R-U policy frontiers for health data de-identification

Area of Science:

• Health Informatics
• Data Privacy
• Computer Science

Background:

• The Health Insurance Portability and Accountability Act (HIPAA) Privacy Rule permits sharing de-identified health data.
• Current methods include formal risk assessment (e.g., k-anonymity) or rule-based policies (Safe Harbor).
• Safe Harbor is interpretable but does not tailor privacy to recipient capabilities, potentially sacrificing data utility.

Purpose of the Study:

• To develop a method for generating a frontier of rule-based de-identification policies that systematically balance re-identification risk and data utility.
• To compare these rule-based frontiers with existing k-anonymity solutions and the Safe Harbor policy.

Main Methods:

• An algorithm was extended to efficiently compose a risk-utility (R-U) frontier using a lattice of policy options.
• Risk was defined proportionally to patient count, and utility proportionally to distribution similarity.
• The method searched 20,000 rule-based policies, comparing results with k-anonymity and Safe Harbor using U.S. state demographic data.

Main Results:

• The generated rule-based frontiers comprised approximately 5000 policies on average.
• A significant portion (2%) of these policies offered superior utility with lower risk compared to Safe Harbor.
• The rule-based frontiers demonstrated a broader range of utility and risk trade-offs than k-anonymity frontiers.

Conclusions:

• Efficient discovery of R-U frontiers for de-identification policies is feasible.
• This approach enables healthcare organizations to customize data protection based on recipient needs and trustworthiness.
• Tailored de-identification policies can enhance both data utility and privacy.