Open-Source Synthetic Data Generation of Clinical Routine Data.

Area of Science:

• Medical Informatics
• Health Data Science
• Computational Medicine

Background:

• Clinical routine data is vital for hospital research but faces privacy barriers for decentralized analysis.
• Synthetic data generation presents a privacy-preserving alternative for accessing and utilizing clinical data.

Purpose of the Study:

• To explore the feasibility of generating synthetic electronic patient records from mixed tabular and event-based clinical data.
• To evaluate the effectiveness of an open-source package in handling complex clinical data structures for synthetic data generation.

Main Methods:

• Utilized an open-source software package for synthetic data generation from electronic patient records.
• Employed a combination of static and time-series-based generative algorithms after data preprocessing and cleansing.
• Evaluated synthetic data quality based on the similarity of marginal distributions compared to original data.

Main Results:

• The study demonstrated the potential of synthetic data generation for clinical research applications.
• Challenges were identified in accurately mimicking the full complexity and distribution of real-world clinical data.
• The chosen software package faced rigorous demands due to the data's quality, complexity, and structure.

Conclusions:

• Synthetic data generation shows promise for enhancing clinical research by addressing privacy concerns.
• Further advancements are necessary to develop more sophisticated algorithms capable of replicating the entirety of clinical routine data.
• Continued research is needed to improve the fidelity and utility of synthetic clinical datasets.