From Fax to FHIR: Ingesting Handwritten and Multilingual Records for Risk Adjustment

Ask any risk adjustment leader where their missed HCCs hide, and the answer is rarely "the structured EHR fields." It is the faxed cardiology consult, the scanned discharge summary from another health system, the handwritten progress note, the lab report in a PDF, the record written in Spanish. Intelligence is only as good as the records behind it — and most of those records are unstructured and inconvenient.

A platform that only reads clean, structured data is coding with one eye closed. Here is what it takes to actually see everything.

Step 1: Get every document, automatically

Before you can read a record, you have to have it. The first job is automated ingestion: pulling every available patient document from your systems — faxes, scanned charts, prior records, labs, PDFs — without manual uploads or someone chasing down files. Completeness is the foundation; a missed page is a missed diagnosis.

Step 2: Read the hard stuff

This is where most tools quietly fail. Capturing a document is not the same as understanding it. Real-world ingestion has to handle:

• Scanned and faxed PDFs — low-quality images, skew, stamps, and overlapping text.
• Handwritten notes — provider shorthand and cursive that defeats traditional OCR.
• Non-English records — documents in Spanish and other languages, common across the communities our clients serve.
• Tables and layouts — medication lists, problem lists, and lab panels where structure carries meaning.

We use modern LLM vision models to interpret layout, tables, and text the way a human reader would — so a handwritten note or a Spanish-language consult becomes usable clinical evidence rather than an unsearchable image. The same approach powers our take on unstructured data as the biggest opportunity in risk adjustment.

Step 3: Structure, de-duplicate, and reconcile

Raw text is not enough. Ingested content is pre-processed into a clean, structured, de-duplicated patient record — so the same lab reported in three documents becomes one fact, and the full history lands on a single timeline. That reconciled record is what feeds every downstream recommendation, pre-visit summary, and audit trail.

Step 4: Make it searchable and traceable

Once a document is understood, two things become possible. First, blazing-fast search across every unstructured record — a Google-like bar over your entire patient archive, invaluable for CDI and audit response. Second, traceability: when a code is surfaced from a handwritten Spanish consult, a coder can click the evidence and see the exact highlighted passage in the original document. That source link is what makes the capture audit-defensible, no matter how messy the original.

Why this is a moat, not a feature

Any vendor can parse a clean CCD. Reading a stack of faxed, handwritten, multilingual records — and turning them into structured, searchable, traceable evidence — is the hard part, and it is exactly where the missed HCCs live. The completeness of what you can read sets the ceiling on everything else your risk adjustment program can do.

*Have a stack of "unreadable" documents? Schedule a live demo and we'll run real handwritten and multilingual records through MedChartScan live._