CPT Studio

Side-by-side CPT code pricing comparison for healthcare payer data — load, compare, export.

Overview

CPT Studio is a Flask web application for comparing CPT (Current Procedural Terminology) code pricing across healthcare insurance payers. It ingests payer "Transparency in Coverage" Machine-Readable Files (MRFs), Excel fee schedules, and CSV price lists, normalizes the negotiated rates, and produces a clean side-by-side comparison that highlights price deltas, savings, and outliers.

The tool is built to handle the realities of payer data: deeply nested JSON, gzipped in-network rate files, and exports large enough (300 MB+) that they have to be streamed or processed in split parts rather than loaded into memory all at once.

Features

• Multiple ingestion paths — upload JSON / Excel (.xlsx/.xls) / CSV files, paste a direct payer URL, or load bundled demo data.
• Index-file parsing — extracts the list of in-network rate files from a payer Transparency-in-Coverage index, then fetches and decompresses the selected gzipped JSON on demand.
• Smart column detection — auto-detects CPT code, price, and description columns when loading Excel/CSV fee schedules.
• Large-file handling — streams files over 300 MB, supports paginated loading, and accepts multi-part (split) uploads stitched together as a single stream.
• Configurable comparison rules — compare by max, min, median, average, by billing class, or per-occurrence, with optional filtering by negotiated type and exclusion of expired rates (as-of a given date).
• Incremental comparison sessions — compare a streamed/multi-part source against a baseline part-by-part, with live status reporting.
• Local caching — downloaded MRF files are cached so repeated fetches of the same URL load almost instantly.
• CSV export — one-click export of a single source, a full comparison, or an incremental comparison result.
• Resilient error handling — detects and clearly reports expired signed payer URLs instead of failing silently.

Tech Stack

• Backend: Python 3.12, Flask
• Data: pandas, openpyxl (Excel), ijson (streaming JSON parsing), gzip
• HTTP: requests
• Frontend: Server-rendered Jinja template (single-page UI, vanilla JS, custom CSS)
• Serving / Deploy: gunicorn; deployable to Render (render.yaml) or Vercel (vercel.json + api/index.py)

Setup / Run

# 1. Clone
git clone https://github.com/ube09/PayerPrice-Compare.git
cd PayerPrice-Compare

# 2. (Optional) create a virtual environment
python -m venv .venv
source .venv/bin/activate   # Windows: .venv\Scripts\activate

# 3. Install dependencies
pip install -r requirements.txt

# 4. Run
python app.py

Then open http://localhost:5000.

Quick start

1. Load Source 1 — upload a payer index JSON (or paste its URL), then click an in-network file to fetch its CPT pricing. For very large exports, use the split-parts uploader.
2. Load Source 2 — upload an Excel or CSV fee schedule (columns like CPT Code / Code, Price / Rate / Amount, optional Description).
3. Compare — choose a comparison rule and view deltas, savings, and outliers side by side.
4. Export — download the comparison (or a single source) as CSV.

No live payer file? Click Load Test Insurance Data to load the bundled sample (test_pricing_data.json) and try the comparison flow immediately. A sample Excel/CSV fee schedule is included (sample_cpt_pricing.xlsx / sample_cpt_pricing.csv).

API endpoints (selection)

Method	Endpoint	Purpose
GET	/	Main interface
POST	/upload	Upload a JSON/Excel/CSV file or a URL
POST	/fetch_pricing	Fetch & parse a gzipped in-network rate file
POST	/load_test_data	Load bundled demo data
POST	/compare	Compare two loaded sources
POST	/stream_compare_upload	Stream-compare a large file vs a baseline
POST	/upload_multipart_part / /finalize_multipart	Split-part upload handling
POST	/load_paginated / /compare_paginated	Page-by-page loading & comparison
GET	/export_source_csv / /export_comparison_csv / /export_incremental_comparison_csv	CSV exports
GET	/sources	List loaded sources

Notes

• Payer signed URLs include Expires / Signature parameters and stop working once expired — download a fresh index file from the payer site if a link fails.
• CPT® is a registered trademark of the American Medical Association; this project is an independent analysis tool.

Author

Usama Bin Ejaz — AI/ML Engineer · Data Scientist · Founder, UBE Labs ("AI that ships.") GitHub: github.com/ube09 · Website: usamabinejaz.com

License

MIT