An LLM-powered compliance triage system that automates the analysis of Medicare Administrative Contractor (MAC) correspondence classifying letter type, extracting structured compliance data, generating prioritized action checklists, and drafting response letters.
Built to eliminate manual triage bottlenecks in Home health provider workflows, where missed deadlines carry direct financial and regulatory consequences.
Medicare Home health providers receive dozens of MAC letters monthly complaints, ADRs, overpayment demands, audit notices. Each has different deadlines, required documentation, and response strategies. Providers manually read each letter, decide priority, pull records, and draft responses.
Miss a 30-day complaint deadline → compliance violation.
Miss a 45-day ADR window → automatic claim denial.
Miss an overpayment response → interest accrual at ~11% annually.
This tool triages letters in seconds, not hours.
Given any MAC/CMS letter (pasted as text), the tool:
- Classifies the letter type: Complaint, ADR, Overpayment, Audit, Denial, or General
- Extracts structured fields: MAC contractor, DCN, HCPCS code, dollar amount at risk, deadline
- Prioritizes HIGH / MEDIUM / LOW with a one-line reason
- Generates a step-by-step action checklist specific to that letter type
- Drafts a professional response letter ready for compliance team review
- Saves structured JSON output for downstream tracking or dashboard use
============================================================
MEDICARE MAC LETTER ANALYSIS
============================================================
LETTER TYPE: COMPLAINT
MAC/CONTRACTOR: CGS DME MAC
JURISDICTION: Jurisdiction X
HCPCS CODE: XXXXX (Orthotic Equipment)
DEADLINE: XX days from receipt
** MEDIUM PRIORITY **
30-day hard deadline for equipment pickup and fax confirmation required.
ACTION CHECKLIST:
[ ] Contact beneficiary within X business days
[ ] Schedule equipment pickup appointment
[ ] Obtain beneficiary signature on pickup confirmation
[ ] Fax pickup documentation to Complaint Screening department
[ ] Log case closure in compliance tracker
DOCUMENTS TO PULL:
[ ] Original delivery documentation
[ ] Beneficiary signature on delivery
[ ] Equipment serial number record
CONTACT INFO:
Phone: XXXXXXX
============================================================
DRAFT RESPONSE LETTER
============================================================
[PROVIDER NAME]
[DATE]
CGS DME MAC Jurisdiction B
Complaint Screening Department
Re: Beneficiary Equipment Pickup Request
Dear CGS Complaint Screening Team,
We are writing to acknowledge receipt of your correspondence dated MM/DD/YYYY
regarding the above-referenced beneficiary and DCN.
We have initiated the equipment pickup process and are coordinating directly with
the beneficiary to schedule retrieval of the orthotic equipment.
Please contact [CONTACT NAME] at [PHONE] with any questions.
Sincerely,
[SIGNATURE]
[PROVIDER NAME]
============================================================
Requirements: Python 3.9+, Anthropic API key (get one here)
# Clone the repo
git clone https://github.com/Chidvy/Medicare-Mac-Analyzer.git
cd medicare-mac-analyzer
# Install dependencies
pip install -r requirements.txt
# Set your API key (Windows)
set ANTHROPIC_API_KEY=your_key_here
# Set your API key (Mac/Linux)
export ANTHROPIC_API_KEY=your_key_herepython analyzer.pyYou will be prompted to either:
- Select one of 3 built-in synthetic sample letters (complaint, ADR, overpayment demand)
- Paste your own letter text
Output prints to terminal and optionally saves to a timestamped JSON file.
| Type | Description |
|---|---|
| COMPLAINT | Beneficiary complaint — equipment pickup, billing dispute |
| ADR | Additional Documentation Request — records to support claim |
| OVERPAYMENT | Payer demanding repayment with interest timeline |
| AUDIT | Pre/post-payment audit — records review required |
| DENIAL | Claim denial with appeal window |
| GENERAL | Informational correspondence |
| Choice | Rationale |
|---|---|
| JSON-first extraction | Enables downstream use: dashboards, trackers, alerts |
| Two-pass architecture | Separate extraction from response generation for cleaner outputs |
| Synthetic samples built-in | Enables live demo without any real provider data |
| Priority classification | Mirrors real compliance triage logic (deadline + dollar amount) |
- Batch processing — analyze a folder of letters, output summary dashboard
- Deadline tracker — parse dates, calculate days remaining, sort by urgency
- Streamlit UI — drag-and-drop letter upload with visual priority queue
- Appeal letter generator — extend to full ADR appeal with medical necessity arguments
- Multi-payer support — extend beyond Medicare to Medicaid MACs and commercial payers
-Large Language Model API (Claude / OpenAI compatible)
- Python 3.9+
- Anthropic Python SDK
Durga Meduri — Business Analytics Manager | MS Business Analytics, UMass Boston
Built from direct observation of compliance workflows in healthcare operations.
LinkedIn | GitHub