Evaluating the utility of data integration with synthetic data and statistical matching

Area of Science:

• Bioinformatics
• Data Science
• Epidemiology

Background:

• Data integration improves dataset utility but poses privacy risks.
• Synthetic data is a potential privacy-preserving solution.
• The role of synthetic data in data integration requires further investigation.

Purpose of the Study:

• To assess synthetic data integration's impact on data utility.
• To evaluate varying common variables in statistical matching.
• To explore synthetic-real dataset combinations in donor-recipient scenarios.

Main Methods:

• Utilized the Korean Genome and Epidemiology Study (KoGES) cohort.
• Generated multiple synthetic datasets with varied common variables.
• Performed statistical matching using the nearest-neighbor hotdeck method.

Main Results:

• All-available matched synthetic data generally outperformed other conditions.
• Clinically relevant matching variables sometimes showed equivalent performance.
• Synthetic data demonstrated comparable model accuracy to real data.

Conclusions:

• Statistically matched synthetic data offers utility comparable to real data.
• This approach can reduce privacy risks while preserving data utility.
• Further research is needed to fully understand performance differences.