Handwritten Census to X12 Claim

This billing platform replaces manual transcription between hospital rounds and clearinghouse submission. Billing reviewers upload marked-up census PDFs: handwritten attending notes, strike-throughs, room and bed scrawls, ICD-10 shorthand. Our vision models turn every encounter on the page into structured data. The fields are patient, MRN, date of service, attending, diagnoses, codes, modifiers, units, and place of service.

Each field carries a confidence score and a box drawn back to the exact pixel it came from. The pipeline recognizes the common census layouts. It settles who the billing attending is from the whole page, not from a single row. Struck-through entries come through marked as rejected. The reviewer sees them, and they stay off the claim.

Clean typed pages go to a cheaper model, so the heavier model is saved for true handwriting. Each document has a fixed cost ceiling.

Eight checks run before anything is sent. They cover code-versus-diagnosis sanity, retired codes, ICD-10 validity against the live CMS dictionary, and fuzzy payer matching. They also cover continuity across multi-day stays, non-billable parent codes, CMS bundling and frequency edits, and a live provider-ID check. The last gate is a full round trip. The finished claim is re-read against the official 837P standard and held back if it will not pass.

Approved encounters then merge into an X12 837P file. One patient across several rounding days becomes a single claim with one line per day. When a payer responds, the tool reads the denial codes, sorts the disputes, and rebuilds the corrected claim. No one has to leave the tool.