Setting up a DataSHIELD Hub for the German Medical Informatics Initiative: Challenges and Lessons Learned

Area of Science:

• Medical Informatics
• Biomedical Research
• Data Science

Background:

• Secondary use of clinical data offers research opportunities but faces data protection and interoperability challenges.
• The German Medical Informatics Initiative (MII) established Data Integration Centers (DICs) for harmonized, secure research data access.
• Privacy-preserving multi-centric analyses are crucial for research without exchanging individual-level patient data.

Purpose of the Study:

• To implement and evaluate a federated DataSHIELD infrastructure for privacy-preserving multi-centric research within the MII.
• To assess the feasibility of conducting analyses on the biomarker NT-proBNP in atrial fibrillation patients across multiple hospitals without data exchange.
• To identify key challenges and derive recommendations for future federated analysis infrastructure deployments.

Main Methods:

• Implemented a federated DataSHIELD infrastructure connecting local Opal servers from participating hospitals to a central hub.
• Standardized datasets were provided by each site based on variable availability and operational readiness.
• Conducted a study on the biomarker NT-proBNP in patients with atrial fibrillation using the federated setup.

Main Results:

• The federated infrastructure enabled GDPR-compliant analyses across multiple hospitals.
• Substantial manual configuration and maintenance were required for the setup.
• Key challenges included version compatibility, limited analytical functionality, and lack of automated deployment and testing.

Conclusions:

• The study demonstrates the feasibility of federated analysis infrastructure within the MII for privacy-preserving research.
• Significant manual effort and technical challenges were encountered, highlighting the need for improvement.
• Recommendations include standardized installation workflows, better version/access management, and modular functionality extensions for more scalable solutions.