Implementing a hash-based privacy-preserving record linkage tool in the OneFlorida clinical research network

Area of Science:

• Health Informatics
• Data Science
• Biomedical Research

Background:

• Accurate patient record linkage is crucial for large-scale health research.
• Existing methods often face challenges with privacy and scalability.
• Deterministic record linkage (RL) offers a precise approach but requires careful implementation.

Purpose of the Study:

• To implement and validate an open-source, privacy-preserving tool for deterministic record linkage (RL) in a real-world clinical research network.
• To assess the performance of the tool in linking and deduplicating large volumes of electronic health records and claims data.
• To demonstrate the utility of enhanced data quality for scientific discovery.

Main Methods:

• Developed OneFL Deduper, a Python-based open-source tool utilizing cryptographic one-way hash functions for privacy-preserving record linkage.
• Learned and validated two deterministic linkage rules using voter registration data and gold-standard datasets.
• Deployed the tool within the OneFlorida Clinical Research Network, linking data from multiple healthcare partners and a state Medicaid program.

Main Results:

• Achieved high precision (97.25–99.7%) and reasonable recall (75.5%) in record linkage using gold-standard datasets.
• Successfully deduplicated approximately 3.5 million records to 1.7 million unique patients across six healthcare partners and Florida Medicaid.
• Demonstrated the benefits of record linkage by examining distinct disease profiles in the linked cohorts.

Conclusions:

• Deterministic record linkage (RL) is essential for improving data quality in research networks, enabling reliable scientific discoveries.
• The OneFL Deduper tool provides a practical, privacy-preserving solution for real-world record linkage challenges.
• Factors such as privacy regulations, data availability, and computational resources influence the construction of RL solutions.