A Multi-layer model for privacy preserving policy making for disclosure of public health data

Alizadeh Mizani, Mehrdad
Health organizations in Turkey collect ever-increasing amount of individual data are valuable source of information for public health research. However, due to privacy risks, they publish data in aggregated rather than individual forms. The lack of standardized policies regarding secondary uses of health data leads to ineffectiveness of available technical methods. As a result, access to and utilization of person-specific datasets by public health researchers become extremely cumbersome. The bias introduced by privacy protection methods also makes data inefficient for epidemiological and public health contexts. We developed a three layer model for evidence-based policy making for secondary uses of health data. The first layer covers the evaluation of anonymized datasets based on clustering analysis independent of the underlying algorithm. The second layer provides the researcher with Representability Vector (RV), which consists of information about factors affecting the interpretation of research results. RV is also a method to gather researcher requirements and context oriented evidence. The third layer, provides a generic framework for policy making with pseudo-contents covering the issues of anonymization and RV. This framework provides a dynamic approach for disclosure of and reporting bias to the researcher while emphasizing the policy issues along with context, evidence, and regulations.