Program Integrity Tool
Fraud is a network crime. See the network.
The Problem
Sophisticated fraud is built to pass claim-by-claim checks.
Each claim in a kickback ring or a phantom billing scheme can look perfectly clean on its own, so rules engines that score rows in a table either miss the scheme entirely or bury investigators in false positives. The collusion only becomes visible at the level of relationships: who refers to whom, who bills for whom, and which patterns repeat across a network.
What Program Integrity Tool Does
The Program Integrity Tool starts with a simple observation: fraud is a network crime. Instead of scoring rows in a table, it recasts claims data, providers, members, pharmacies, diagnoses, procedures, and payments, into a living map of explicit relationships.
On that map, every provider and member gets a mathematical signature based on how they behave and who they connect to. Similar signatures land close together, so coordinated networks stand out as tight clusters that no individual claim would ever reveal. That is how the platform exposes systemic fraud rings while cutting false positives.
And every flag is explainable. Investigators trace any alert back to the original claims and claim lines behind it, review evidence in visual dashboards with drill-down views, and keep full human oversight of every decision. The platform operates in full compliance with HIPAA and CMS Data Use Agreement requirements.
Credentials
| ✓ | CMS Crushing Fraud Chili Cook-Off Finalist |
|---|---|
| ✓ | HIPAA compliant |
| ✓ | CMS Data Use Agreement compliant |
| ✓ | NIST 800-53 compliant |
| ✓ | Human-in-the-loop oversight |
Recognition
CMS Crushing Fraud Chili Cook-Off: Finalist
CMS sponsored the Crushing Fraud Chili Cook-Off as a market-based research challenge: harness explainable AI and Machine Learning to find anomalies in Medicare claims data that translate into novel indicators of fraud, at scale, with humans meaningfully in the loop.
From a national field of 259 submissions, CMS selected ten finalist teams, including Amida, alongside KPMG, Milliman, and Stanford/UCSF and Abt. In the competition phase, Amida's team applied the graph-based detection techniques behind the Program Integrity Tool to Medicare Fee-for-Service Hospice, Part B, and DME claims through CMS Limited Data Sets.
Why Graphs
Same Data. Different Picture.
Four prescription fills that each look routine. Connected, they are one member drawing controlled substances from seven sources in thirty days, and only the network shows it.
The Schemes
Six Schemes. One Lens.
Every scheme looks different in the data, and each one betrays itself as a shape in the network.
How It Flows
From Claims Table to Fraud Network
How graph-based detection works, without the math.
- STEP 1ConnectClaims become a map: providers, members, pharmacies, diagnoses, and payments
- STEP 2ProfileEvery entity gets a mathematical signature from its behavior and connections
- STEP 3ClusterSimilar signatures group together, and coordinated networks stand out
- STEP 4ExplainEvery flag traces back to the original claims, with evidence attached
How It Works
Key Capabilities
- Circular referral ringsKickback schemes where providers steer patients to each other surface as loops in the referral network, ranked by claim volume and paid dollars.
- Phantom billingServices billed but never rendered: after-death claims, impossible provider days, transportation with no matching medical visit, and more.
- Pharmacy shopping and opioid diversionMembers cycling through pharmacies, prescribers, and emergency rooms to obtain controlled substances, spotted at the network level.
- Hospice and end-of-life schemesNon-terminal enrollment, upcoded levels of care, and stays that defy clinical patterns.
- DME supplier millsIdentity harvesting, sudden volume spikes from new suppliers, and equipment billed far beyond valid orders.
- Provider outlier profilingStatistical peer comparison that flags providers whose billing behavior stands far outside their specialty and region.
How We Compare
Rules-Based Detection vs. Program Integrity Tool
| Rules-Based Detection | Program Integrity Tool |
|---|---|
| Scores claims one row at a time | Sees the network: rings, clusters, and coordination |
| Black-box risk scores | Every flag explained and traceable to source claims |
| Alert fatigue from false positives | Clustering cuts noise; investigators get prioritized cases |
| Static rules that fraudsters learn to evade | Patterns derived from behavior, not fixed checklists |
| Finds problems only after payment | Supports pre-payment review and post-payment recovery |
Why It Wins
The Differentiator
Rules engines score claims one at a time, and sophisticated schemes are built to pass exactly those checks. The Program Integrity Tool works at the level where collusion actually happens: the network. It surfaces coordinated fraud rings with fewer false positives, explains every flag with evidence an investigator can defend, and keeps humans in control of every decision.
Works Well With
Complementary Products
TorQE
T-MSIS Outcome Review and Quality Engine. Catch errors before they reach CMS.
Learn moreLighthouse
An API-first validation engine: your pipeline calls Lighthouse mid-run, it checks the batch against your rule library where the data lives, and returns a clear gate: proceed or stop.
Learn moreTessarix™
The cloud-native data lakehouse that unifies Medicaid, CHIP, Medicare, and waiver data.
Learn moreSee the Networks Hiding in Your Claims.
Let us show you a new way to reveal illicit patterns in your claims data.