# How to anonymize insurance claim files

A claim file is a claim form, an EOB, letters, IDs and records that all refer to the same claimant and member numbers. Testing a claims agent on it means anonymizing the whole file consistently.

## What matters for claim files

- **One identity per person across the file.** The claimant's name, policy and member numbers must be replaced the same way on the claim form, the EOB and every letter, or cross-document checks stop working.
- **Keep the clinical and financial substance.** Procedures, diagnoses, amounts and dates of service are often what the workflow tests. Replace identities, not the facts the task depends on.
- **Keep the documents usable.** Scans stay scans, PDFs stay PDFs, layouts stay the same, so OCR and extraction behave as in production.
- **Review.** A reviewer should see every change, override where needed, and confirm before the file is shared.

## Doing it with CasePatch

Upload the claim documents at [the CasePatch home page](/) (PDF, PNG or JPG; up to 10 pages per document). CasePatch detects names, member and account numbers, addresses, dates of birth, phone numbers, signatures and other identifiers, replaces them in place with format-valid synthetic values, and keeps one replacement per person across the case. The public Medicare Summary Notice and card sample on the home page shows a worked example.

Use test or specimen documents in the hosted beta, and run CasePatch in your own environment for production claim files. CasePatch's output is not a compliance determination; review every page. See also [healthcare documents](/guides/de-identify-healthcare-documents.html) and [AI agent evals](/guides/anonymize-production-data-for-ai-agent-evals.html).
