JMIR Publications

ABSTRACT

Background:

Synthetic data generation (SDG) is increasingly used in health research to address privacy concerns and data sharing barriers, particularly in cross-border and multi-institutional contexts. Despite growing interest, questions remain around the methodological soundness, legal status, and real-world trustworthiness of synthetic health data.

Objective:

This paper examines how SDG is operationalized across seven European projects in the HealthData4EU cluster. It identifies common methodological, compliance, and acceptance challenges, and highlights emerging strategies to ensure the quality and utility of synthetic health data.

Methods:

A qualitative multiple case study approach was used, combining workshop presentations with structured input from each project. Projects were analyzed across four aspects: methodological challenges, data quality assurance, trust and transparency strategies, and contributions to the wider synthetic data ecosystem.

Results:

Results reveal recurring challenges across projects: tension between privacy and utility, lack of shared quality validation standards, institutional readiness gaps, and legal ambiguity under GDPR. Projects employ varied strategies, including federated machine learning and validation pipelines, differential privacy, co-design processes, and metadata transparency tools, but implementation remains uneven. The absence of harmonised legal and evaluation frameworks continues to hinder broader uptake.

Conclusions:

To scale SDG from isolated pilots to trusted European infrastructure, four priorities are critical: shared evaluation metrics, legal clarity, sustainable platforms, and embedded stakeholder engagement. While rooted in specific projects, the insights offer broader relevance for global SDG efforts, as the core tensions identified are likely to recur across health systems.