|
Abstract:
|
Protecting confidentiality is an essential duty of all Health Departments. However, existing guidelines have important limitations, including: 1) poor specification of disclosure, 2) failure to consider multiple data releases, 3) arbitrary rules-of-thumb and 4) conflation of privacy risk and reliability. We address these issues in an increasingly common setting: a queriable web-based reporting system for health events and vital statistics. We compare traditional approaches to more rigorous statistical methods. Two microdata methods (K-anonymity and L-diversity) are applicable but must be modified to address multiple tabular releases, for which we use a guidance matrix (Dandekar). The resulting method is easily applied and in our example verifies that the proposed releases contain no disclosure risk (hospitalizations, births), or only require minor modifications to ensure safety (deaths).
|
