> ## Documentation Index
> Fetch the complete documentation index at: https://docs.glialhealth.com/llms.txt
> Use this file to discover all available pages before exploring further.

# De-identification

> Automatically mask PHI before results leave the Folio pipeline.

When you set `deidentify=true` on a submission, Folio detects and masks protected health information (PHI) across all surfaces of the extracted result — field values and table cells — before anything is stored or returned. The original document bytes are never stored after masking completes.

## Enabling de-identification

Add `deidentify=true` as a form field:

```bash theme={null}
curl -X POST https://api.glialhealth.com/v1/documents \
  -H "Authorization: Bearer sk_test_..." \
  -F file=@patient_record.pdf \
  -F document_type=lab_report \
  -F deidentify=true
```

The result's `deidentified` field will be `true` to confirm masking was applied.

## Typed-token masking

Each detected PHI entity is replaced with a typed token that preserves the category of information without exposing the actual value:

| Token       | PHI category                                       |
| ----------- | -------------------------------------------------- |
| `[NAME]`    | Person names                                       |
| `[DATE]`    | Dates (birth dates, test dates, appointment dates) |
| `[ADDRESS]` | Street addresses, postal codes, cities             |
| `[PHONE]`   | Phone and fax numbers                              |
| `[EMAIL]`   | Email addresses                                    |
| `[ID]`      | Patient IDs, health card numbers, MRNs, SSNs       |
| `[OTHER]`   | Any other PHI that doesn't fit a specific category |

### Before / after example

**Before de-identification (`deidentify=false`):**

```json theme={null}
{
  "extract": {
    "patient_name": {
      "value": "Jane Doe",
      "confidence": 0.99,
      "found": true,
      "page": 1,
      "bbox": {"x": 0.05, "y": 0.12, "width": 0.15, "height": 0.02}
    },
    "date_of_birth": {
      "value": "1985-03-22",
      "confidence": 0.97,
      "found": true,
      "page": 1,
      "bbox": {"x": 0.05, "y": 0.14, "width": 0.11, "height": 0.02}
    }
  },
  "deidentified": false
}
```

**After de-identification (`deidentify=true`):**

```json theme={null}
{
  "extract": {
    "patient_name": {
      "value": "[NAME]",
      "confidence": 0.99,
      "found": true,
      "page": 1,
      "bbox": {"x": 0.05, "y": 0.12, "width": 0.15, "height": 0.02}
    },
    "date_of_birth": {
      "value": "[DATE]",
      "confidence": 0.97,
      "found": true,
      "page": 1,
      "bbox": {"x": 0.05, "y": 0.14, "width": 0.11, "height": 0.02}
    }
  },
  "deidentified": true
}
```

Bounding boxes, confidence scores, and `found` flags are preserved — only the `value` is masked.

## Masking scope

De-identification applies to every surface of the result:

* **`extract` field values** — each field's `value` string is scanned and PHI tokens are substituted.
* **`tables` cell values** — every cell in every extracted table is scanned.

## Secondary-use and regulatory context

De-identification is designed to support secondary use of health data under Quebec Law 25 (Act to modernize legislative provisions as regards the protection of personal information) and similar frameworks that require personal information to be anonymised or de-identified before use for research, analytics, or model training. Consult your privacy and legal team to confirm that Folio's token-based approach meets the specific de-identification standard required for your use case.

<Note>
  De-identification is performed on extracted text, not on the document image itself. If you retain the original PDF or image file outside of Folio, those files still contain the original PHI.
</Note>