Medicare Fraud: Following the Money in Part D

How Medicare Fraud Works

Medicare fraud comes in many forms, but in Part D (prescription drugs), the most common patterns include:

• Pill mills — Providers running high-volume opioid practices with minimal patient evaluation, prescribing controlled substances to patients who don't need them or selling them on the black market.
• Kickback schemes — Prescribing expensive brand-name drugs when cheaper generics exist, often in exchange for manufacturer payments or pharmacy rebates.
• Identity fraud — Billing Medicare for prescriptions never written, using stolen patient or provider identities.
• Unnecessary prescribing — Writing prescriptions for drugs patients don't need, generating revenue through volume.
• Upcoding — Prescribing more expensive versions of drugs when cheaper alternatives would be clinically appropriate.

What Our Data Analysis Reveals

By analyzing 1,380,665 Medicare Part D prescribers across $275.6 billion in drug costs, we've identified several concerning patterns:

🔴 372 Excluded Providers Still Prescribing

The OIG maintains a List of Excluded Individuals/Entities (LEIE) — providers convicted of healthcare fraud, patient abuse, or other offenses. Despite being excluded, 372 of these providers still appear as active Medicare Part D prescribers in 2023 data. This raises serious questions about CMS enforcement.

🤖 Machine Learning Catches What Rules Miss

Our ML fraud detection model, trained on 281 confirmed fraud cases from the LEIE, flagged 4,100+ providers with prescribing patterns that statistically resemble known fraud. Of these, 2,579 were not caught by traditional rule-based scoring — suggesting a significant blind spot in conventional approaches.

💊 The Opioid Signal

Opioid prescribing is the strongest single indicator in our model. Providers who prescribe opioids at rates far above their specialty peers — particularly combined with benzodiazepines — are disproportionately represented among confirmed fraud cases. We track 450,343 opioid prescribers, with 113,169 prescribing at rates above 20%.

💰 Cost Outliers Signal Waste

Some providers prescribe drugs costing 10-50x more per patient than their specialty average. While some have legitimate clinical reasons (treating rare diseases, complex cases), the statistical overlap with fraud is significant. Our cost outlier analysis breaks this down.

How We Detect Suspicious Patterns

OpenPrescriber uses two complementary approaches:

1. 10-Component Rule-Based Scoring — Combines specialty-adjusted opioid z-scores, cost outliers, brand preference, elderly antipsychotic use, LEIE cross-referencing, drug diversity, and dangerous combinations into a transparent 0-100 score.
2. Bagged Decision Tree ML Model — 20 decision trees trained on confirmed fraud labels, achieving 83% precision and 67% recall in cross-validation.

Neither system makes accusations — they identify statistical patterns for further investigation.

The $275 Billion Question

If even 3% of Medicare Part D spending involves fraud or waste, that's $8+ billion per year. Better data transparency — making prescribing patterns visible to researchers, journalists, and the public — is one of the most effective tools for accountability.

Every dollar wasted on fraudulent prescriptions is a dollar not available for legitimate patient care.