Disassociation for electronic health record privacy

Area of Science:

• Health Informatics
• Medical Data Privacy
• Biostatistics

Background:

• Electronic Health Record (EHR) data sharing is crucial for large-scale medical research and public health improvement.
• Current methods for anonymizing EHR data risk patient re-identification through diagnosis codes.
• Existing privacy-preserving techniques often lead to significant data utility loss.

Purpose of the Study:

• To propose a novel approach, disassociation, for secure EHR data dissemination.
• To prevent patient re-identification by transforming records into carefully selected subsets.
• To minimize data utility loss compared to existing privacy-preserving methods.

Main Methods:

• The study introduces 'disassociation,' a technique that splits EHR records into subsets.
• This method transforms data to prevent linkage based on sensitive information like diagnosis codes.
• Extensive experiments were conducted using real-world EHR data.

Main Results:

• The disassociation approach effectively prevents patient re-identification risks.
• It achieves significantly lower data utility loss than current methods.
• The transformed data remain highly useful for clinical case count studies and general medical analysis.

Conclusions:

• Disassociation offers a robust solution for privacy-preserving EHR data sharing.
• This method enhances the utility of shared EHR data for diverse research applications.
• It supports broad data dissemination for advancing public health initiatives.