Definition

De-Identification

De-identification is the process of removing or transforming identifiers in health data so that it no longer identifies individuals. Data that meets the HIPAA de-identification standard is no longer protected health information.

2 min readReviewed September 14, 2026

Also known as: HIPAA de-identification, Safe Harbor method, Expert Determination method, De-identified data

Key facts

Standard
45 CFR 164.514(a) and (b)
Methods
Safe Harbor or Expert Determination
Safe Harbor list
18 categories of identifiers must be removed
Guidance
HHS Office for Civil Rights de-identification guidance, issued in 2012
Legal effect
De-identified data is outside the HIPAA Privacy Rule

What is de-identification?

HIPAA lets covered entities and business associates use and share health data without patient authorization once it is de-identified. The core test is whether there is a reasonable basis to believe the remaining data could identify a person.

De-identification differs from a limited data set. A limited data set removes direct identifiers but may keep dates and some geographic detail such as city, state and ZIP code, so it remains protected health information (PHI) and requires a data use agreement.

How HIPAA de-identification works

HIPAA recognizes two methods:

  • Safe Harbor: remove 18 types of identifiers, including names, geographic subdivisions smaller than a state, all date elements except year that relate to the individual, contact details, record and account numbers and full-face photos, and have no actual knowledge that the rest could identify someone.
  • Safe Harbor details: the first three digits of a ZIP code may be kept only if that area contains more than 20,000 people, and ages over 89 must be grouped into a single category of 90 or older.
  • Expert Determination: a qualified expert applies accepted statistical and scientific methods, concludes that the risk of identification by the anticipated recipient is very small, and documents the analysis.
  • Re-identification codes: a covered entity may assign a code for re-linking if the code is not derived from the individual's information and the method is not disclosed.

Why de-identification matters

Most commercial health data flows depend on it:

  • Real-world data: claims and EHR data licensed for research and commercial analytics is typically de-identified, often through expert determination.
  • Data linkage: tokenization links records about the same person across data sets without names, but linking can raise re-identification risk and is commonly evaluated under expert determination.
  • AI training: developers need a clear de-identification basis before training models on clinical notes, which can hide identifiers in free text.
  • Contracts and state law: data agreements often ban re-identification, and some state laws regulate de-identified data beyond HIPAA.

Sources

All glossary terms