Why Internal Medicine Is Ground Zero for Medicare Fraud

When we trained a machine learning model on 8,300+ convicted Medicare fraudsters and asked it to find 500 active providers with matching billing patterns, we expected the results to span dozens of specialties. Instead, one specialty dominated everything: Internal Medicine.

263 of the 500 flagged providers — 52.6% — are internists. The next closest specialty,Family Practice, accounts for 135 (27%). Together, these two primary care specialties make up nearly 80% of all AI-flagged providers. Every other medical specialty combined accounts for the remaining 20%.

This isn't because internists are inherently dishonest. It's because Internal Medicine is the perfect storm for billing abuse: high patient volume, broad procedure codes, and extraordinary discretion in what constitutes a "necessary" service.

The Numbers Don't Lie

Internal Medicine is the largest single specialty in Medicare by provider count — but it doesn't represent 53% of all providers. It represents roughly 15% of the Medicare physician workforce. The fact that it accounts for more than half of all fraud-pattern flags means the rate of suspicious billing in Internal Medicine is disproportionately high.

Why Internal Medicine? Volume + Discretion = Opportunity

Three structural features of Internal Medicine make it uniquely vulnerable to billing abuse:

The classic fraud playbook in Internal Medicine is simple: see a patient for a basic checkup, document it as a comprehensive evaluation, and bill 99214 instead of 99213. Do this 30 times a day, 250 days a year, and you've added $300,000 in fraudulent billing that's nearly invisible in the data — unless you're comparing it to what convicted fraudsters did.

The 99213/99214 Problem

The two most common billing codes in all of Medicare are 99213 (established patient, low complexity) and 99214 (established patient, moderate complexity). Together, they account for over $50 billion in annual Medicare payments. Our upcoding detector tracks these patterns nationally.

The difference? About $40 per visit and a judgment call. A legitimate internist seeing a patient for a blood pressure check bills 99213. A fraudulent one documents additional "complexity" — reviewing systems, discussing treatment options that may or may not have happened — and bills 99214.

Among our 263 flagged Internal Medicine providers, the average 99214-to-99213 ratio is significantly higher than the national average for internists. They bill the more expensive code at rates that match convicted upcoding fraudsters.

The Top 5 Flagged Internists

These five Internal Medicine providers have the highest fraud probability scores in our model. Their billing patterns are statistically closest to convicted Medicare fraudsters.

Internal Medicine vs. Family Practice

Family Practice is the second-most flagged specialty at 135 providers (27%). Together with Internal Medicine, these two primary care specialties account for 398 of 500 flags — 79.6%.

The pattern makes sense. Both specialties share the same structural vulnerabilities: high volume, broad codes, subjective documentation. But Internal Medicine flags at nearly twice the rate of Family Practice. Why?

Internists also tend to treat older, sicker Medicare patients with multiple chronic conditions — which provides legitimate justification for higher-complexity billing codes. The line between "this patient is genuinely complex" and "I'm documenting complexity that doesn't exist" is where fraud lives. And in Internal Medicine, that line is exceptionally blurry.

What This Means for Medicare

The concentration of fraud-pattern flags in Internal Medicine suggests that CMS and the OIG should focus audit resources disproportionately on this specialty — not because internists are bad actors, but because the structure of IM billing creates the widest window of opportunity for abuse.

Automated billing audits that compare a provider's 99213/99214 ratio to their peers, flag unusual services-per-beneficiary rates, and monitor markup ratios could catch the most egregious patterns before they cost taxpayers millions. Our model proves the math works — the question is whether anyone will use it.