What DOGE Should Know About Medicare Fraud: $854B in Data, 500 AI Flags
Published February 2026 · 15 min read
The Bottom Line
The Department of Government Efficiency (DOGE) is looking for waste. Medicare is the biggest target: $854.8 billion in physician payments alone, an estimated $60+ billion lost to fraud annually, and 1.72 million providers billing the system. We built an AI model that flagged 500 providers matching convicted fraudster patterns. Here's the roadmap DOGE needs.
⚠️ Important Context
All data on this page comes from publicly available CMS Medicare payment records. Unusual billing patterns may reflect legitimate medical practices (such as high-volume drug administration where each unit is counted as a separate service), data reporting differences, or group practice billing. Inclusion on this page does not constitute an accusation of fraud or wrongdoing. Only law enforcement and regulatory agencies can determine whether billing patterns represent fraud. Providers flagged by our statistical model have billing patterns similar to previously convicted providers, but many may have perfectly legitimate explanations.
Medicare Is the Biggest Efficiency Problem in Government
Forget the $400 hammers. Medicare is where the real money is. The program spent approximately $1 trillion in total in 2023 — with $854.8 billion going to physician and supplier payments alone. The HHS Office of Inspector General estimates that $60 billion or more is lost to fraud, waste, and abuse every year. Some estimates run as high as $100 billion.
That means Medicare fraud alone likely exceeds the entire budget of the Department of Homeland Security. If DOGE is serious about cutting government waste, this is ground zero.
What the Data Actually Shows
We analyzed every line of the CMS Medicare Physician and Other Supplier Payment dataset — 96 million rows of billing data covering 1.72 million providers. We then trained a machine learning model on 8,300+ confirmed fraud cases from the LEIE (List of Excluded Individuals/Entities) and DOJ prosecution records.
The model (a Random Forest classifier with an AUC of 0.83) flagged 500 currently active providers whose billing patterns statistically match convicted Medicare criminals. These aren't random — at least 6 of the flagged providers were subsequently charged by the DOJ in the largest healthcare fraud takedown in history.
DOGE Dashboard: Medicare by the Numbers
$854.8B
Total Payments
1.72M
Providers
500
AI Fraud Flags
$60B+
Est. Annual Fraud
$3.22T
Billed Charges
$2.14T
Written Off
8,300+
Confirmed Fraudsters
53
Active Investigations
The 5 Biggest Fraud Patterns DOGE Should Target
1. Wound Care Schemes ($14.6B in DOJ Takedowns)
Wound care is the #1 Medicare fraud vector in 2025. Skin substitute products and debridement procedures have been exploited through phantom billing, kickback schemes with manufacturers, and billing for products never applied. The DOJ's June 2025 takedown — the largest ever — was dominated by wound care cases.
Our data confirms it: we found providers billing skin substitutes at 60x markup ratios, nurse practitioners in Phoenix billing $514 million for 2,974 patients, and Beverly Hills plastic surgeons billing wound care instead of cosmetic procedures.
Read: The Wound Care Industrial Complex →
Read: The Arizona Wound Care Ring →
2. Impossible Billing Volumes
We identified providers billing 400+ services per day — one every 1-2 minutes for 8 hours straight. The most extreme case: a single provider who billed 2.47 million services in 2023, or 9,862 per working day. That's one service every 2.9 seconds.
These aren't rounding errors. They're either outright fraud or systemic billing abuses that CMS should catch automatically but doesn't. A simple volume cap — already used by private insurers — would save billions.
Read: 9,862 Services Per Day →
3. COVID Test Billing Abuse
When CMS created the K1034 code for over-the-counter COVID tests at ~$12 each, some providers saw an ATM. Individual providers billed millions in COVID tests — with some shipping unrequested tests to seniors' homes and billing Medicare. The fraud was so pervasive that NPR documented cases of seniors receiving tests they never ordered.
Read: The COVID Test Gold Rush →
4. The Upcoding Epidemic
Upcoding — billing for a more expensive service than was actually provided — is the most common form of Medicare fraud. The difference between a Level 3 office visit (99213, ~$92) and a Level 4 (99214, ~$130) is subjective, and providers know it. Our analysis of the $117.7 billion office visit economy found systematic shifts toward higher-level codes across specialties.
Explore: The Upcoding Detector →
5. Geographic Fraud Hotspots
Medicare fraud isn't evenly distributed. Five states — California, Florida, New York, Texas, and New Jersey — account for over half of all AI-flagged providers. South Florida alone has been called the "Medicare fraud capital of the world." Targeted enforcement in these hotspots would yield the highest return on investment.
What DOGE Could Actually Do
Based on our data analysis, here are concrete, implementable actions:
Deploy AI Pre-Payment Review
Our model flags suspicious claims with 0.83 AUC accuracy. CMS currently does most fraud detection after payment. Shifting to pre-payment AI screening — like credit card fraud detection — could prevent billions in improper payments before they go out the door.
Implement Real-Time Volume Caps
No solo practitioner can legitimately bill 400+ services per day. Private insurers already use automated volume limits. CMS doesn't. Adding this single rule would immediately flag the most egregious billing abuse.
Audit the Top 500 Flagged Providers
Our model identified 500 providers whose billing patterns match convicted fraudsters. A targeted audit of these 500 — rather than random sampling — would yield significantly higher fraud recovery per audit dollar spent. At least 6 were already confirmed by DOJ action.
Close the Wound Care Loophole
Skin substitute products have virtually no billing limits. A single nurse practitioner can bill $22 million in wound care products with no automated review. Prior authorization for high-cost skin substitutes would save billions immediately.
Make the Data Public and Searchable
CMS publishes raw data files, but they're nearly impossible for the public to use. That's why we built OpenMedicare. If CMS provided a real-time, searchable interface — like what we've built — public oversight would multiply government enforcement efforts.
The $2.14 Trillion Nobody Talks About
Here's a number DOGE should know: providers submitted $3.22 trillion in charges to Medicare in 2023. Medicare paid $854.8 billion. The other $2.14 trillion was simply written off. This 3.8x markup ratio — providers charging nearly 4x what they get paid — is the biggest hidden number in American healthcare.
While the writeoff doesn't directly cost taxpayers (Medicare pays the allowed amount, not the charged amount), these inflated charges flow into private insurance pricing, out-of-network billing, and uninsured patient bills. The system-wide waste is staggering.
Read: The $2.1 Trillion Writeoff →
Why This Matters Now
The Department of Government Efficiency has positioned itself as a data-driven reformer. Medicare fraud is the ultimate test case: it's measurable, the data is public, AI tools already exist, and the savings potential is enormous. If DOGE can't make a dent in Medicare fraud — with $60+ billion in annual losses and ready-made AI detection tools — it's hard to argue it can reform anything.
The data is here. The tools are here. The question is whether there's political will to go after the healthcare industry's most profitable actors.
Explore the Data Yourself
- 🚨 AI Fraud Watchlist — 500 flagged providers
- 🔍 Search any Medicare provider
- 📊 State-by-state spending data
- 🔬 Full fraud analysis hub
- 🤖 How we built the ML model
Data Sources
- • Centers for Medicare & Medicaid Services (CMS)
- • Medicare Provider Utilization and Payment Data (2014-2023)
- • CMS National Health Expenditure Data
Note: All data is from publicly available Medicare records. OpenMedicare is an independent journalism project not affiliated with CMS.