Private Continuous Survival Analysis with Distributed Multi-Site Data

Area of Science:

• Epidemiology
• Data Science
• Health Informatics

Background:

• Effective disease surveillance relies on large-scale epidemiological data for improved health outcomes.
• Multi-site data sharing is crucial but faces challenges in privacy protection and decentralization.
• Existing privacy solutions often depend on a central site, posing risks and failing to support dynamic data analysis.

Purpose of the Study:

• To propose a novel privacy-protecting approach for decentralized, dynamic epidemiological analysis.
• To address the limitations of centralized privacy solutions and static data assumptions.
• To enable timely clinical interventions and policy decisions through secure data sharing.

Main Methods:

• Developed a decentralized privacy-preserving framework for distributed epidemiological data.
• Applied the solution to continuous survival analysis using the Kaplan-Meier estimation model.
• Integrated differential privacy to ensure robust data protection.

Main Results:

• The proposed method supports dynamic epidemiological analysis in a decentralized manner.
• The approach provides strong privacy guarantees without relying on a central site.
• Evaluations on a COVID-19 dataset demonstrated the high usability of the results.

Conclusions:

• This work presents a viable solution for privacy-preserving, dynamic, multi-site epidemiological analysis.
• The decentralized approach enhances security and overcomes the limitations of traditional methods.
• The findings are crucial for advancing real-time disease surveillance and public health decision-making.