CMS has long released provider-level, CPT-coded utilization or payment data. (See exactly which CPT codes your doctor was paid for in 2023...) Medicaid has usually been a beast unto itself, in 50 silos in the 50 states. But now, CMS releases a multi-gigabyte trove of Medicaid data. One banner application is to let the public find fraud.
However, there may be a lot of commercial uses for the data. You might be able to find who were the most active prostatectomy surgeons in Medicaid, or the most active at giving a particular IV cancer drug.
See FOOTNOTE for Medicaid and agentic AI.
See coverage at Axios here. See coverage at Newsday here. See some download sites at HHS here, here. The granularity at the latter is listed as Medicaid, January 2018-December 2024, Provider NPI x HCPCS/CPT code x Month. The format is 10 gigabyte files, so special data management skills are needed to use it. (Medicare doctor-CPT-HCPCS data is offered in a cloud database here.)
Two Thoughts (1) Fraud Is Often Blatant. (2) Commercial uses?
First, going back to when I was a Medicare medical director 2004-2008, fraud is often blatant. (At the time, billing for 10 $200-500 medical diagnostic tests on the same day was an instant fraud flag and a major form of fraud.) In the past years 2019-2025, there have been billions of dollars of genetic test fraud, but it was all blatantly obvious, like billing for six whole sequencing long rare genes for $8000 in a 90 year old in a nursing home and billing that same panel on 15,000 Medicare patients. Recently, labs were caught billing insane multi-thousand dollar codes like 81408 by a MAC in one year, and the SAME insanely behaving lab was allowed to billing ten million dollars the NEXT year on a different, essentially ADJACENT, set of gene codes. Here. The point is, it didn't require super AI detection to find this stuff, a ten year old could find it in 2 minutes. Another example was massively unbelievable billing of urinary catheters in incredible volumes. Here, here. While CMS did eventually pick that up (link previous), not mentioned is that they got away with zillions of dollars, enough to upend ACO economics, before being picked up.
Second, I don't have the data skills to work with a 10 GB database, but this could have commercial uses as well, such as tracking top providers of HCPCS drug codes or specialty surgical procedures for free in Medicaid data.
###
