Applied Session: “Real-World” De-Identification of Transactional Data Extracted from Electronic Health Records - Breaking the Curse of Dimensionality

No ratings

Presented at Privacy&SecurityConference 2019 by

Dr. Moselle

Researchers who seek to detect better practices or predict better outcomes from "real world" data extracted from clinical information systems must be supplied with very high dimensional datasets that are deeply refractory to privacy protection via application of 'classic' data de-identification tools (e.g., "k-anonymization"). Nevertheless, the privacy risk model out of which these tools are fashioned is a foundational component of methods that can provide meaningful operational definitions of key constructs such as "identifiable", or "risk", or "low risk" - or "de-identified"! This presentation will cover a "meta-k-anonymization framework" that is intended to scale out to the privacy challenges associated with "real world" (almost invariably high dimensional) health datasets whose analytical content must be protected in order to derive products that warrant application back to the points of service.