Impact of Anonymity Indicator "PFDOC Entropy l-Diversity" on Medical Data Utility

In clinical research articles, statistics about the patients under the study are presented in the data format called "patient characteristics". Patient characteristics provide insights for interpreting the study results. However, disclosing patient characteristics with low l of "Patient Family Detect on Overall Category (PFDOC) entropy l-diversity", an indicator of patient characteristics anonymity, raises privacy concerns.In this paper, we examine a scenario in which medical institutions withhold publication of their articles that include low l of PFDOC entropy l-diversity patient characteristics based on the existing patient characteristics publication model. In this scenario, the number of published medical articles would decrease, and this reduction may impact data utility. Therefore, we check two quantitative utility indicators. The first utility indicator is the number of published papers without being withheld. The second utility indicator is the degree of difference between the average of the published population data contained in patient characteristics and the true population ratio, such as the prevalence of diabetes in a particular patient group. According to the numerical results, when articles are excluded by 1 ≦ l ≦ 1.5, the data loss is more than 5% and the average difference is more than 0.5% if a category with a population ratio of 15% or less or 85% or more is included in the patient characteristics. This study provides guidance on the adoptable l of PFDOC entropy l-diversity with consideration of utility.