Metrics and methods for the de-identification of health information

点击翻译按钮获取中文摘要

With the increasing investments in electronic health records, and the rise in their deployment, more personal health information (PHI) is being collected and stored electronically. There is considerable demand for that data for secondary purposes, such as research, public health, quality improvement, and improving patient safety. Even when disclosures of PHI are permitted without patient consent, providers are reluctant to do so. In general, providers, privacy officers, and research ethics boards will require that health data be de-identified when disclosed for secondary purposes. In our previous work we have developed re-identification risk metrics and methods for the optimal de-identification of cross-sectional data for managing identity disclosure. The objective of the proposed program of work is to extend and improve on these metrics and methods to develop: metrics with more realistic re-identification attack assumptions, more efficient de-identification of geospatial information based on the clustering of small areas, and the de-identification of longitudinal data sets by building on previous work on the de-identification of transactions.