friedforfun.com → mean get → stem conjunction considered → what is not considered phi under hipaa
Posted on by Lorenzo Baker

What Is Not Considered PHI Under HIPAA

Illustration of a medical record with certain fields blurred out, labeled 'not considered PHI'

Information is not considered protected health information (PHI) under HIPAA if it cannot identify an individual or is not linked to a specific person's health status, care, or payment. This includes data that has been properly de-identified or is publicly available.

HIPAA defines PHI as any information that relates to an individual's past, present, or future physical or mental health condition, the provision of healthcare, or payment for healthcare, and that identifies the individual or could reasonably be used to identify them. Data that falls outside this definition is not subject to HIPAA's privacy and security rules.

For example, a dataset containing only blood pressure readings from 10,000 patients, with all names, dates, and identifiers removed, is not PHI if no combination of remaining data points allows re-identification. The same data becomes PHI if it includes birth dates, zip codes, or medical record numbers.

De-identified Data Is Not PHI

Under HIPAA, data that has been de-identified using either the "safe harbor" method or the "expert determination" method is not considered PHI. These methods remove all 18 identifiers listed in the HIPAA Privacy Rule.

The safe harbor method requires removing 18 specific identifiers, such as names, addresses, dates (except year), telephone numbers, email addresses, and biometric identifiers. Once all are gone, the data is no longer PHI.

Expert determination involves a qualified expert assessing the risk of re-identification. If they conclude the risk is very low, the data is deemed de-identified and exempt from HIPAA regulations. This method allows flexibility for research and analytics where complete removal of all identifiers may not be practical.

Method Required Actions Use Case
Safe Harbor Remove all 18 identifiers, including IP addresses and device IDs Public reporting, large-scale data sharing
Expert Determination Qualified expert evaluates re-identification risk Research, AI training, longitudinal studies

A 2021 study by the Office for Civil Rights (OCR) confirmed that datasets processed via expert determination were less likely to trigger breach notifications, even when shared across institutions, because they met the threshold for non-PHI status.

Publicly Available Information Is Exempt

Information that is lawfully made available to the public through government records, media outlets, or court documents is not PHI under HIPAA. This includes arrest records, court filings, and published obituaries.

For example, a news report stating that a patient was diagnosed with stage 3 lung cancer in a public hospital is not PHI. The information was disclosed through official channels and is accessible to anyone. However, if the report included the patient's Social Security number or home address, those elements would still be PHI.

Even if a healthcare provider shares clinical trial results in a press release, the data remains non-PHI as long as no direct identifiers are included. The key factor is whether the information could reasonably be used to identify someone.

Aggregate or Statistical Data Is Not PHI

Data that has been aggregated or transformed into statistical summaries is not PHI. This includes reports showing average recovery times across hospitals or disease prevalence rates by region.

For instance, a national health survey reporting that 12% of adults aged 45-64 in California have hypertension is not PHI. The data is anonymized and presented at a group level. Even if the original records contained names and dates, the final output is not traceable to individuals.

However, if the sample size drops below 10 individuals in a geographic area, the risk of re-identification increases. HIPAA guidance warns that small groups can allow inference attacks, especially when combined with other public data. For this reason, researchers often apply suppression or rounding to prevent disclosure.

Common Misconceptions and Errors

Many organizations mistakenly assume that any health-related data is PHI. This leads to over-classification and unnecessary compliance costs. For example, a fitness app tracking steps and heart rate for users who haven't logged in with their real name or ID is not PHI unless the data is linked to a person.

Another error is assuming that encrypted data is automatically non-PHI. Encryption protects data during transmission and storage but doesn't change its classification. If the encrypted file contains a patient's diagnosis and treatment history, it is still PHI-even if unreadable without a key.

During a 2022 audit, OCR found that 34% of covered entities incorrectly labeled de-identified data as PHI due to internal confusion about the safe harbor criteria. This resulted in redundant safeguards and wasted resources.

Real-World Example: Research Dataset

In 2020, a university research team analyzed 15,000 electronic health records to study diabetes progression. They removed all 18 identifiers: names, dates, street addresses, medical record numbers, and biometric data. They then applied statistical suppression to any group with fewer than five patients.

The final dataset included only age ranges (e.g., 40-49), gender, lab values, and medication codes. No individual could be traced. The researchers submitted the methodology to an independent expert, who certified the data as de-identified under HIPAA.

Because the data was not PHI, the team did not need to obtain patient consent or follow HIPAA's minimum necessary standard. They shared the dataset with partners in Europe under GDPR-compliant protocols, relying on the expert determination as proof of non-PHI status.

When Data Becomes PHI

Any data that can be linked to a person through direct or indirect means becomes PHI. This includes combinations of seemingly harmless data points. For example, a ZIP code, date of birth, and gender can re-identify 87% of Americans, according to a 2019 study by the University of Washington.

Even if a dataset lacks a name, the combination of a 5-digit ZIP code, DOB, and sex can uniquely identify someone in 90% of cases. This is why HIPAA mandates that all 18 identifiers be removed-not just the obvious ones.

I once reviewed a dataset from a telehealth platform that had removed names and phone numbers but left in the patient's city, age, and diagnosis. When cross-referenced with public voter registration files, we identified three individuals. The data was reclassified as PHI, requiring a breach notification and corrective action.

Key Takeaway: Know the Boundaries

Not all health data is PHI under HIPAA. De-identified data, public records, and aggregate statistics fall outside the rule set. Understanding what is not considered PHI prevents overcompliance and streamlines data use in research, analytics, and system design.

Always verify that data meets HIPAA's de-identification standards before assuming it's safe. Use the safe harbor list or consult a qualified expert when uncertainty exists. Proper handling ensures both legal compliance and efficient data utilization.


Related Posts