Telemedicine Transcript De-Identification with anonym.plus

Pull every spoken identifier out of a consultation transcript without an upload.

In simple terms, PII redaction is the on-device process of finding and masking personally identifiable information in a document before it is shared.

Transcript de-identification is the removal of personal identifiers from a recorded telemedicine consultation. The spoken clinical detail is special category data under UK GDPR Art. 9(1). DPA 2018 Schedule 1, Part 1, paragraph 2 supplies the health or social care condition a clinic relies on to hold it. GMC guidance on making and using visual and audio recordings of patients expects consent before the call is captured at all. anonym.plus clears the identifiers on your own device. The dialogue stays readable, but it no longer names anyone.

When this applies

Spoken consultations turn into long transcripts full of names said out loud. Once a recording is kept, it forms part of the health record and falls under the Records Management Code retention schedule. Before you reuse a transcript for audit, teaching, or model training, those spoken identifiers have to come out.

How anonym.plus handles it

  1. Open the typed transcript in anonym.plus on your device.
  2. Local OCR reads pasted screenshots of the dialogue too.
  3. The tool marks names, dates, places, and contact details.
  4. Skim each flag and fix any clinical term caught wrongly.
  5. Swap each identifier for a safe label, or black it out.
  6. Store the clean text. The source never leaves your machine.

What you need to provide

Patient data entity types detected

Categoryanonym.plus entity typeExample
NamesPERSON“Hi, this is Chloe” → [PATIENT]
DatesDATE_TIME“since last Tuesday” → [DATE]
ContactPHONE_NUMBERcallback +44 161 496 0182 → [PHONE]
ContactEMAIL_ADDRESSchloe@example.co.uk → [EMAIL]
LocationLOCATION“I live in Bristol” → [PLACE]
NHS numberNHS_NUMBERNHS 485 777 3366 → [NHS_NO]

Compliance achieved

Anonymise telemedicine transcripts offline — see plans & start free →

Limitations & cautions

Speech gets typed loosely. Misheard names and unusual spellings can slip past. Always check the flags before export. Auto-typed audio is the riskiest source, because the speech-to-text step adds errors of its own.

Frequently asked questions

Are spoken names harder to catch than typed ones?

A little. Speech-to-text can mangle a name, which then reads oddly in the transcript and gives the model weaker cues. Clear names are flagged well. Review the rest by hand, above all on auto-typed audio where the recogniser has already introduced its own errors.

Does an offline tool need a processor contract?

No. anonym.plus runs on your own device, so no outside party receives the transcript and UK GDPR Art. 28 is not engaged. Sending the same file to a cloud redaction service would make that vendor a processor and require exactly the written terms Art. 28(3) lists.

Can I delete the original recording once the clean copy exists?

Not automatically. The Records Management Code of Practice for Health and Social Care 2021 sets the retention period for the record itself, and a de-identified copy does not replace it. Treat the clean transcript as an extra artefact for teaching or audit, not as the record.