Medical Benefit Drug Policy Tracker

🧠 Medical Benefit Drug Policy Tracker

** 🚀 Inspiration **

The healthcare system works smoothly for pharmacy benefits, but completely breaks down when it comes to medical benefit drugs. Instead of structured data, everything lives inside long, inconsistent policy PDFs.

Analysts today manually read multiple documents just to answer simple questions like “Is this drug covered?” or “What criteria apply?”. This process is slow, repetitive, and error-prone.

We built this project to eliminate that friction and turn static documents into instant answers.

🏗️ What We Built

We created an AI-powered system that ingests medical policy documents and converts them into structured, comparable insights.

Our platform allows users to:

Search drug coverage across multiple payers Compare policies side-by-side Extract key details like prior authorization and step therapy Ask natural language questions and get direct answers

Instead of reading documents, users interact with clean, actionable data.

⚙️ How We Built It

Our pipeline works as follows:

Document ingestion from policy PDFs Text extraction and cleaning NLP-based entity extraction Data normalization across payers Query and comparison interface Unstructured Policies→Parsed Text→Structured Data→Searchable Insights Unstructured Policies→Parsed Text→Structured Data→Searchable Insights

We focused on extracting high-value fields like drug names, coverage criteria, and authorization requirements, then standardizing them for comparison.

⚔️ Challenges We Faced

Inconsistent formats Every payer structures policies differently, making parsing unreliable.

Complex medical language Policies are written in dense clinical terms that are difficult to interpret programmatically.

Normalization problem The same rule can be written in completely different ways across payers, making comparison hard.

Noise in documents A large portion of policy text is irrelevant, so isolating meaningful information was a key challenge.

📚 What We Learned Real-world AI is more about data quality than model complexity Normalization is harder than extraction Simplicity in UI is critical for usability Solving a real workflow problem matters more than building flashy features 💡 What’s Next* Real-time policy change tracking Expansion to more payers and policies Improved clinical reasoning and comparisons Integration into healthcare decision workflows

🎯 Final Thought

This project transforms medical policy analysis from a manual, document-heavy task into an instant, intelligent experience—saving time and enabling better decisions at scale.

Built With

• fastapi
• html
• llm
• pandas
• python
• react