As health data are increasingly reused and transformed for new research and AI applications, ethical governance must consider not only where data came from, but whether the ethical relationships and obligations associated with those data remain meaningful as their uses change.

This paper identifies and names the Continuity Trap—a governance error in which evidence of continuity in one domain, such as provenance or documentation, is treated as sufficient evidence that ethical continuity has been preserved overall. The paper develops four domains through which ethical continuity can be assessed: provenance, semantics, authorization, and relational standing.

Rather than proposing that every secondary use of health data should undergo complete re-review, the paper proposes a targeted, trigger-based approach to continuity review, directing attention to the domains most likely to have changed as data are linked, transformed, modeled, or redeployed. This has particular relevance to contemporary health AI, where datasets and models may move across institutions, purposes, populations, and stages of the data lifecycle.

For BridgELSI, the paper contributes to a broader governance approach that seeks to ensure that ethical oversight remains meaningful as data science and AI-enabled health research evolve.