Post-Payment Fraud Detection

Prosecution intelligence for Medicare fraud investigators.

Vigil scores every Medicare provider in the country for fraud risk across six independent signals — then shows its work. Every lead carries full feature attribution, provenance back to CMS and OIG records, and a case package built to survive cross-examination.

Built on CMS Part B & Part D, OIG LEIE, and SAM.gov public records — every score traceable to source.

Provider Risk Dossier
Specimen
NPI 1•••••8821 · Internal Med · FL
Provider #1138-A
Composite
94/100
High risk
LEIE exclusion history
PRIOR
Billing anomaly (peer dev.)
92
Rx pattern (opioid ratio)
88
Referral-network structure
76
Address cluster
84
SAM.gov debarment
FLAG
Provenance verified6 sources · SHAP attribution
1,230,274
Medicare providers scored
Every billing provider in the country
30,679,033
Referral relationships analyzed
Provider-to-provider network edges
79,746
Address clusters flagged
Multi-provider shell-entity candidates
15.2%
Recall on a temporal holdout
Against 2023+ exclusions never seen in training
The problem

The fraud is documented. The evidence isn’t.

The government loses more than $100 billion a year to improper Medicare and Medicaid payments. Recovering it falls to the investigators who build the cases — Medicaid Fraud Control Units, state Attorneys General, and the qui tam bar — where the outcome turns almost entirely on the quality of the evidence.

The analytics that exist were built to deny claims and market to payers, not to support a prosecution. They surface risk with closed-box scores no expert can defend on cross-examination, or they catalog providers without ever saying which ones are worth investigating.

For the investigator building the case, that gap is the difference between a lead that stalls and a provider you can charge.

How it works

Score every provider. Investigate the worst. Document the case.

01

Score

Every Medicare provider in the country is scored across six independent signals — prior exclusions, billing anomalies, prescription patterns, referral structure, address clustering, and debarments — into one 0–100 composite.

02

Investigate

Drill into any provider with full SHAP feature attribution, the address cluster they sit inside, and their referral network — so you see not just the score, but exactly which facts produced it.

03

Document

Export a case package: a cover sheet, per-provider analysis, loss estimate, methodology appendix, and a complete audit log of every query — formatted to hand to DOJ or a grand jury.

Every lead leaves as a case package — not a dashboard.

One provider, one document, built to be read by people who will never log into the tool: a line prosecutor, a relator’s counsel, a reviewing judge. The numbers trace back to source CMS, OIG, and SAM.gov records, with download timestamps and a sealed audit log.

  • PDF cover sheet plus per-provider analysis
  • Methodology appendix included in every export
  • Complete, timestamped audit log of every query
Case Package
Provider #1138-A
01Cover sheet & summary of findingsp.1
02Composite score & signal attributionp.2
03Peer-group billing comparisonp.4
04Referral network & address clusterp.7
05Prescription pattern analysisp.11
06Loss estimate & methodology appendixp.14
07Source provenance & audit logp.19
SHA-256 sealed23 pp · audit log attached
Methodology

Validated the way a court would test it.

A model that scores high on the data it was trained on proves nothing. So Vigil is validated on a temporal holdout: trained only on OIG exclusions recorded before 2023, then measured against exclusions that landed in 2023 and later — providers the model had never seen.

On that test, the composite ranking recalls 15.2% of genuinely excluded providers in its highest-risk tier — a forward-looking measure, not a number tuned in hindsight.

The full methodology document — feature definitions, weighting, and validation — is included in every case package, so the analysis is defensible on its face.

Temporal holdout · 2023+ exclusions
15.2%
Recall against unseen exclusions
Trained on pre-2023 LEIE records; scored against exclusions it had never encountered.
6
Independent signals
100%
Scores traceable to source
Who it’s for

Built for the people who bring the case.

MFCU investigators

Medicaid Fraud Control Units inside state Attorneys General offices. Vigil ranks every provider and shows the analysis behind each lead, so investigative hours go to the cases most worth pursuing — not to assembling the data.

State Attorneys General

Offices deciding which providers to charge. Vigil delivers a clear loss estimate, a documented basis for every score, and a sealed audit trail that holds up under scrutiny.

False Claims Act attorneys

Plaintiffs’-side qui tam counsel building healthcare whistleblower cases. Vigil turns a relator’s tip into a defensible, evidence-backed complaint — the kind DOJ is far more likely to intervene on.

See how the worst providers in the country score.

Walk the demo, or get in touch about access for your firm or unit. Engagements are scoped individually.