On this page

HIPAA expert determination is one of two ways to de-identify health information under the Privacy Rule: a qualified expert determines that “the risk is very small that the information could be used, alone or in combination with other reasonably available information, by an anticipated recipient to identify an individual who is a subject of the information,” and documents the methods and results (45 CFR 164.514). The words “anticipated recipient” carry the weight when a telehealth or pharmacy partner shares data with a drugmaker, because HHS tells experts to consider data sources “available to a recipient” (HHS de-identification guidance). Curve Compliance describes a direct-to-patient measurement method in which the drugmaker receives aggregate results by channel, campaign and partner.

Book a call. Measuring a direct-to-patient program? In the method Curve describes, partners send step events with no name or contact details, and the drugmaker receives aggregate results by channel, campaign and partner. Curve signs a BAA on every plan and holds SOC 2 Type II. Book a call with Curve.

This page is general information, not legal advice. It summarizes 45 CFR 164.514 and the HHS guidance as they read on November 20, 2026.

What does HIPAA expert determination require?

The standard sits at 45 CFR 164.514(a): health information that does not identify a person “and with respect to which there is no reasonable basis to believe that the information can be used to identify an individual” is not individually identifiable health information (45 CFR 164.514). HHS explains that the Privacy Rule lets a covered entity or its business associate create such information by following that standard (HHS de-identification guidance). Under the Expert Determination method in 164.514(b)(1), “A person with appropriate knowledge of and experience with generally accepted statistical and scientific principles and methods for rendering information not individually identifiable” applies those methods, finds the risk very small for the anticipated recipient, and “Documents the methods and results of the analysis that justify such determination.”

The other route is the Safe Harbor method in 164.514(b)(2): remove 18 listed identifiers of the person and of relatives, employers or household members, and have no actual knowledge that what remains could identify someone. Neither route reaches zero. HHS notes that both methods “yield de-identified data that retains some risk of identification,” and that “Although the risk is very small, it is not zero” (HHS de-identification guidance). For how de-identified data differs from anonymized and aggregated data, see de-identified data under HIPAA.

QuestionExpert determinationSafe Harbor
Where in the rule45 CFR 164.514(b)(1)45 CFR 164.514(b)(2)
What it takesA qualified expert's documented analysisRemoving 18 listed identifiers, with no actual knowledge that the rest could identify someone
Who judges the riskThe expert, for a specific anticipated recipientThe rule itself, through the fixed list
Dates and placesFiner detail can stay if the expert finds the risk very smallAll date elements except the year come out, and so do places smaller than a state, with a narrow three-digit ZIP code exception
Fit for marketing dataFlexible, but tied to who receives the data and what else they holdSimple, but removes detail such as exact dates that campaign reporting may want
After the first releaseRepeat releases to the same recipient may need a fresh lookActual knowledge that data could identify someone still counts

Who counts as an expert, and how small is very small?

HHS says “There is no specific professional degree or certification program for designating who is an expert at rendering health information de-identified.” When it enforces the rule, OCR reviews the expert's professional experience, academic or other training, and actual experience using de-identification methods (HHS de-identification guidance).

There is no fixed threshold either. HHS says “There is no explicit numerical level of identification risk that is deemed to universally meet” the very small standard, and that an expert sets acceptable risk “based on the ability of an anticipated recipient to identify an individual.” A determination also ages. The rule sets no expiration date, but HHS notes that “technology, social conditions, and the availability of information changes over time,” and that for future releases to the same recipient, such as monthly reporting, an expert should examine whether they “should be subject to additional or different de-identification processes consistent with current conditions” (HHS de-identification guidance). A partner report that goes to a drugmaker every month is that kind of repeat release.

Why does the data a recipient already holds change the answer?

Re-identification usually works by linkage. HHS describes three conditions: the de-identified data must be distinguishing, there must be “a naming data source,” and there must be a mechanism to relate the two (HHS de-identification guidance). Its principles for experts weigh replicability, data source availability and distinguishability, and in HHS's words “experts are advised to consider how data sources that are available to a recipient of health information (e.g., computer systems that contain information about patients) could be utilized for identification of an individual.”

HIPAA's de-identification rules are written for covered entities and their business associates, which in a direct-to-patient program can include a telehealth practice or a pharmacy partner. When such a partner prepares data for a drugmaker, the drugmaker is the recipient the expert has to picture. If that drugmaker also runs a patient support program or a pharmacy service of its own, its systems may hold records about some of the same people. A feed that looks harmless on its own, such as one row per paid fill with a date, a three-digit ZIP code, a product and a partner, could line up against those records. HHS makes the general point that a risk found acceptable “for one particular data set in the context of a specific environment may not be appropriate for the same data set in a different environment” (HHS de-identification guidance). The same report can be very low risk for one recipient and higher for another.

What practices reduce re-identification risk in shared marketing data?

HHS describes three broad classes of mitigation: suppression (removing a feature, a record or a single value), generalization (for example, a five-digit ZIP code shortened to three digits, or exact ages grouped into ranges) and perturbation (replacing values with different but equally specific ones). It notes that agencies set policies on small cell counts, but “OCR does not designate a universal value for k,” which “should be set at a level that is appropriate to mitigate risk of identification by the anticipated recipient of the data set” (HHS de-identification guidance). An expert may also limit distribution through an agreement in which the recipient “agrees not to attempt identification of the subjects.” For how small-group rules work in practice, see small-group rules for de-identified answers.

Turned into questions a drugmaker can ask before any partner data arrives:

  • Aggregates first. Can results come by channel, campaign and partner rather than one row per patient? HHS gives the example that an average age reported across a health plan is not PHI, because it identifies no member.
  • Minimum cell sizes. What is the smallest number of people behind any reported figure, and what happens to smaller cells? HHS leaves the number to the expert and the recipient.
  • Coarse time windows. Are results reported by week or by month rather than by the date of each fill, so a row can't be matched to a dispensing date in the recipient's own records?
  • Fields that match recipient records. Are ZIP codes, exact dates, order numbers and other fields the recipient's systems also hold left out?
  • A contract ban on re-identification. Does the data use agreement bar attempts to identify anyone, and limit who at the recipient can see the reports?
  • A review schedule. When does the expert revisit the analysis as the data, the recipient and the available data sources change?

None of these replaces the expert's analysis. HHS says a data use agreement “does not substitute for any of the specific requirements of the Expert Determination Method” (HHS de-identification guidance). Whether a data flow needs an expert determination, a data use agreement or both is a call for the partner's privacy officer and the drugmaker's legal team, so talk to your counsel before any partner data reaches your team.

How Curve Compliance helps

Curve's direct-to-patient measurement method is built around aggregates. In this method, each partner sends a stripped-down event at each step (visit, consult, prescription, paid fill) carrying the campaign that brought the patient, with no name or contact details. The drugmaker receives aggregate results by channel, campaign and partner, including cost per paid fill.

On the ad platform side, Curve strips PHI before data reaches an ad platform and flags PHI-like patterns so they can be stopped at the source. Each platform receives only a fixed list of fields, and contact identifiers are off by default and SHA-256 hashed when enabled. Curve signs a BAA on every plan, holds SOC 2 Type II (security), and its team does the setup, typically live in about a week, case by case. Book a call to walk through what each partner would send and what the drugmaker would receive.

For a measurement setup that reports in aggregate, see direct-to-patient pharma attribution.

For the ad platform side of the same program, see direct-to-patient ad platform data.

Sources, checked November 20, 2026: 45 CFR 164.514; HHS OCR de-identification guidance (content last reviewed February 3, 2025).

Frequently Asked Questions

What is HIPAA expert determination?

It is one of the two HIPAA methods for de-identifying health information: a qualified expert applies statistical and scientific methods, finds the risk very small that an anticipated recipient could identify anyone, and documents the analysis. The other method is Safe Harbor, which removes 18 listed identifiers.

What is the difference between expert determination and Safe Harbor?

Safe Harbor removes a fixed list of 18 identifiers and requires no actual knowledge that the rest could identify someone. Expert determination lets more detail stay when a documented analysis shows the risk is very small for the anticipated recipient (45 CFR 164.514).

Does HIPAA set a number for very small risk?

No. HHS says there is no explicit numerical level that universally meets the standard and that OCR does not designate a universal value for k, so the expert sets the level for the anticipated recipient (HHS de-identification guidance).

Who can perform an expert determination?

HHS says there is no specific professional degree or certification program for the role. OCR looks at the expert's professional experience, training and actual experience with de-identification methods.

Does an expert determination expire?

The Privacy Rule sets no expiration date, but HHS notes that some practitioners put time limits on their determinations, and that repeat releases to the same recipient, such as monthly reports, may need additional or different de-identification as conditions change.

Is a data use agreement required for de-identified data?

No. HHS says the Privacy Rule doesn't require one, but a covered entity may ask the recipient to sign one with terms such as a ban on re-identification, and the agreement doesn't replace the expert's analysis.

Why does it matter what data the recipient already holds?

Re-identification works by linking a data set to a source that names people. HHS advises experts to consider the data sources available to the recipient, so a drugmaker with its own patient records is a different recipient from one without them.

Talk to Curve Compliance

Book a call. Building measurement for a direct-to-patient program? In the method Curve describes, the drugmaker receives aggregate results by channel, campaign and partner. Curve signs a BAA on every plan, holds SOC 2 Type II, and its team does the setup, typically live in about a week, case by case. Book a call with Curve.