Lambert here: Strong on rhetoric, very very light on links. I’m all for open source, but I’m cautious enough to remember when “millions of people with laptops and basic data skills” fingered the wrong guy in the Boston Marathon bombing case.
The HHS DOGE team did something the federal government has never done — they open-sourced the largest Medicaid dataset in department history. Provider-level claims data from 2018 through 2024, organized by billing code, available for free download at opendata.hhs.gov.
Lambert here: If the data is clean and well-attested, sure.
Within hours, the internet went to work. One data analyst [who?] scanning just 0.16% of the providers in the dataset found roughly $90 billion in likely fraudulent payments. New York state appears to be the worst offender. Let that sink in — $90 billion from a fraction of a percent of the data.
Medicaid covers over 90 million Americans and costs taxpayers north of $800 billion a year. For decades, fraud detection has been left to inspector general reports and congressional hearings that move at glacial speed. Meanwhile, providers have exploited complex billing codes — particularly high-reimbursement categories like autism spectrum disorder therapies — to extract millions in payments for services that were either never delivered or wildly exaggerated. The Minnesota autism diagnosis fraud scheme, which the DOGE team explicitly referenced, is a case study: a network of providers systematically billing for fabricated therapies, laundering the money through shell companies, and sending it overseas.
What makes this release powerful is the approach. Rather than relying on a small team of government auditors to catch needle-in-a-haystack anomalies, they’re crowdsourcing fraud detection to millions of people with laptops and basic data skills. A provider billing an anomalous number of autism evaluations relative to peers in the same geography stands out immediately. Modern data analytics tools can identify these patterns in minutes — something the traditional auditing infrastructure has failed to do for decades.
This is the open-source ethos applied to government accountability. Sunlight as a disinfectant at scale. And the early results suggest the fraud is not a rounding error — it may be one of the largest ongoing thefts of public money in American history.
Lambert here, irritatedly donning his partisan hat: Forgive my cynicism, but Medicaid fraud is “shooting fish in a barrel”-level fraud detection. After all, a Medicaid fraudster was the 45th Governor of Florida (and a Republican). And it looks like little Marco is following in Rick Scott’s muddy footsteps. Sunlight shines everywhere. Does this particular brand?
Add new comment