HIPAA-Compliant Research De-Identification Policy Template for Academic Pathology Departments
Policy purpose: This template sets department procedures for preparing pathology information for research. The covered entity’s Privacy Officer, research compliance office, and institutional review board (IRB) or Privacy Board should review and adapt it to institutional policy and applicable law before implementation.
Which HIPAA de-identification method should the department use?
The department shall select one of the two methods in the HIPAA Privacy Rule: Safe Harbor or Expert Determination. Under either method, information must not identify an individual, and there must be no reasonable basis to believe it can be used to identify one. 45 CFR 164.514(a)–(b) and HHS guidance on de-identification describe these methods.
- Use Safe Harbor when the department can remove every required identifier and is not aware that the remaining information could identify a person alone or in combination with other information. This may suit carefully controlled exports with limited fields and no identifying text or image content.
- Use Expert Determination when research requires information that Safe Harbor would remove or generalize, or when the data’s complexity creates a meaningful identification risk. A qualified expert must conclude that the risk of identification by the anticipated recipient is very small, using generally accepted statistical or scientific principles and methods, and document the methods and results.
For pathology datasets containing detailed timelines, rare diagnoses, narrative reports, images, or unusual combinations of clinical and demographic features, the department should consider Expert Determination when Safe Harbor would not preserve necessary research utility. The chosen method must be documented for each dataset or release; a prior determination should not automatically be applied to a materially different dataset or recipient.
How should the policy address the 18 Safe Harbor identifiers in pathology data?
Before releasing a dataset under Safe Harbor, the department shall remove the following identifiers of the patient and of the patient’s relatives, employers, or household members. The requirement applies throughout the dataset and its associated files, not only to fields labeled as identifiers.
- Names.
- Geographic subdivisions smaller than a state, subject to the rule’s limited exception for qualifying first-three-digit ZIP codes.
- All elements of dates directly related to an individual, except the year, and ages over 89 and date elements indicative of such ages.
- Telephone numbers.
- Fax numbers.
- Email addresses.
- Social Security numbers.
- Medical record numbers.
- Health plan beneficiary numbers.
- Account numbers.
- Certificate or license numbers.
- Vehicle identifiers and serial numbers, including license plate numbers.
- Device identifiers and serial numbers.
- Web Universal Resource Locators (URLs).
- Internet Protocol (IP) addresses.
- Biometric identifiers, including finger and voice prints.
- Full-face photographs and comparable images.
- Any other unique identifying number, characteristic, or code, unless the code meets the requirements for permitted re-identification under 45 CFR 164.514(c).
In pathology workflows, the department shall check case and accession identifiers, slide labels, specimen-container labels, requisitions, report headers and footers, image overlays, barcodes, file names, embedded image metadata, scanner exports, and linked tables. Treat a patient-linked accession or case number as an identifier requiring removal unless it qualifies under the regulation’s specific coded-data provisions.
The department shall also assess whether remaining information could identify a patient in the circumstances of the release. For example, an unusual diagnosis combined with a specific location or a highly distinctive clinical history may create identification concerns even after the listed fields are removed. The regulation’s 18-item list and Safe Harbor conditions are set out in 45 CFR 164.514(b)(2).
How should dates, ages, and free-text fields be handled?
For Safe Harbor, remove month and day from dates directly related to an individual; only the year may generally remain. Remove other date information more specific than the year. If a date or other field would reveal that a person is over age 89, recode the age as 90 or older and remove date elements that would reveal the person’s age more precisely.
Check dates in pathology reports, accession and collection records, procedure histories, image metadata, and longitudinal datasets. A year retained in one field may still imply an age over 89 when combined with another date or a known event. Assess the combined information, not just each field in isolation. HHS de-identification guidance discusses date elements and ages over 89.
Safe Harbor applies to recognizable identifiers in free text as well as structured fields. The department shall screen pathology reports, gross descriptions, microscopic findings, comments, clinical histories, and other narrative fields for identifiers such as names, dates, locations, and distinctive personal details. HHS states that an identifier must be removed regardless of where it appears in the record. HHS guidance on free-text fields.
Use validated text-processing tools and, where needed, human review to check free text and image content. A tool’s output alone does not establish that Safe Harbor has been met. If the department cannot reliably identify and remove recognizable identifiers, it shall not label the material Safe Harbor de-identified; it must use another permitted pathway or obtain an Expert Determination.
What documentation is required for Expert Determination?
The HIPAA requirement is that the qualified expert document the methods and results of the analysis justifying the determination. The regulation does not prescribe a specific form or require a particular credential or certificate. The department shall retain the determination and supporting documentation and make it available to the appropriate regulator upon request. See 45 CFR 164.514(b)(1).
To make the determination reviewable and reproducible, the department’s documentation should record:
- The expert’s identity and relevant knowledge and experience.
- The dataset, fields, images, and version assessed, and the intended recipient and use.
- The statistical or scientific methods, assumptions, data sources, and risk factors considered.
- Any transformations, suppression, generalization, or other controls applied to reduce risk.
- The expert’s conclusion that the risk is very small for the anticipated recipient, with the analysis supporting that conclusion.
- Any conditions or limitations on the determination, including material changes that require reassessment.
Keep the determination with the project’s release record and limit the release to the dataset and circumstances it covers. If the dataset, recipient, available external information, or release context changes materially, refer it for reassessment rather than assuming the earlier conclusion still applies.
Ready to simplify HIPAA compliance?
Join thousands of organizations that trust Accountable to manage their compliance needs.
Get Started Free TodayWhen should researchers use a limited data set, authorization, or a waiver?
When researchers need identifiable health information, the covered entity may use or disclose it for research with an individual’s valid HIPAA authorization, or under an applicable Privacy Rule permission. A limited data set is a separate option: it is still protected health information (PHI), is not de-identified, and may be used or disclosed for research with a compliant data use agreement (DUA). See HHS guidance on HIPAA and research.
- Use de-identified data when the research can be conducted without PHI and the data satisfy Safe Harbor or Expert Determination.
- Use a limited data set when research requires certain information retained in an LDS—such as dates or town, state, and ZIP code—that Safe Harbor would not permit, while direct identifiers listed in the regulation are removed. The covered entity must obtain a DUA limiting permitted uses and disclosures and requiring appropriate safeguards and restrictions on identifying or contacting individuals. See 45 CFR 164.514(e).
- Use authorization when the research use or disclosure of PHI is covered by an individual’s valid authorization and no other applicable permission is being relied upon.
- Use a waiver or alteration of authorization only when an IRB or Privacy Board has approved it under 45 CFR 164.512(i)(1)(i) and the required documentation is obtained. Among other conditions, the board must determine that the PHI use or disclosure presents no more than minimal privacy risk, that the research could not practicably be conducted without the waiver or alteration, and, if practicable, that it could not practicably be conducted without access to and use of the PHI.
For a waiver, the minimal-risk finding must be based on at least an adequate plan to protect identifiers from improper use or disclosure, an adequate plan to destroy them at the earliest opportunity consistent with the research unless retention is justified or required by law, and adequate written assurances against impermissible reuse or disclosure. Retain the required waiver documentation and follow institutional review procedures. The waiver of HIPAA authorization is distinct from any waiver of informed consent that may be required under other rules.
Researchers and the Privacy Office shall confirm which pathway applies before PHI is accessed or disclosed. A DUA for a limited data set is not a substitute for authorization or waiver when the data do not qualify as a limited data set; likewise, an agreement does not make PHI de-identified.
What safeguards should govern access to de-identified research data?
The department shall assign a data custodian to approve, document, and oversee each release. Access should be limited to authorized project personnel and the approved dataset, for the project’s stated purpose. Use institution-approved storage and transfer systems, apply role-based access controls, and maintain appropriate access and release records under institutional policy.
For datasets that remain PHI, including limited data sets, apply applicable HIPAA Privacy and Security Rule requirements and institutional controls. For a limited data set, the DUA must satisfy the applicable requirements of 45 CFR 164.514(e), including permitted-use restrictions and safeguards. The department shall ensure that researchers and their agents follow the agreement.
For information that has been properly de-identified, the HIPAA Privacy Rule no longer treats it as PHI. The department may nevertheless impose institutional confidentiality, security, access, and no-reidentification conditions as a risk-management practice or through agreements. Do not describe those additional controls as universal HIPAA requirements for all de-identified data.
Maintain a release record identifying the approved method, dataset version, recipient, purpose, applicable determination or DUA, and any limitations. Report suspected unauthorized access, disclosure, or attempted re-identification through institutional incident-reporting procedures so the Privacy Office can assess the event and determine required actions.
What are the limits of de-identification and re-identification safeguards?
De-identification reduces identification risk but does not make that risk zero. Unusual pathology findings, rare diseases, distinctive narratives, small populations, detailed timelines, and information available from outside sources may increase the chance of linkage. Expert Determination must consider the anticipated recipient and information reasonably available to that recipient; the result should not be treated as a guarantee against identification in every setting. HHS guidance explains that properly de-identified data can retain some risk.
If the covered entity assigns a code to permit re-identification, the code must not be derived from or related to information about the individual, and it must not otherwise be translatable to identify the individual. The covered entity must not use or disclose the code for another purpose or disclose the re-identification mechanism. See 45 CFR 164.514(c).
The department should keep any re-identification key separate from research data, restrict access to authorized personnel, and prohibit researchers from seeking access to the key. These are recommended safeguards for the department’s program; they do not replace the regulatory conditions for a permitted code. If a covered entity or business associate successfully identifies a person in information it maintained as de-identified, the information may again be PHI. Other applicable laws, agreements, research approvals, or institutional rules may also restrict use even where HIPAA does not.
In practice, the department should select the appropriate pathway, document the decision, inspect the full pathology record—including free text and images—and control access according to the data’s status. Safe Harbor, Expert Determination, limited data set agreements, and authorization or waiver are distinct routes; applying the right one is central to protecting patients while supporting research.
FAQs
What are the two HIPAA methods for de-identifying research data?
The two methods are Safe Harbor, which requires removal of the specified identifiers and no actual knowledge that the remaining information could identify an individual, and Expert Determination, which requires a qualified expert to determine that the risk of identification is very small and document the methods and results.
Does Safe Harbor require removing identifiers from pathology free text?
Yes. Recognizable Safe Harbor identifiers must be removed wherever they appear, including free-text pathology reports and comments. The department must also consider whether it has actual knowledge that remaining information could identify a patient.
How should dates and ages over 89 be handled under Safe Harbor?
Remove date elements more specific than the year for dates directly related to an individual. Recode ages over 89 as 90 or older, and remove date elements that would reveal a more precise age. Check combinations of dates and other fields for implied ages.
Can a limited data set be used for research without de-identifying it?
Yes. A limited data set may be used or disclosed for research without meeting the de-identification standard, but it remains PHI. The covered entity must remove the identifiers specified for limited data sets and have a compliant DUA with the recipient.
Table of Contents
- Which HIPAA de-identification method should the department use?
- How should the policy address the 18 Safe Harbor identifiers in pathology data?
- How should dates, ages, and free-text fields be handled?
- What documentation is required for Expert Determination?
- When should researchers use a limited data set, authorization, or a waiver?
- What safeguards should govern access to de-identified research data?
- What are the limits of de-identification and re-identification safeguards?
- FAQs
Ready to simplify HIPAA compliance?
Join thousands of organizations that trust Accountable to manage their compliance needs.
Get Started Free Today