Privacy Risks of Sharing Data from Environmental Health Studies

Area of Science:

• Environmental health
• Data privacy
• Bioinformatics

Background:

• Research data sharing offers benefits like resource efficiency and reproducibility.
• Sharing environmental health data, unlike genetic or medical data, is understudied regarding privacy risks.
• Participant re-identification is a key concern when linking study data with external datasets.

Purpose of the Study:

• To evaluate the vulnerability of environmental health (EH) data to linkage with external datasets.
• To investigate if environmental measurements can infer participant characteristics, such as geographic location, thereby increasing privacy risks.
• To assess the privacy implications of sharing EH research data.

Main Methods:

• Identified 12 prominent EH studies and reviewed their data types and overlap with external data.
• Utilized k-means clustering and principal component analysis on environmental measurements from two case studies.
• Assessed the accuracy of inferring participants' region of residence from household air and dust chemical measurements.

Main Results:

• All 12 studies contained data types (e.g., geographic location, medical data, occupation) with potential for linkage.
• Participant region of residence was inferred with 80%-98% accuracy using environmental measurements.
• Environmental health studies frequently collect data vulnerable to linkage with publicly available records.

Conclusions:

• Unsupervised sharing of EH research data poses significant privacy risks due to potential re-identification.
• Exposure measurements can be used to infer subgroup membership, increasing the likelihood of data linkage.
• Legal and policy protections are essential to mitigate the harms of re-identification from shared EH data.