The Data

In February 2025, the HHS DOGE team released the largest Medicaid dataset the department has ever made public: 227 million rows of aggregated claims data covering January 2018 through December 2024. The dataset contains provider-level Medicaid billing organized by NPI (National Provider Identifier), HCPCS procedure code, and month — showing unique beneficiaries served, claims filed, and total dollars paid.

The raw data is a 2.9 GB Parquet file hosted on opendata.hhs.gov. Without tooling, it's essentially unusable for most people. This site makes every provider, every city, every billing code, and every state browsable with zero technical knowledge.

Total Medicaid spending in the dataset: $1.09 trillion across 617,503 billing providers.

Data Sources

Three public datasets, all freely available:

  • Medicaid Provider Spending — 227 million rows from HHS Open Data. Each row: billing NPI, servicing NPI, HCPCS code, month, unique beneficiaries, claims, total paid. Rows with fewer than 12 claims are suppressed for patient privacy.
  • NPPES — The National Plan and Provider Enumeration System maps NPI numbers to provider names, addresses, specialties, and organizational relationships. 9.37 million providers.
  • HCPCS Code Descriptions — Maps billing codes like T1019 to human-readable descriptions ("Personal care services per 15 minutes").

What This Site Shows

Everything here is derived directly from the public data above. We aggregate, index, and cross-reference it to make it searchable:

  • Provider pages — Total spending, monthly trends, billing codes, yearly breakdowns, organizational relationships, and automated billing flags for 617K+ providers.
  • State and city pages — Geographic breakdowns of Medicaid spending with top providers and billing patterns.
  • Billing code pages — National spending by procedure code with the top providers billing each code.
  • Specialty pages — Spending by provider type (e.g., Home Health Agency, Acute Care Hospital) with state-level breakdowns.
  • Provider network pages — All entities sharing the same authorized official, useful for identifying multi-entity operations.

Billing Flags

Some provider pages display automated billing flags. These are statistical signals, not evidence of fraud. They include:

  • Claims-per-beneficiary outliers — Providers filing significantly more claims per patient than peers in the same specialty.
  • Single-code concentration — Providers billing over 90% of their revenue from a single procedure code with very few codes overall.
  • OIG exclusion — Providers appearing on the HHS Office of Inspector General exclusion list who still show Medicaid payments in the dataset.
  • Entity proliferation — Authorized officials controlling large numbers of billing entities across multiple states.

Legitimate explanations exist for all of these patterns. A home health agency billing one code is normal. A large health system having many NPIs is expected. These flags surface statistical outliers for further review — nothing more.

What This Data Cannot Tell You

Important limitations:

  • This data shows what was billed, not what was medically appropriate. High billing is not inherently suspicious.
  • Claims with fewer than 12 occurrences are suppressed, so small providers may be underrepresented.
  • The dataset covers Medicaid only — not Medicare, private insurance, or out-of-pocket spending.
  • Provider names and addresses come from NPPES enrollment data, which may be outdated.
  • Billing amounts reflect payments made, not amounts initially billed (which are often higher).

Technical Details

This is a single Go binary serving pre-aggregated data from a 3.3 GB SQLite database. The 227 million raw rows are aggregated into six summary tables at build time. No external services, no JavaScript frameworks, no tracking beyond basic analytics. Pages are server-rendered HTML.

Contact

Built by Patrick White. Questions or feedback: [email protected]. Follow @MedicaidSpend on X for updates.

This site is not affiliated with or endorsed by HHS, CMS, or DOGE.