← medicaid-provider-spending

Random Sample

Visuals in this post: EyePic Case Per-Enrollee Spending Concentration Top Providers Outlier Screen Provider Lookup Random Sample Methodology
data.nycuriosity.com › Medicaid Provider Spending › Random Sample
Medicaid July 2026

See 1,000 raw rows exactly as the government publishes them

Aggregates can hide what a dataset actually is. This is a reproducible random sample of unmodified rows from the federal claims file, restricted to New York, Texas, and California providers, with every original column shown as published.

Download CSV (members)
1,000 raw rows from the federal file, exactly as published
A random sample of unmodified records from providers in New York, Texas, and California: no aggregation, no cleaning, no filtering beyond the state restriction. This is what one row of the 238-million-row dataset looks like. Click a column header to sort.
Provider State Billing Provider NPI Servicing Provider NPI HCPCS Code Claim From Month Total Patients Total Claim Lines Total Paid
Created by Tal Roded · NYCuriosity · Source: HHS Open Data, Medicaid Provider Spending by HCPCS

Every row is one billing provider, one servicing provider, one procedure code, one month. The federal file has no names, no states, and no diagnoses; identity comes only from the two NPI columns, which is why every analysis on this site joins them to the public NPPES registry. The Provider State column here is that join applied to the billing NPI; the other seven columns are verbatim from the source file.

Sampling: 1,000 rows drawn uniformly at random (fixed seed 0.42, so the draw is reproducible) from the 58.1 million rows whose billing NPI resolves to a NY, TX, or CA practice location in NPPES. No other selection was applied: rows with $0 paid amounts and rows from tiny providers appear at their natural rates. Because of the source file's privacy suppression, every row represents at least 12 claim lines and 12 beneficiaries.

Column meanings: exactly as documented by HHS. TOTAL_PATIENTS counts unique beneficiaries within the row's cell; TOTAL_CLAIM_LINES counts claim lines, not claims; TOTAL_PAID sums claim line-level paid amounts for the cell in dollars.

Sources: HHS Open Data, Medicaid Provider Spending by HCPCS (2026-02-09 release); CMS NPPES (July 2026). Replication steps in the methodology.