Claims & Billing›Claims Processing›Fraud Detection
JD

Fraud Detection

ML-powered anomaly detection and risk scoring for claims fraud, waste, and abuse

23
High Risk
$184,221 at risk
Score ≥ 85
67
Medium Risk
$412,887 at risk
Score 60-84
312
Low Risk
$1.2M monitored
Score < 60
Fraud Patterns
Upcoding31 flags
Duplicate Billing18 flags
Phantom Billing14 flags
Unbundling27 flags
Flagged Claims
High & medium risk — requires investigation
Claim IDProviderAmountRisk ScorePatternML Conf.
CLM-2025-047892
Dr. James Park
NPI: 1847221190
 $4,820
94
Upcoding 🤖 97%
CLM-2025-046102
Metro Imaging Ctr
NPI: 1902341122
 $12,440
88
Phantom Billing 🤖 91%
CLM-2025-045881
Southside Clinic
NPI: 1673849012
 $7,115
85
Unbundling 🤖 89%
CLM-2025-044990
Dr. Liu Ping
NPI: 1234567891
 $2,280
72
Duplicate 🤖 78%
CLM-2025-044112
Northgate PT
NPI: 1987654321
 $5,890
68
Upcoding 🤖 74%
AI Assistant

Fraud Detection uses machine learning models trained on historical claim patterns to identify potential fraud, waste, and abuse (FWA). Claims receive a risk score 0–100; high scores trigger immediate investigation workflow.

  • High (85-100): Auto-suspend claim, immediate review
  • Medium (60-84): Flag for manual review before payment
  • Low (0-59): Continue normal processing, monitor
  • Upcoding: Billing higher-intensity code than service rendered
  • Phantom Billing: Billing for services not provided
  • Unbundling: Splitting bundled procedures to increase payment
  • Duplicate Billing: Submitting same claim multiple times

The ML confidence indicator shows how certain the model is about its fraud pattern classification. Claims with both high risk score AND high confidence should be prioritized for investigation.