De-identifying clinical text

openmed.deidentify detects PHI/PII and rewrites the text so it can be shared, stored, or analyzed without exposing patients. It runs fully on-device after a one-time model download, no network calls, no telemetry, no raw PHI leaving the process. This is the single most important OpenMed entry point for privacy work; everything else (policies, audit, multilingual, date-shifting) layers on top of it.

When to use this skill

Reach for deidentify when you need to transform text, replace, mask, remove, hash, or date-shift the identifiers. If you only need to locate PHI spans without changing the text, use extract_pii (see extracting-pii-entities). To restore masked text later, use reidentify (see reidentifying-text).

Quick start

import openmed

note = (
    "Patient John Doe (MRN 1234567) was seen on 2024-03-02 by Dr. Alice Reed. "
    "Contact: john.doe@example.com, 617-555-0142."
)

result = openmed.deidentify(
    note,
    method="mask",                 # mask | remove | replace | hash | shift_dates
    confidence_threshold=0.7,      # safety default; raise to reduce false negatives' impact
    policy="hipaa_safe_harbor",    # optional bundled profile (see below)
)

print(result.deidentified_text)
# Patient [NAME] (MRN [ID_NUM]) was seen on [DATE] by Dr. [NAME]. ...

for e in result.pii_entities:
    # NEVER log e.text / e.original_text, those are raw PHI. Use offsets + label.
    print(e.canonical_label, e.start, e.end, round(e.confidence, 3))

deidentify returns a DeidentificationResult with these fields (note the exact names):