A computational model to protect patient data from location-based re-identification.

Area of Science:

• Health Informatics
• Bioinformatics
• Data Privacy

Background:

• Healthcare organizations must protect patient anonymity when disclosing sensitive data like DNA.
• Traditional de-identification methods are vulnerable to re-identification using public information.
• Patient location visit patterns ('trails') pose a threat to DNA data anonymity when disclosed by multiple organizations.

Purpose of the Study:

• To develop computational methods for disclosing patient-specific DNA records that are resistant to trail re-identification.
• To provide health care organizations with solutions to prevent patient anonymity threats in shared databases.
• To enable the disclosure of DNA records while ensuring provable protection against re-identification.

Main Methods:

• Introduced a formal model called k-unlinkability to define degrees of patient anonymity.
• Developed algorithms for coordinating data disclosure among health care organizations to maintain k-unlinkability.
• Evaluated algorithm efficacy using patient population data from hospital discharge databases and real-world metrics.

Main Results:

• It is not necessary to suppress all records violating k-unlinkability; only portions of trails require suppression.
• Applying protection algorithms significantly increases the percentage of disclosable DNA records while maintaining k-unlinkability (e.g., 95% disclosure achieved).
• Experimental findings demonstrate the effectiveness of the proposed methods across various patient populations.

Conclusions:

• Patient anonymity can be formally protected in shared healthcare databases.
• Significant quantities of patient-specific DNA data can be disclosed with guaranteed protection against trail re-identification.
• The developed methods offer configurable privacy protection levels, enabling informed policy formulation.