Linking temporal medical records using non-protected health information data.

Area of Science:

• Medical Informatics
• Data Science
• Health Research

Background:

• Multi-institutional collaborations are vital for medical research, enabling advanced analytics but posing data integration challenges.
• Current data integration methods often rely on protected health information (PHI), which is difficult to share and may be inaccurate.
• Existing solutions typically do not account for the temporal nature of electronic health records (EHR).

Purpose of the Study:

• To develop a novel approach for linking longitudinal patient data using non-PHI.
• To address the limitations of current data integration techniques in collaborative research settings.
• To improve the utility and privacy of patient data in multi-institutional studies.

Main Methods:

• Proposed a technique that links patient longitudinal data using non-PHI.
• Leveraged diagnosis dependencies and patterns to identify and link patient records.
• Evaluated the approach as a standalone method and in combination with Privacy Preserving Record Linkage (PPRL).

Main Results:

• The novel approach effectively links patient longitudinal data using non-PHI.
• Diagnosis patterns proved to be significant indicators for temporal record linkage.
• The method demonstrated effectiveness in experimental evaluations on real-world datasets, improving data integration.

Conclusions:

• The proposed technique offers a robust solution for temporal record linkage using non-PHI.
• This method enhances data integration for multi-institutional research, particularly with EHR data.
• The approach can be used independently or to augment existing PPRL solutions, resolving ambiguities.