Unlocking Data for Statistical Analyses and Data Mining: Generic Case Extraction of Clinical Items from i2b2 and...

Area of Science:

• Medical Informatics
• Systems Medicine
• Data Science

Background:

• Modern IT and Data Warehouses (DWH) are vital for analyzing complex diseases in medical science.
• Existing DWH solutions like i2b2 and tranSMART have limitations in data privacy and handling complex disease data.
• Data import/export challenges arise from diverse schemas, parameter definitions, and varying data quality across clinics.

Purpose of the Study:

• To address the limitations of standard data export from medical Data Warehouses (DWH).
• To enable broader statistical and data mining analyses by overcoming the key-value pair data format.
• To develop a method for transforming clinical data into a usable format for external analytical tools.

Main Methods:

• Development of the Generic Case Extractor (GCE) concept.
• Pivoting key-value pairs from clinical facts into a row-oriented format per patient.
• Implementation of complex pivotisation routines to ensure temporal consistency and handle repeated parameters.

Main Results:

• The Generic Case Extractor (GCE) successfully transforms clinical data into an instance-based representation.
• The transformed data is suitable for direct use in statistical software (SPSS, SAS) and data mining tools (Weka, RapidMiner).
• Temporal consistency and handling of follow-up data are addressed through complex pivotisation.

Conclusions:

• The Generic Case Extractor (GCE) overcomes significant data export limitations in medical informatics.
• This facilitates more comprehensive data analysis and knowledge discovery in systems medicine.
• The GCE is integrated into a broader software platform for systems medicine applications.