A coded row can remain distinctive
A rare diagnosis, precise date, location, and unusual care sequence may identify someone when combined with other information. A separate code key is another possible route back to identity.
The risk depends on the data, access conditions, and available outside information, not just the absence of a name column.
The legal standard needs a jurisdiction
For HIPAA in the United States, HHS describes two de-identification methods: Safe Harbor and Expert Determination. Those have specific requirements. A general claim that data are “anonymous” is not a substitute for identifying which standard and process were used.
Other jurisdictions and agreements may use terms differently.
Questions for a dataset handoff
- Which identifiers were removed or transformed?
- Does anyone retain a re-linking key?
- Were dates, locations, and rare combinations considered?
- What uses and access controls are permitted?
These questions are relevant even when the dataset is intended for a beneficial project.
De-identification also changes analytic possibilities. Coarser dates may protect privacy while limiting time-to-event analysis. Documenting those transformations helps users avoid treating altered fields as exact original observations. Privacy protection and scientific interpretation both depend on knowing what changed.
