ClearBill Project

ClearBill is a patient-facing web app that helps users understand and question medical bills. It identifies dispute-worthy patterns and explains them, providing recommendations for review, not determinations of correctness.

Tech Stack

• Next.js 14 (App Router)
• TypeScript
• Tailwind CSS
• Supabase (Postgres + Storage)
• Gemini (text + vision)
• Deployed on Vercel

Core Product Principles

• No chargemaster data
• No CMS fee schedule validation
• No insurance contract logic
• No claims of “correct” or “incorrect” pricing
• Outputs are recommendations for review, not determinations

Validation Logic (Allowed Signals Only)

• Duplicate or overlapping charges
• Vague or unclear descriptions
• Temporal inconsistencies (dates outside visit)
• Place-of-service inconsistencies
• LLM-based identification of illogical procedure combinations
• OCR confidence checks

Scoring

• Provide a Fairness Score that summarizes dispute-worthy patterns
• Score must be transparent and explainable
• Scoring reflects “worth reviewing”, not correctness

Dispute Letter

• Professional, neutral tone
• Requests clarification or itemization
• No legal threats
• Include user guidance on who typically receives the letter

Privacy

• Bills stored securely
• No third-party sharing

Project Phases

Phase 1: Upload Medical Bill & Extract Data

Goal: Enable users to upload medical bills (PDF, images) and extract key data using OCR. Scope:

• UI: Simple upload interface.
• Backend: API route to receive uploaded files.
• Service: OCR processing to extract text from the bill.
• Data Model (Initial): Basic structure to store raw uploaded file and extracted text. TODOs:
• Implement secure file storage (Supabase Storage).
• Integrate Gemini Vision for OCR.
• Design a more robust data model for extracted bill details.

Phase 2: Detect Dispute-Worthy Patterns & Explain Clearly

Goal: Analyze extracted data for dispute-worthy patterns and provide clear explanations. Scope:

• Backend: Service to apply validation logic.
• Service: Logic for Fairness Score calculation.
• UI: Display identified patterns, explanations, and Fairness Score. TODOs:
• Implement all allowed validation logic (duplicate charges, vague descriptions, temporal inconsistencies, etc.).
• Refine Fairness Score calculation and explanation.

Phase 3: Generate Dispute Letter & Guide Next Steps

Goal: Generate a professional dispute letter and guide the user on next actions. Scope:

• Backend: Service to generate dispute letter.
• UI: Display generated letter, options to edit, and instructions for sending. TODOs:
• Implement dispute letter generation using LLM.
• Develop comprehensive user guidance for next steps.