When health datasets collected years or even decades ago are reused for modern data science research, a critical question arises: do those stored data still accurately and legitimately represent the people and communities they came from? Existing ethics frameworks largely focus on privacy and consent at the point of collection, but say little about whether data remain fit for use as time passes and contexts change.

This paper addresses that gap by introducing the concept of representational veracity — the fidelity of stored data to the evolving identities, characteristics, and normative expectations of the people who originally contributed them. Drawing on philosophical accounts of how things persist through time, the authors identify four domains where the adequacy of stored data can erode: material provenance, informational descriptors, normative authorization, and relational community. The paper translates this analysis into a practical governance framework — including an assessment matrix, triggers for heightened ethical review, and documentation standards — illustrated through genomics and data science case studies, with particular attention to equity implications for historically marginalized populations.