Medicare Telehealth Program Integrity System

Combined Risk Score Framework for Identifying Anomalous Billing Patterns

$31.70B
Annual Medicare Improper Payments
7.66%
FY 2024 Improper Payment Rate
88x
Telehealth Usage Increase (Pandemic)
1,714
High Risk Providers (HHS OIG)

KEY PERFORMANCE INDICATORS

91.3%
Accuracy
100%
Precision
42.0%
Recall
0.59
F1 Score
0.94
AUC Score

MODEL VALIDATION RESULTS

Method Result Status
Temporal Stability Test 92.0% PASS
False Positive Rate (Critical) 0.0% OPTIMAL
Cross Validation (k fold = 10) 91.3% ± 2.4% PASS
Bootstrap Analysis (n = 1,000) 95% CI Stable PASS
Null Model Improvement 47x PASS

RISK DISTRIBUTION RESULTS

Critical Risk
Score 76 to 100
35 providers (0.07%)
$420.2M excess billing
High Risk
Score 51 to 75
465 providers (0.93%)
$312.5M identified
Medium Risk
Score 26 to 50
9,440 providers (18.88%)
Below threshold
Low Risk
Score 0 to 25
40,060 providers (80.12%)
Normal patterns

GEOGRAPHIC DISTRIBUTION

State Critical Risk Providers Percentage
Florida 11 31%
California 5 14%
Texas 3 9%
Other States 16 46%

Geographic concentration aligns with DOJ Healthcare Fraud Strike Force priority regions, providing independent validation of detection methodology.

FEATURE IMPORTANCE (ML MODEL)

RISK SCORING COMPONENTS

METHODOLOGY HIGHLIGHTS

Six Federal Databases Dual Validation Peer Benchmarking 148 Records/Second

DATA SOURCES (6 FEDERAL DATABASES)

  • • CMS Medicare Provider Utilization (1.1M providers)
  • • NPPES National Provider Registry (8.2M providers)
  • • OIG LEIE Exclusion Lists (81,914 exclusions)
  • • PECOS Provider Enrollment (2.3M records)
  • • Medicare Geographic Variation (306 HRRs)
  • • Medicare Telehealth Trends Data

TECHNICAL PERFORMANCE

Processing Speed 148 records/second
Total Processing Time 5 minutes 47 seconds
Dataset Size 50,000 providers
Service Volume 322.4 million services
Data Completeness 99.7%

DATA QUALITY METRICS

Valid NPI Identifiers 99.97%
Complete Payment Data 98.47%
Geographic Matching 97.91%
Analysis Period 2019 to 2023
Random Seed 42 (reproducible)

VALIDATION SUMMARY

Temporal Stability

• Temporal Stability: 92.0% PASS
• Pattern consistency: 2019 to 2023
• Bootstrap: 1,000 iterations

Detection Performance

• Null Model Improvement: 47x
• Cross Validation: 91.3% ± 2.4%
• False Positive Rate: 0.0%
$420.2M EXCESS BILLING IDENTIFIED
Analysis of 50,000 Medicare telehealth providers covering 322.4 million services (2019 to 2023) identified 35 critical risk providers (0.07%) with billing patterns between 200% and 30,500% above peer benchmarks. The methodology achieved 91.3% accuracy with 100% precision in critical risk identification through dual validation across six federal databases.

Note: This system identifies statistical anomalies in billing patterns. Determination of actual improper payments requires investigation by authorized agencies.