De-identification removes or transforms the information in a document that identifies a person, so the document can be used or shared without exposing them. For unstructured documents such as PDFs, Word files, text files and notes, that means finding names, dates, IDs and addresses inside free text and tables, then replacing them. Re-Doc replaces them with consistent synthetic data and keeps the original layout. Scanned pages and images are redacted with black boxes.
Illustrative example with invented data. The exact transformations, for example reducing dates to a year, follow the de-identification rules your team sets.
The goal is simple. The difficulty is that identifying details are scattered through free text, and one miss can undo the work.
“De-identification” is the US HIPAA term. “Anonymisation” is the EU GDPR term and is a stricter test: anonymous data is outside the GDPR only if people can no longer be identified by any means reasonably likely to be used. “Pseudonymisation” swaps identifiers for codes but keeps a way back, so the data stays personal data under the GDPR.
In a database, a name lives in one column. In a discharge summary, a contract or a letter, it appears in a sentence, a header, a signature block and a table, spelled in more than one way. Finding every instance is the hard part.
Blacking out text removes the information, but it also removes context. Reviews, research and AI workflows need documents that still read naturally. Replacing identifiers with synthetic values keeps the document usable.
If one person gets two different replacement names in one document, or the same replacement as someone else, the result is confusing or misleading. Replacement has to be consistent across the whole document and across a set of documents.
Re-Doc does the document-level work. Your team keeps the decisions and the review.
A PDF, DOCX, TXT file or a scan, in the web app or through the API.
Identifiers are found in the text and replaced with consistent synthetic data, or redacted where you need removal. Scans are redacted with black boxes. The layout stays as it was.
You check the result and edit anything that needs a human decision.
Download a document in the same format, ready for sharing, research or testing.
The same approach applies wherever a document has to travel without the person in it.
Discharge summaries, referral letters and clinical notes shared for research, quality review or training. See healthcare.
Reports prepared for public release or data sharing. See clinical trial anonymization.
Real documents are often too sensitive to feed into AI tools or test environments. De-identified copies keep realistic structure without the real people.
The HIPAA Privacy Rule, at 45 CFR 164.514(b), gives two ways to treat health information as de-identified.
Expert Determination. A person with appropriate knowledge of and experience with generally accepted statistical and scientific principles and methods determines that the risk is very small that the information could be used, alone or in combination with other reasonably available information, by an anticipated recipient to identify an individual, and documents the methods and results of the analysis.
Safe Harbor. The identifiers of the individual, and of relatives, employers and household members, listed below are removed, and the organisation does not have actual knowledge that the remaining information could be used alone or in combination with other information to identify the individual.
The 18 identifier types in the Safe Harbor method are:
Removing these types is not the whole test. The “actual knowledge” condition means free-text details outside the list can still matter, and the other route, Expert Determination, exists for cases where Safe Harbor is too blunt or not enough. Choosing the method, and signing off the result, belongs to your privacy officer or a qualified expert.
The words are used loosely. This is how they differ in practice.
For a longer walkthrough, read redaction vs anonymization vs pseudonymization. For technical guidance on de-identification techniques and governance, see NIST SP 800-188.
Reviewed by the Re-Doc team. Last reviewed 1 October 2026. This page is general information, not legal advice. Check the current text of the regulations, and your own obligations, with your privacy or legal counsel.
Most de-identification of documents is still done by hand, one page at a time. Re-Doc gives your reviewers a consistent first pass.
| # | Typical tools in the market | Re-Doc | Status |
|---|---|---|---|
| 01 | Reading every page and blacking out identifiers by hand. | Identifiers found across the whole document in one pass, then reviewed. | ● Covered |
| 02 | Black boxes that make text hard to read. | Realistic synthetic values that keep the text readable. | ● Covered |
| 03 | Different treatment of the same person in different places. | The same person gets the same replacement throughout. | ● Covered |
| 04 | Reformatting after the edit, or a flattened image. | Same file format and layout as the original. | ● Covered |
The first 10 pages are free in the web app. For larger sets or custom deployment needs, contact our team: we will run a scoped pilot on a sample and share results you can check.
