Health Synthetic Data to Enable Health Learning System and Innovation: A Scoping Review.

Area of Science:

• Health Informatics
• Machine Learning
• Data Science

Background:

• Advancements in machine learning have spurred interest in health synthetic data.
• Electronic medical records (EMRs) present challenges in access and usage for research.
• The utility and governance of health synthetic data require further investigation.

Purpose of the Study:

• To conduct a scoping review on the evaluation and governance of health synthetic data.
• To understand the current status of health synthetic data research and application.
• To identify gaps in the standardization and scalability of health synthetic data generation.

Main Methods:

• A scoping review was performed.
• PRISMA guidelines were followed for systematic literature search and selection.
• Evaluations of synthetic data utility and governance were analyzed.

Main Results:

• Properly generated synthetic health data demonstrates low privacy leak risk and comparable quality to real data.
• Health synthetic data generation is currently performed on a case-by-case basis, lacking scalability.
• Regulations, ethics, and data sharing policies for health synthetic data are largely implicit, despite existing common principles.

Conclusions:

• Synthetic health data holds significant potential for research, provided robust generation methods are employed.
• There is a need for standardized approaches to health synthetic data generation and governance.
• Clearer guidelines and policies are necessary to facilitate the ethical and effective sharing of synthetic health data.