RxHCC FWA detection Agent https://rxhccagent.streamlit.app/

Kaggle 5-Day AI Agents Intensive Vibe Coding Capstone — Agents for Good

An autonomous multi-agent system that investigates Medicare Part D claims for Fraud, Waste & Abuse (FWA) before payment is released — using the Google Gen AI SDK + Gemini 2.0 Flash. https://rxhccagent.streamlit.app/

Architecture

User submits claim
      │
      ▼
fwa_orchestrator  (Gemini 2.0 Flash)
  ├── claim_analyzer   → lookup_icd10_code
  ├── risk_scorer      → calculate_rxhcc_risk_score, check_drug_combination, get_provider_billing_history
  └── report_writer    → generate_fwa_report

Fraud Patterns Detected

• Gender-Diagnosis Mismatches: e.g., Male patient billed for female-specific conditions.
• Controlled Substance Cocktails: Opioid + benzodiazepine + muscle relaxant "holy trinity" (pill-mill signature).
• Billing Anomalies: Provider billing in the 99th percentile versus peers for controlled substances.
• Claim Cost Mismatches: Claim amounts 3–10x the provider's own historic average.

Quick Start

1. Clone the repository:

   git clone https://github.com/sechan9999/rxhcc_agent.git
   cd rxhcc_agent

2. Install dependencies:

   pip install -r requirements.txt

3. Set up environment variables:

   export GOOGLE_API_KEY="your_api_key_here"

4. Run the Streamlit application:

   streamlit run app.py

Repository Structure

File	Description
tools.py	5 ADK tool functions (ICD-10 lookup, provider history, drug combo checker, risk scorer, report generator)
agents.py	4 Google ADK agents & Orchestration logic
app.py	Streamlit demo user interface with 3 pre-loaded FWA investigation scenarios
sample_claims.json	4 labeled test claims for validation
requirements.txt	Python library dependencies
KAGGLE_WRITEUP.md	Capstone writeup for the Agents for Good track

Track

Agents for Good — Medicare fraud costs $60–100B/year. This agent system catches potential fraud before the money leaves.