Developing High Performance Secure Multi-Party Computation Protocols in Healthcare: A Case Study of Patient Risk...

Area of Science:

• Computational cryptography
• Health informatics
• Data security

Background:

• Clinical use cases often require sensitive data exchange across multiple institutions.
• Existing methods for data linkage and analysis may compromise patient privacy or require trusted intermediaries.
• Secure multi-party computation (MPC) offers a potential solution for privacy-preserving collaborative analytics.

Purpose of the Study:

• To demonstrate the viability of MPC using garbled circuits for clinical data analysis.
• To develop and test MPC protocols for patient risk stratification metrics.
• To evaluate the security and efficiency of these protocols in a cross-institutional setting.

Main Methods:

• Implementation of two MPC protocols based on Yao's garbled circuits.
• Utilizing private set intersection (PSI) for record linkage.
• Employing cuckoo hashing for optimized performance.
• Testing with large, realistic synthesized datasets.

Main Results:

• Successful computation of high utilizer identification (PSI-HU) and comorbidity index calculation (PSI-CI) metrics.
• Protocols achieved run times of minutes, significantly faster than traditional methods.
• Demonstrated provable security against computationally bounded adversaries in a semi-honest setting.

Conclusions:

• MPC using garbled circuits is a viable and efficient technology for secure cross-institution clinical data exchange and analysis.
• The developed protocols enable fast and secure computation of critical patient risk stratification metrics.
• These MPC protocols eliminate the need for a trusted third party, enhancing data security and collaboration.