A software package for the application of probabilistic anonymisation to sensitive individual-level data: A proof of principle with an example from the ALSPAC birth cohort study

Abstract

Individual-level data require protection from unauthorised access to safeguard confidentiality and security of sensitive information. Risks of disclosure are evaluated through privacy risk assessments and are controlled or minimised before data sharing and integration. The evolution from ‘Micro Data Laboratory’ traditions (i.e. access in controlled physical locations) to ‘Open Data’ (i.e. sharing individual-level data) drives the development of efficient anonymisation methods and protection controls. Effective anonymisation techniques should increase the uncertainty surrounding re-identification while retaining data utility, allowing informative data analysis. ‘Probabilistic anonymisation’ is one such technique, which alters the data by addition of random noise. In this paper, we describe the implementation of one probabilistic anonymisation technique into an operational software written in R and we demonstrate its applicability through application to analysis of asthma-related data from the ALSPAC cohort study. The software is designed to be used by data managers and users without the requirement of advanced statistical knowledge.