Eastridge-Analytics/pharmaceutical-fraud-detection

Pharmaceutical Claims Fraud Detection

Detect co-pay card misuse patterns in pharmaceutical claims data using advanced graph analytics and network analysis.

Features

• —Multi-Pattern Detection: Identifies various fraud patterns including high-value claims, impossible travel, and suspicious frequency patterns
• —Interactive Network Graphs: Visualize relationships between cards, patients, pharmacies, and drugs
• —Real-time Analytics Dashboard: Monitor claims volume, amounts, and distributions
• —CSV Upload & Processing: Robust file handling with multiple fallback methods
• —Export Capabilities: Download fraud detection reports and analysis results

Fraud Patterns Detected

1. 1.High Value Claims: Claims exceeding typical reimbursement amounts
2. 2.High Frequency Usage: Cards used multiple times per day across different locations
3. 3.Impossible Travel: Same card used at geographically distant locations within unrealistic timeframes
4. 4.Widespread NDC Distribution: Drugs appearing across many different pharmacies (potential diversion)
5. 5.Network Analysis: Complex relationships between suspicious entities

Usage

1. 1.Upload Data: Drag and drop your pharmaceutical claims CSV file
2. 2.Configure Parameters: Adjust detection thresholds for your use case
3. 3.Run Analysis: Execute fraud detection algorithms
4. 4.Explore Results: Review detected patterns and network visualizations
5. 5.Export Reports: Download detailed fraud analysis reports

Expected Data Format

Your CSV should include these columns:

Required:

• —claim_id: Unique claim identifier
• —card_id: Insurance card identifier
• —patient_id: Patient identifier
• —pharmacy_id: Pharmacy identifier
• —paid_amount: Amount paid for the claim
• —claim_dt: Claim date/timestamp

Optional (for enhanced analysis):

• —ndc: National Drug Code
• —qty: Quantity dispensed
• —pharmacy_name: Pharmacy name
• —city, state: Location information
• —lat, lon: Geographic coordinates

Technology Stack

• —Streamlit: Interactive web application framework
• —NetworkX: Graph analysis and community detection
• —Plotly: Interactive data visualizations
• —Pandas: Data manipulation and analysis
• —st-link-analysis: Network graph visualization

Use Cases

• —Insurance Fraud Investigation: Identify suspicious claim patterns
• —Regulatory Compliance: Monitor for potential violations
• —Risk Assessment: Evaluate fraud risk across different entities
• —Network Analysis: Understand relationships between fraudulent actors
• —Operational Intelligence: Optimize fraud detection processes

Enterprise Applications

This tool demonstrates core capabilities that can be scaled for enterprise fraud detection systems including:

• —Real-time fraud scoring
• —Integration with existing claims processing systems
• —Advanced machine learning models
• —Distributed processing for large datasets
• —Automated alert systems

Contact Eastridge Analytics for enterprise implementations and custom fraud detection solutions.