Docs / Working cases

Uploading documents & how extraction works

Uploading

From a case’s Documents tab, upload PDFs of whatever you’ve collected — the written order, chart notes, face-to-face encounter notes, test results, delivery tickets. When you upload, you pick the document type; the AI also classifies each document independently and, when it’s confident, its classification takes effect — so a mislabeled upload usually self-corrects.

Types AffirmDME works with include SWO (standard written order), Chart notes, F2F notes (encounter documentation), and Other.

The Documents tab: uploaded files with classification and extraction status

What extraction does

Each document is read by AffirmDME’s document AI (running on AWS Bedrock — see how we handle PHI), which pulls the structured facts: patient identity, prescriber and NPI, order date and signature, encounter dates, diagnosis codes, sleep-study values (AHI/RDI), blood-gas values (SpO2/pO2), and more. Those extracted facts are what the checklist evaluates.

The Extraction tab shows what was read from each document, so you can always see why a requirement passed or failed.

Statuses and timing

Documents move Queued → Extracting → Complete (or Failed). Most documents finish in about a minute; the checklist re-evaluates automatically as each one completes, so the readiness score updates live.

Limits and re-runs

Up to 20 documents per case, and a bounded number of extraction runs per case — generous for real cases (typically 4–12 documents)
A failed or oddly extracted document can be reprocessed from its row on the Documents tab
Deleting a document removes its extraction and re-evaluates the checklist