Know a company with years of records? Refer it and earn up to $100K
Health data resource

HIPAA 18 Identifiers: The Full List and How to Remove Them

All 18 HIPAA identifiers quoted from 45 CFR 164.514(b)(2), what each covers in a record, the actual knowledge rule, and identifiers in text, images and audio.

Short answerChecked

The 18 HIPAA identifiers are what the Safe Harbor method in 45 CFR 164.514(b)(2) requires removing: names; geographic units smaller than a state; date elements except the year, and ages over 89; phone and fax numbers; email addresses; Social Security numbers; medical record, health plan beneficiary and account numbers; certificate and license numbers; vehicle and device identifiers; web URLs; IP addresses; biometric identifiers; full-face photos; and any other unique number, characteristic or code.

Where the list is
45 CFR 164.514(b)(2)(i)items (A) to (R)
Whose identifiers
Patient and familyplus employers and household members
Second condition
No actual knowledge164.514(b)(2)(ii)
A dispatch clipboard and barcode scanner in a warehouse office
Key points
  • The list covers the patient and the patient's relatives, employers and household members.
  • Removing all 18 is not enough on its own: Safe Harbor also requires no actual knowledge that what remains could identify someone.
  • Partial identifiers fail. Initials, the last four digits of a Social Security number and unsalted hashes still count.
  • Identifiers in notes, burned into images and spoken in audio count the same as those in labeled fields.

What are the 18 HIPAA identifiers?

The 18 HIPAA identifiers, also called the 18 PHI identifiers, are the items the Safe Harbor method of de-identification requires removing from health information. They come from 45 CFR 164.514(b)(2)(i), which requires that "The following identifiers of the individual or of relatives, employers, or household members of the individual, are removed" (45 CFR 164.514). The list reaches past the patient: a spouse's phone number or an employer's name in a note counts too.

Removing them is half of Safe Harbor. The other half is that the covered entity "does not have actual knowledge that the information could be used alone or in combination with other information to identify an individual who is a subject of the information." The other route, Expert Determination, uses no fixed list; HIPAA de-identification compares the two methods.

An identifier is not PHI on its own. HHS explains that "Identifying information alone, such as personal names, residential addresses, or phone numbers, would not necessarily be designated as PHI", and gives a phone book as the example (HHS OCR guidance). A name becomes PHI when it sits next to information about a person's health, care or payment for care.

For trial data, removing these identifiers is only one condition: clinical trial data and AI licensing covers the consent forms, IRB rules and trial agreement that also decide what a sponsor may do.

A token from a linkage service is a code too. Our Datavant trust check quotes what that firm says about its tokens and HIPAA de-identification.

What does each of the 18 PHI identifiers cover?

The middle column quotes 45 CFR 164.514(b)(2)(i)(A) to (R) as the regulation words them (eCFR). The right column shows where each one turns up in clinical records.

#Identifier, as the regulation words itWhere it turns up in health records
1NamesPatient, guarantor and emergency contact fields; note headers; letters; scanned forms
2All geographic subdivisions smaller than a State, including street address, city, county, precinct, zip code, and their equivalent geocodes, except for the initial three digits of a zip code if, according to the current publicly available data from the Bureau of the Census: (1) The geographic unit formed by combining all zip codes with the same three initial digits contains more than 20,000 people; and (2) The initial three digits of a zip code for all such geographic units containing 20,000 or fewer people is changed to 000Home and billing addresses, geocodes added for analytics, towns named in notes
3All elements of dates (except year) for dates directly related to an individual, including birth date, admission date, discharge date, date of death; and all ages over 89 and all elements of dates (including year) indicative of such age, except that such ages and elements may be aggregated into a single category of age 90 or olderBirth, visit, admission, discharge, procedure, lab and prescription dates; ages written in notes
4Telephone numbersDemographics, call logs, after-visit summaries
5Fax numbersReferral and records-request cover sheets
6Electronic mail addressesPatient portal accounts, appointment reminders
7Social security numbersRegistration, billing and insurance records
8Medical record numbersEHR exports, specimen labels, report headers, file names
9Health plan beneficiary numbersInsurance cards, eligibility checks, claims
10Account numbersBilling statements, payment plans, collections
11Certificate/license numbersCopies of driver's licenses taken at registration
12Vehicle identifiers and serial numbers, including license plate numbersAccident and injury intake notes
13Device identifiers and serial numbersImplant records, pacemaker and pump serial numbers, home monitoring devices
14Web Universal Resource Locators (URLs)Links to personal pages or profiles in messages and notes
15Internet Protocol (IP) address numbersPatient portal and telehealth session logs
16Biometric identifiers, including finger and voice printsVisit audio, voicemail, fingerprint check-in systems
17Full face photographic images and any comparable imagesChart photos, facial dermatology and wound photos, 3D head renderings
18Any other unique identifying number, characteristic, or code, except as permitted by paragraph (c) of this sectionClinical trial record numbers, barcodes, unsalted hashes, a rare job title

How do ZIP codes, dates and ages work under Safe Harbor?

Items 2 and 3 are the ones most often handled wrong, because part of each may stay.

  • ZIP codes. The first three digits may stay only where the area they cover "contains more than 20,000 people" under current Census data; otherwise they become 000. HHS published 17 restricted three-digit codes from 2000 Census data, and says covered entities "should not, however, rely upon this listing or the one found in the August 14, 2002 regulation if more current data has been published" (HHS OCR guidance).
  • Dates. Every date element below the year goes, for any date tied to the patient. In HHS's example, January 1, 2009 can appear only as 2009. Test dates are included: "Dates associated with test measures, such as those derived from a laboratory report, are directly related to a specific individual and relate to the provision of health care."
  • Ages over 89. These collapse into one category of 90 or older, and so do dates that imply them. In HHS's example, a patient born in 1910 and seen in 2010 has the birth year reported as on or before 1920.

These rules are why Safe Harbor data cannot answer timing questions, such as days from diagnosis to treatment. An Expert Determination can keep shifted dates or intervals instead; see HIPAA de-identification.

What counts as "any other unique identifying number, characteristic, or code"?

Item 18 is the catch-all. HHS gives examples of each part:

  • Numbers: "Clinical trial record numbers are included in the general category" of other unique identifying numbers.
  • Codes: "a code derived from a secure hash function without a secret key (e.g., "salt") would be considered an identifying element." EHR and e-prescribing barcodes count too, because they "are often designed to be unique for each patient, or event in a patient's record".
  • Characteristics: an occupation can identify someone, such as a record listing a patient as "current President of State University."
  • Parts of identifiers: a data set with "patient initials, or the last four digits of a Social Security number" fails Safe Harbor.

One kind of code is allowed. 45 CFR 164.514(c) lets a covered entity keep a re-identification code if "The code or other means of record identification is not derived from or related to information about the individual and is not otherwise capable of being translated so as to identify the individual", and it "does not use or disclose the code or other means of record identification for any other purpose, and does not disclose the mechanism for re-identification" (45 CFR 164.514(c)). A random study number in a locked lookup table fits. HHS says that under Safe Harbor "a code or other means of record identification that is derived from PHI would have to be removed", so a hash of the medical record number does not.

What does the actual knowledge condition mean?

HHS defines actual knowledge as "clear and direct knowledge that the remaining information could be used, either alone or in combination with other information, to identify an individual who is a subject of the information" (HHS OCR guidance). If the organization knows the stripped data still points to someone, Safe Harbor fails, even with all 18 identifiers gone. HHS gives four examples:

  • A revealing occupation left in the record, such as a well-known public role.
  • A recipient with a relative in the data: a researcher who "had a family member in the data" and enough context to recognize the case.
  • A rare event that made the news: "Rare clinical events may facilitate identification in a clear and direct manner."
  • Being told the recipient can re-identify: the covered entity then "has actual knowledge because it was informed outright that the recipient can identify a patient".

Knowing that re-identification research exists is not enough on its own: HHS says "mere knowledge of these studies and methods, by itself," does not create actual knowledge.

Doctor and staff names are a related question. "There is no explicit requirement to remove the names of providers or workforce members of the covered entity or business associate", HHS says, but the covered entity "would need to consider whether additional personal names contained in the data should be suppressed to meet the actual knowledge specification."

Where do identifiers hide in free text, images and audio?

Labeled fields are the easy part. HHS says "an identifier listed in the Safe Harbor standard must be removed regardless of its location in a record if it is recognizable as an identifier", and warns that "Clinical narratives in which a physician documents the history and/or lifestyle of a patient are information rich and may provide context that readily allows for patient identification" (HHS OCR guidance).

Signals can identify people as well: the ECG and EEG guide summarizes the studies, and the genomic data guide covers DNA.

Record typeWhere identifiers hideWhat a careful pass does
Notes and lettersNames of relatives and employers, towns, dates written in prose, strings such as 5/97 that may or may not be datesAutomated detection, then a person reads a sample and judges unclear strings in context
Scanned documents and PDFsLetterheads, signatures, fax headers, handwritingText recognition before redaction; image-only pages reviewed by hand
Medical images (DICOM)Header fields, text burned into pixels, faces in 3D head and neck renderingsA DICOM confidentiality profile plus pixel cleaning
Clinical photosFaces and other features that could identify a personCrop or mask; full-face and comparable images must go
Audio and videoSpoken names, dates and places; the voice itselfRemove spoken identifiers from transcript and audio; treat the voice as an identifier
File names and metadataRecord numbers and names in file names, folder paths and document propertiesRename from a random key; strip document metadata

Context decides some cases. In HHS's examples, a note saying a patient is "the Executive Vice President of the state university" must be cleaned, while a stand-alone "Newark, NJ" in intake notes "may be retained in the data" because it is unclear what the place refers to. HHS also asks: "is 5/97 a date or a ratio?"

In images, DICOM says burned-in annotation is routine in ultrasound, and that head and neck renderings may allow identification (DICOM PS3.15 E.3; E.3.2). HIPAA de-identification covers the DICOM profile in detail.

Audio is the hardest case, because the list includes "Biometric identifiers, including finger and voice prints". Removing names from a transcript does not de-identify the recording itself. For visit recordings in particular, see telehealth visit recordings.

18 identifiers example: checking one discharge summary (illustrative)

An illustration written for this page: a check of a fictional discharge summary. Bracketed placeholders stand in for values; no real patient or record is shown.

Text in the summaryIdentifierSafe Harbor action
"[NAME] was admitted on [DATE]"1 Names; 3 DatesRemove the name; keep the year only
"MRN [NUMBER]" in the page header8 Medical record numbersRemove; use a random study code if records must stay linked
"Lives with her daughter, a teacher in [TOWN]"2 Geographic subdivisions; a relative's detailsRemove the town; check whether the family detail still points to one person
"Age 93"3 Ages over 89Replace with "90 or older"
"Pacemaker serial [NUMBER] checked"13 Device identifiersRemove the serial; keep "pacemaker checked"
"Follow up by phone at [NUMBER]"4 Telephone numbersRemove
"Seen by Dr. [NAME], cardiology"Not on the list (provider name)May stay; remove if it narrows the patient down
"Former mayor of [TOWN]"18 Unique characteristic; actual knowledgeRemove

After the pass, a second reviewer reads a sample of summaries, because the most identifying detail here (the former mayor) is on no field list.

Common mistakes with the 18 HIPAA identifiers

Most misses come from reading the list as a set of database columns.

  • Removing identifiers for the patient only, and leaving names and numbers of relatives, employers or household members.
  • Keeping the month and day of visit or lab dates because they are "not birth dates".
  • Keeping five-digit ZIP codes, or three-digit codes for areas of 20,000 people or fewer.
  • Leaving ages over 89, or birth years that imply them.
  • Treating initials, partial Social Security numbers or unsalted hashes as safe.
  • Cleaning database fields but not notes, scanned letters, file names or image pixels.
  • Removing names from a transcript while keeping the original voice recording.
  • Using the 2002 restricted-ZIP list instead of current Census data.

Are de-identified health records also a data asset?

They can be, once the identifiers are handled. AI data buyers have asked Origin for longitudinal EHR records (patient records over time with diagnoses, medications, labs and outcomes), imaging with reports (CT, MRI, X-ray and pathology slides), behavioral health and specialty care records, and, as the newest request, doctor-patient conversation audio with transcripts. A buyer has also told Origin what these deals need to get through: de-identification signed off by an independent third party under HIPAA's Expert Determination method, not by the seller; IRB approval, or patient consent that covers commercial licensing; and a clear chain of rights showing who collected the data, who owns it and who can license it.

What must come out first: all 18 identifiers wherever they sit, including notes, burned-in image text and voices; any detail the practice knows could still single out a patient; and records the practice cannot show it has the right to share. Safe Harbor's year-only dates also remove the timing in longitudinal records, so the choice of method matters before any export. See health data licensing requirements and licensing patient records.

Origin does not perform Expert Determination or IRB review. Origin is an independent data partner: it helps a practice work out what it holds, which of the three requirements it already meets and what is missing. Your practice decides what is shared and signs any license itself, and a buyer may pay Origin when a deal closes.

Does your company hold records like these? See what your data is worth.

Common questions

What are the 18 PHI identifiers?

Names; geographic subdivisions smaller than a state; all date elements except the year, and ages over 89; telephone numbers; fax numbers; email addresses; Social Security numbers; medical record numbers; health plan beneficiary numbers; account numbers; certificate and license numbers; vehicle identifiers; device identifiers; web URLs; IP addresses; biometric identifiers such as finger and voice prints; full-face photos and comparable images; and any other unique identifying number, characteristic or code.

Are dates of service one of the 18 HIPAA identifiers?

Yes. The list covers all elements of dates except the year for any date directly related to the patient, which includes visit, admission, discharge, procedure and lab test dates. HHS says test dates count because they relate to the patient and to the care given. Under Safe Harbor only the year may stay. An Expert Determination can keep more date detail if the expert finds the risk very small.

Is a three-digit ZIP code allowed under HIPAA Safe Harbor?

Sometimes. The first three digits may stay only if the area formed by all ZIP codes sharing them has more than 20,000 people under current Census data. Otherwise the three digits must be changed to 000. HHS published 17 restricted three-digit codes based on 2000 Census data, but tells covered entities to use more current data where it has been published.

Are patient initials allowed in de-identified data?

Not under Safe Harbor. HHS says a data set that contains patient initials, or the last four digits of a Social Security number, does not meet the Safe Harbor requirement, because parts or derivatives of a listed identifier still count. Replace initials with a random code that is not derived from the patient's information, or remove them. An Expert Determination would have to assess them case by case.

Do doctor names have to be removed under Safe Harbor?

Not as a rule. HHS says only the names of the patients and of their relatives, employers and household members must be removed; there is no explicit requirement to remove the names of providers or staff. A covered entity must still consider whether other names could identify a patient under the actual knowledge test, and other laws or confidentiality duties may call for removing them anyway.

Is a voice recording a HIPAA identifier?

The voice can be. The Safe Harbor list includes biometric identifiers, including finger and voice prints, so a recording of a patient's voice is not de-identified just because spoken names, dates and numbers were removed. Under Safe Harbor, a clean transcript is easier to de-identify than the audio. Keeping usable audio usually means an Expert Determination that addresses the voice itself.

Is a hashed patient ID allowed under Safe Harbor?

Generally no. HHS says a code derived from a secure hash function without a secret key is an identifying element, and that under Safe Harbor a code derived from PHI generally has to be removed. What Safe Harbor allows is a re-identification code that is not derived from patient information, is kept secret and is not used for anything else. An expert can approve keyed hashes under Expert Determination.

What does actual knowledge mean in HIPAA Safe Harbor?

It means clear and direct knowledge that the remaining data could identify a patient, alone or combined with other information. HHS gives examples: a revealing occupation left in the record, a recipient with a relative in the data, a rare clinical event reported in the news, or being told the recipient can re-identify patients. Knowing that re-identification research exists is not enough by itself.

Is a phone number by itself PHI?

No. HHS says identifying information alone, such as names, home addresses or phone numbers, is not necessarily PHI; a phone book listing is not PHI because it is not related to health information. The same number becomes PHI when a covered entity or business associate holds it together with information about the person's health, the care they received or payment for that care.

Sources and further reading

What is your company’s data worth?

AI companies pay $20K to $5M for company records. Six questions, no files.