Not My Debt

Inspiration

Medical debt is often confusing because the truth is spread across bills, insurance explanations, receipts, and collection notices. We wanted to help patient advocates understand both the individual evidence and the community context surrounding medical-debt collection pressure.

What it does

Not My Debt helps advocates:

Explore reported medical-debt complaint patterns across U.S. states.
Compare complaint rates with population, poverty, and uninsured rates.
Review nearby hospital and financial-assistance information.
Reconcile bills, EOBs, receipts, and collection notices.
Identify possible discrepancies without making unsupported legal conclusions.
Generate a source-linked evidence packet for review.

How we built it

We built a Next.js and React interface with TypeScript, Recharts, and a bundled SVG state map. The evidence engine is written in Python and connects to the frontend through stateless API routes.

The Community Explorer uses versioned snapshots from:

CFPB 2025 medical-debt complaints
2024 ACS five-year estimates
CMS hospital directory data
Three reviewed North Carolina assistance examples

The case workflow uses fictional documents for the demo and keeps case data in memory.

Challenges we ran into

The biggest challenge was preserving accuracy while combining datasets with different definitions, time periods, identifiers, and geographic boundaries.

We also had to handle ambiguous billing evidence carefully. An EOB describes patient responsibility, but it does not prove payment. A complaint may identify a collector without identifying the hospital involved. These distinctions shaped both the data model and the interface language.

Accomplishments that we're proud of

We created a working end-to-end flow from community exploration to individual evidence review.

The app includes:

A national complaint-rate map
Linked state trends and comparison charts
Searchable hospital directories
Reviewed assistance resources
A printable community briefing
Evidence reconciliation with source passages
Conservative handling of missing, duplicate, and conflicting records
A response packet generator

The project passes 153 automated tests plus Python linting, TypeScript checks, ESLint, and a production build.

What we learned

Public complaint data is useful for finding patterns, but it is not a complete measure of harm or proof that a hospital caused a problem.

We also learned that clear provenance matters as much as the visualization. Every measure needs its period, denominator, source, and limitations stated directly where people use it.