SHARE: system design and case studies for statistical health information release

Area of Science:

• Biomedical Informatics
• Data Privacy
• Statistical Computing

Background:

• Protecting patient privacy in electronic health records (EHRs) is crucial for data sharing and research.
• Traditional anonymization methods often fall short, necessitating robust privacy-preserving techniques.
• Differential privacy offers a strong theoretical framework for statistical data release.

Purpose of the Study:

• To introduce SHARE, a novel system for releasing statistical health information from EHRs with differential privacy.
• To evaluate the feasibility and utility of SHARE on real-world biomedical datasets.
• To demonstrate the application of differential privacy in the medical domain for aggregate data release.

Main Methods:

• SHARE implements state-of-the-art methods for releasing multidimensional histograms and longitudinal patterns.
• The system was tested on two distinct real-world datasets: SEER breast cancer and Emory EHR.
• Differential privacy guarantees were applied to ensure data utility while protecting individual privacy.

Main Results:

• SHARE effectively handles heterogeneous medical data, producing useful aggregate statistics.
• Low Kullback-Leibler divergence (below 0.5 and 0.01 for 7D and 3D data cubes from SEER) indicates high fidelity.
• Relative error for longitudinal pattern queries on the EeMR dataset ranged from 0 to 0.3, demonstrating accuracy.

Conclusions:

• SHARE represents a pioneering system for releasing differentially private aggregate statistics in the medical domain.
• The proof-of-concept system is designed for integration into large-scale medical data warehouses.
• While promising, challenges persist in applying differential privacy for higher-dimensional medical data analysis.