Taming EHR data: using semantic similarity to reduce dimensionality.

Area of Science:

• Health Informatics
• Data Science
• Medical Data Mining

Background:

• Medical care data holds significant potential for health management and clinical research.
• Data heterogeneity and complexity hinder the application of data mining techniques, leaving valuable insights untapped.

Purpose of the Study:

• To develop a methodology for reducing the dimensionality of primary care data.
• To enhance the suitability of medical data for visualization, mining, and clustering.

Main Methods:

• The study employed a combination of ontology-based semantic similarity and principal component analysis (PCA).
• This methodology maps primary care data into a low-dimensional space.
• Anonymised patient data from primary care in Salford, UK, was utilized.

Main Results:

• The developed methodology successfully reduced the dimensionality of primary care data.
• Diagnosis codes were effectively used to map patients into an informative low-dimensional space.
• The approach facilitates further data exploration and medical hypothesis formulation.

Conclusions:

• The proposed methodology makes complex primary care data more amenable to data mining and visualization.
• This technique offers a valuable tool for exploring medical data and generating new research hypotheses.
• The findings support the untapped potential of medical data for advancing healthcare and research.