Access our complete Medicare spending datasets for research, journalism, and analysis. All 19 datasets are free to download.
All data is sourced from public CMS records (2014–2023). Licensed under CC BY 4.0. Please cite OpenMedicare when using our datasets. See our API documentation for programmatic access.
Pre-formatted CSV files ready for Excel, Google Sheets, or any data tool.
Top 100 Medicare providers by total payments — NPI, name, specialty, state, payments, and services.
All states and territories with total Medicare payments, provider counts, services, and markup ratios.
All 132 medical specialties with total payments, provider counts, and markup ratios.
All 500 watchlist providers with risk scores, billing flags, and payment totals.
Essential Medicare spending datasets — providers, states, specialties, and trends.
Anomaly detection and billing investigation datasets.
Fraud feature matrix for 1.14M providers (services/day, markup, etc.)
Machine learning fraud model v2 results — 500 flagged providers with fraud probabilities, risk factors, and markup ratios. Model AUC: 0.83.
Detailed breakdowns by geography, place of service, utilization, and more.
Individual detail files are also available for each provider, procedure, state, and specialty. Access them via the site or directly:
All datasets are available as static JSON files at predictable URLs — no authentication needed. Perfect for building dashboards, research tools, or data journalism projects.
View API Documentation →Key fields used across our datasets.
| Field | Type | Description |
|---|---|---|
| npi | string | National Provider Identifier — unique 10-digit ID assigned to every Medicare provider |
| total_payments | number | Sum of Medicare payments received across all services and years |
| services_per_day | number | Total services ÷ 250 working days — flags impossibly high billing volumes |
| markup_ratio | number | Submitted charges ÷ Medicare payments — how much more than the Medicare rate was billed |
| risk_score | number | Composite anomaly score (0–100) combining multiple billing flags (watchlist only) |
| specialty_zscore | number | Standard deviations from specialty median — measures how far a provider deviates from peers |
| total_charged | number | Sum of submitted charges (what provider billed before Medicare adjustments) |
| total_services | number | Total service line items billed to Medicare |
| total_beneficiaries | number | Unique Medicare patients served |
| hcpcs_code | string | Healthcare Common Procedure Coding System code identifying specific services |
| specialty | string | CMS-assigned provider specialty classification |
| fraud_probability | number | ML model v2 predicted probability of fraud (0–1). Only present in ml-v2-results.json for flagged providers. |
| risk_rank | number | Rank among flagged providers (1 = highest risk). Only in ml-v2-results.json. |
| top_risk_factors | string[] | Human-readable descriptions of why the provider was flagged (e.g., "Matches confirmed fraud profile") |
| services_per_bene | number | Average services per beneficiary — high values indicate potential over-utilization |
| avg_markup | number | Average markup ratio across all services (watchlist providers) |
| flags | object[] | Array of billing flags with type, description, and severity (watchlist providers) |
Our data is derived from the CMS Medicare Provider Utilization and Payment dataset. You can access the original source files directly from CMS:
CMS Medicare Provider Data →Last Updated: February 2026 (data through 2023, the latest CMS release)
Note: All data is from publicly available Medicare records. OpenMedicare is an independent journalism project not affiliated with CMS.