Assessing privacy risks in population health publications using a checklist-based approach.

Area of Science:

• Population health research
• Data privacy and security
• Statistical analysis

Background:

• Increasing access to detailed datasets in population health research raises privacy concerns.
• A proposed checklist aims to help researchers identify and mitigate privacy risks in analysis outputs.
• This study evaluates the usability and reliability of this checklist approach.

Purpose of the Study:

• To investigate the usability and reliability of a researcher-applied checklist for identifying privacy risks in population health analysis outputs.
• To assess whether different users identify the same privacy risks using the checklist on sample publications.

Main Methods:

• A checklist was applied to 100 academic population health publications.
• Five readers evaluated the publications.
• Interrater agreement was measured using Cohen's kappa.

Main Results:

• Counts, summary statistics, plots, and model outputs were the most common statistical outputs.
• The checklist identified 128 outputs (22.6%) with potential privacy concerns, mainly related to small counts.
• No substantial actual privacy concerns were found upon contextual review, and interrater agreement was generally good.

Conclusions:

• The checklist is a reliable tool for assisting population health researchers in anonymizing analysis outputs.
• This approach shows potential for development into a broadly applicable standard for consistent data confidentiality protection.