Data Element Mapping in the Data Privacy Era.

Area of Science:

• Bioinformatics
• Health Informatics
• Data Science

Background:

• Secondary use of health data is hindered by semantic heterogeneity.
• Data privacy concerns limit the sharing of sensitive health information.
• Aligning local terminologies with international standards is crucial for data interoperability.

Purpose of the Study:

• To develop a machine learning approach for aligning biological data elements.
• To enable privacy-preserving sharing of aggregated health data features as open data.
• To address semantic heterogeneity in health data for secondary use.

Main Methods:

• A three-step methodology involving feature engineering, a blocking strategy, and supervised learning was employed.
• Machine learning models were trained to align biological data elements.
• Aggregated features were used to maintain data privacy.

Main Results:

• The proposed machine learning methodology demonstrated initial success in aligning biological data elements.
• The approach showed potential for overcoming semantic heterogeneity in health data.
• Modest but encouraging results were achieved in the first phase of development.

Conclusions:

• Machine learning offers a viable solution for aligning biological data elements despite semantic heterogeneity.
• The developed approach supports privacy-preserving open data sharing for secondary health data use.
• Further development is warranted to enhance the performance and scalability of this method.