margin, occupancy, Medicare and Medicaid shares of days and cost per visit. Ratios are NULL when the denominator is missing or not positive.
- 3
Keep values as filed, with nothing capped, and flag the 16,985 reports that carry no values. Rebuilt annually after each CMS release.
What is in each record
| Field | Notes |
|---|---|
| RPT_REC_NUM, CCN, FISCAL_YEAR, FY_BEGIN_DATE, FY_END_DATE | One row per filed report. CCN is six-character text with leading zeros |
| FACILITY_NAME, ADDRESS, CITY, STATE, ZIP, COUNTY, CBSA_CODE | Facility business address as filed. COUNTY and CBSA_CODE on hospital and SNF rows |
| TYPE_OF_CONTROL, OWNERSHIP_GROUP, PROVIDER_TYPE, CCN_FACILITY_TYPE | Ownership decoded. PROVIDER_TYPE and CCN_FACILITY_TYPE on hospital rows |
| OPERATING_MARGIN, TOTAL_MARGIN, OCCUPANCY_RATE, MEDICARE_DAYS_SHARE, MEDICAID_DAYS_SHARE | Computed. Not capped, so some values are extreme |
| BEDS, TOTAL_DAYS, DISCHARGES, FTE_EMPLOYEES_ON_PAYROLL | Days and discharges also by payer |
| COST_OF_CHARITY_CARE, BAD_DEBT_EXPENSE, COST_OF_UNCOMPENSATED_CARE | HOSPITAL_COST_REPORT |
| DSH_ADJUSTMENT, IME_PAYMENT, OUTLIER_PAYMENTS, COST_TO_CHARGE_RATIO | Hospital Medicare payments. OUTLIER_PAYMENTS is empty from FY2019 in the CMS files |
| NET_PATIENT_REVENUE, TOTAL_OPERATING_EXPENSE, NET_INCOME, TOTAL_ASSETS, TOTAL_LIABILITIES | Income statement and balance sheet, with the same names in all three report tables |
| SNF_DAYS_MEDICARE, NF_BEDS, SNF_AVG_LENGTH_OF_STAY, WORKSHEET_C_COSTS | SNF_COST_REPORT. WORKSHEET_C_COSTS is not total facility cost |
| TOTAL_VISITS, COST_PER_VISIT, MEDICARE_VISITS_SHARE, VISITS_LUPA_EPISODES | HHA_COST_REPORT: visits, patient census and cost for each discipline by payer |
| PROVIDER | 34,244 CCNs with the latest name, address and ownership, and FIRST_FISCAL_YEAR and LAST_FISCAL_YEAR |
| CODE_VALUES, SOURCE_FILES, DATA_DICTIONARY | 69 decodes, the 30 CMS files with checksums, and one row for each of the 522 columns |
Who uses it
- Healthcare lenders, investors and M&A teams screening provider finances
- Hospital and post-acute strategy teams benchmarking peers
- Health economists and policy researchers
Pricing
$990 a year, including each annual refresh.
No personal contact data is collected or sold in any QuestionPunk dataset.
Questions
- Is this data not free from the government?
- Yes. The source files are public and we link to them. You pay for the cleaning, the joins, the documented schema and the refresh, not for the data itself.
- How is it delivered?
- As Parquet or CSV files, by download link or to your own cloud bucket. The same tables load into Snowflake with the scripts we include.
- Can I see the data before paying?
- Yes. Ask for a sample and we send the Vermont rows from every table, plus the full data dictionary.
- Are the numbers audited?
- No. Providers report them to CMS and CMS publishes them without edits, so a report may contain filing errors. The file does not say whether a report is as filed, amended or settled. Nothing is capped or cleaned, so some ratios are extreme.
- Why does it stop at fiscal year 2023?
- Fiscal year 2023 is the latest year CMS has published as a public use file. Later years exist only in raw HCRIS files, which this package does not parse. FISCAL_YEAR is the federal fiscal year in which the reporting period begins.
- Does it include hospice, dialysis or clinic cost reports?
- No. CMS publishes public use files only for hospitals, skilled nursing facilities and home health agencies. Home health starts in 2020 because CMS publishes that file only for form CMS-1728-20.
- Does it include personal data?
- No. Every row is a facility report. The public use files have no preparer, officer or contact columns, and a facility name that identifies an individual practitioner is withheld.
- What is the license?
- The cost report public use files are US government works in the public domain, published by CMS. CMS does not endorse this product.
Tell us what you need
Geography, size and fields. A two-line email is enough, and a person replies.
Request a sampleMore from the catalog
- University syllabi
- ZIP code demographics
- Census tract demographics
- ZIP code crosswalk
- US care facilities
- CDC PLACES health estimates
- Medicare spending
- US college data
- US salaries by occupation
- US housing cost and rent benchmarks
- FDA drug reference data
- US K-12 schools and districts
- IRS income and migration data
- US medical code sets
- Bank and credit union branches
- US health shortage areas
- ACA marketplace plans
- FDA Total Diet Study
- US jobs and commuting flows
- US air quality
- FDA medical device data
- CMS Provider of Services and CLIA labs