Protecting patient privacy in survival analyses

Area of Science:

• Biomedical Informatics
• Computer Science

Background:

• Survival analysis is crucial for clinical decisions, calculating patient survival probabilities.
• Sharing exact survival curves risks revealing individual patient data and study participation.
• Protecting patient privacy in survival analysis is imperative.

Purpose of the Study:

• To develop a privacy-preserving framework for survival analysis.
• To ensure provable privacy protection against adversaries using differential privacy.
• To evaluate the performance of privacy-protecting solutions for the Kaplan-Meier model.

Main Methods:

• Developed a framework based on differential privacy.
• Applied the framework to the Kaplan-Meier nonparametric survival model.
• Empirically evaluated the framework using epidemiology and synthetic datasets.

Main Results:

• The proposed methods significantly reduce privacy risk compared to non-private approaches.
• The utility of survival curves is retained while enhancing privacy.
• Demonstrated the feasibility of privacy-protecting survival analyses.

Conclusions:

• The developed framework offers strong privacy protections for survival analyses.
• The methods preserve the usefulness of survival data for biomedical research.
• Future work will focus on enhancing the utility of privacy-preserving solutions.