Generating Enriched Synthetic German Hospital Claims Data - A Use Case Driven Approach

Area of Science:

• Health Informatics
• Medical Data Science

Background:

• Clinical data is highly sensitive and requires robust privacy protection.
• Access to realistic clinical data is essential for developing advanced data analysis methods and decision support systems.

Purpose of the Study:

• To present a mixed approach for generating semantically and clinically realistic synthetic patient data.
• To enable students and researchers to develop clinical decision support systems using modern data analysis methods without compromising patient privacy.

Main Methods:

• Utilizing existing synthetic data, extracting and adapting patient visit and diagnosis information to German claims data encoding.
• Performing statistical analysis of real German hospital data to identify distributions of procedures and measurements.
• Semi-automatically transferring identified data distributions to synthetic patient records.

Main Results:

• Generated a synthetic dataset that is semantically and clinically realistic.
• Enabled the application of patient-level prediction algorithms for developing clinical decision support systems.
• Ensured complete protection of patient data privacy.

Conclusions:

• The mixed approach successfully generates realistic synthetic clinical data.
• This method facilitates research and education in clinical decision support systems without privacy risks.
• It provides a valuable, privacy-preserving resource for data analysis and algorithm development.