Published August 13, 2026 | Version v1

When health records become representations of people

Authors/Creators

Description

Health records are increasingly linked across services and over time, and used as training material for models that generate claims about the people they describe. This position paper argues that composition — linkage combined with unstructured or derived content — changes what a health record is, and that the governance frames it examines do not yet treat composition itself as the event that creates a distinct governable representational object. It proposes a three-layer account of what a composed record contains, and identifies where governance obligations can attach to each layer.

The paper extends the classification proposed in The Trusted Hub (https://doi.org/10.5281/zenodo.20613962), responding to feedback on that preprint and to subsequent developments in Canadian standards. Mental health data is used as the design case, on the grounds that a framework which holds for records of that kind is more likely to hold for less demanding ones.

Files

When health records become representations of people.pdf

Files (192.0 kB)

Name Size
md5:40aafc3e3c43da915fc1bca6dcd9df74
192.0 kB Preview Download

Additional details

Additional titles

Subtitle
Composition, inference, and why de-identification no longer settles the question

Related works

Is new version of
Preprint: 10.5281/zenodo.20613962 (DOI)

Dates

Copyrighted
2026-08-13