---
title: "Clinical Notes De-Identification (Date-Shifted)"
description: "Stop sending sensitive data to third-party APIs. Philterd ships open source PII and PHI redaction that runs inside your cloud. For HIPAA, GDPR, and CCPA work."
url: "https://philterd.ai/policies/healthcare/clinical-notes-deid/"
---

[← All policies](https://philterd.ai/policies/)

Healthcare · Philterd

# Clinical Notes De-Identification (Date-Shifted)

De-identify clinical notes for research, ML training, or analytics — preserving temporal relationships via per-patient date shifting.

[View policy →](https://philterd.ai/policies/healthcare/clinical-notes-deid/#policy-json) [Download JSON →](https://raw.githubusercontent.com/philterd/pii-redaction-policies/main/policies/philterd/healthcare/clinical-notes-deid.json) [View source on GitHub](https://github.com/philterd/pii-redaction-policies/blob/main/policies/philterd/healthcare/clinical-notes-deid.md)

**v2.0.0** Updated 2026-10-26 Philter `>=4.0.0` By Philterd

HIPAA PHI clinical notes date shifting research

## The policy

The full `clinical-notes-deid.json` file. The same content you’d get by downloading. Copy any part of it, or use the buttons in the hero to grab the whole file.

```
{
  "config": {
    "splitting": {
      "enabled": true,
      "threshold": 8000
    }
  },
  "ignored": [],
  "identifiers": {
    "person": {
      "phEyeConfiguration": {
        "modelPath": "ph-eye-pii-en-small",
        "labels": [
          "name"
        ]
      },
      "phEyeFilterStrategies": [
        {
          "strategy": "REDACT",
          "redactionFormat": "[PATIENT]",
          "condition": "confidence > 0.60"
        }
      ]
    },
    "physicianName": {
      "physicianNameFilterStrategies": [
        {
          "strategy": "REDACT",
          "redactionFormat": "[PROVIDER]"
        }
      ]
    },
    "date": {
      "onlyValidDates": true,
      "dateFilterStrategies": [
        {
          "strategy": "SHIFT",
          "shiftDays": 90
        }
      ]
    },
    "phoneNumber": {
      "phoneNumberFilterStrategies": [
        {
          "strategy": "REDACT",
          "redactionFormat": "[PHONE]"
        }
      ]
    },
    "ssn": {
      "ssnFilterStrategies": [
        {
          "strategy": "REDACT",
          "redactionFormat": "[SSN]"
        }
      ]
    },
    "emailAddress": {
      "emailAddressFilterStrategies": [
        {
          "strategy": "REDACT",
          "redactionFormat": "[EMAIL]"
        }
      ]
    },
    "hospital": {
      "hospitalFilterStrategies": [
        {
          "strategy": "REDACT",
          "redactionFormat": "[FACILITY]"
        }
      ]
    },
    "identifiers": [
      {
        "classification": "mrn",
        "pattern": "\\bMRN[\\s:#]*\\d{5,}\\b",
        "caseSensitive": false,
        "identifierFilterStrategies": [
          {
            "strategy": "REDACT",
            "redactionFormat": "[MRN]"
          }
        ]
      },
      {
        "classification": "age-over-89",
        "pattern": "\\b(?:9\\d|1[0-4]\\d)(?=[\\s-]*(?:years?|yrs?|y\\.?o\\.?)(?:\\b|$))",
        "caseSensitive": false,
        "identifierFilterStrategies": [
          {
            "strategy": "REDACT",
            "redactionFormat": "[AGE>89]"
          }
        ]
      }
    ]
  }
}
```

## Example

Input

```
Dr. Garcia saw Mr. Smith (MRN 47291) on 2025-03-14 for follow-up of his 2024-12-08 admission. Patient is 72 years old.
```

Output

```
[PROVIDER] saw [PATIENT] ([MRN]) on 2025-04-26 for follow-up of his 2025-01-19 admission. Patient is 72 years old.
```

## Entities this policy acts on

NAME PROVIDER DATE AGE PHONE SSN EMAIL HOSPITAL MRN

## What this policy does

Tailored for **research and ML use cases** where you need patient privacy *and* the temporal structure of clinical events to remain analyzable. Differs from the strict HIPAA Safe Harbor policy in three ways:

1. **Per-patient date shifting** instead of full date redaction. Each patient’s dates are shifted by the same random offset (±90 days), preserving intervals between events while breaking linkage to the actual calendar dates.
2. **Replacement tokens instead of full redaction** — `[PATIENT]`, `[PROVIDER]`, `[FACILITY]` — makes the text more readable for human reviewers and easier for downstream NLP.
3. **Confidence-gated name detection** (`confidence > 0.60`) reduces over-redaction of common English words that Philter’s NER occasionally misfires on in clinical text.

Ages under 90 are preserved (clinically relevant). Ages 90+ get the `[AGE>89]` token per HIPAA Safe Harbor §164.514(b)(2)(i)(C). Ages 90 and over are matched when written as a number followed by “years”, “yrs”, or “y.o.”, such as `93 years old` or `95-year-old`. A bare number, such as “she’s 93”, is not matched.

## When to customize

- **Date shift window.** ±90 days works for most cohort studies. For oncology timelines or longitudinal studies that span years, widen the window to ±365 or more.
- **Replacement tokens.** If your downstream pipeline expects specific tokens (e.g., spaCy’s `PERSON` or HuggingFace’s `[NAME]`), edit the `redactionFormat` fields.
- **Confidence threshold.** `> 0.60` is conservative. For higher-recall (catch more names at the cost of more false positives), lower to `> 0.40`. For higher-precision research datasets where false positives are worse than false negatives, raise to `> 0.80`.
- **MRN regex.** Same caveat as the Safe Harbor policy — adjust for your EHR’s format.

## When NOT to use this policy

- **For publication or sharing outside your covered entity.** This policy is *more* permissive than Safe Harbor — it keeps year-month-day structure (just shifted) and uses semantic tokens that are easier to re-identify than full redaction. For external sharing, use [hipaa-safe-harbor.json](https://philterd.ai/policies/healthcare/hipaa-safe-harbor/) instead.
- **For PCI or financial workloads.** Use the finance/ policies instead.

## Compliance notes

Date shifting is a recognized de-identification technique under the [HHS Expert Determination method](https://www.hhs.gov/hipaa/for-professionals/privacy/special-topics/de-identification/index.html#expert), but it does NOT meet HIPAA Safe Harbor on its own. Sharing data redacted with this policy outside the covered entity requires a qualified statistician’s certification under 45 CFR 164.514(b)(1).

For internal research use within the covered entity, this policy + an executed Business Associate Agreement (where applicable) is generally sufficient. Confirm with your IRB or compliance office.

## References

- [HHS Guidance on Expert Determination](https://www.hhs.gov/hipaa/for-professionals/privacy/special-topics/de-identification/index.html#expert)
- [Date Shifting Best Practices (NIST SP 800-188 draft)](https://csrc.nist.gov/publications/detail/sp/800-188/draft)

## Using this policy with Phileas

The name filter names `ph-eye-pii-en-small`, the model bundled with Philter 4.0. With Phileas, set its `modelPath` to the directory of a downloaded copy of the model, [philterd/ph-eye-pii-en-small](https://huggingface.co/philterd/ph-eye-pii-en-small).

## Use this policy

Download and load into your running Philter instance:

```
# Download the policy
curl -O https://raw.githubusercontent.com/philterd/pii-redaction-policies/main/policies/philterd/healthcare/clinical-notes-deid.json

# Upload it to Philter as clinical-notes-deid
curl -sk -X POST "https://localhost:8080/api/policies?name=clinical-notes-deid" \
     -H "Authorization: Bearer $PHILTER_API_KEY" \
     -H "Content-Type: application/json" \
     --data @clinical-notes-deid.json

# Redact text with it
curl -sk "https://localhost:8080/api/filter?p=clinical-notes-deid" \
     -H "Authorization: Bearer $PHILTER_API_KEY" \
     -H "Content-Type: text/plain" \
     --data-binary "your text here"
```

For Philter 4.0. No Philter instance yet? The [quick start](https://philterd.ai/dev/start/) runs one and shows where `PHILTER_API_KEY` comes from · Want to tune this policy against your data? [Talk to the team](https://philterd.ai/policies/healthcare/clinical-notes-deid/#).

[← All policies](https://philterd.ai/policies/)
