Related Experiment Video
Updated: May 4, 2026
10:43
Eye-tracking Technology and Data-mining Techniques used for a Behavioral Analysis of Adults engaged in Learning Processes
Published on: June 10, 2021
4.4K
Employing heat maps to mine associations in structured routine care data
Dennis Toddenroth1, Thomas Ganslandt2, Ixchel Castellanos3
1Chair of Medical Informatics, University of Erlangen-Nuremberg, Krankenhausstr. 12, 91054 Erlangen, Germany.
Artificial Intelligence in Medicine
|January 7, 2014
Summary
Heat maps enhance the interpretation of electronic medical record (EMR) data, revealing hidden causal effects. This visualization method simplifies the discovery of new medical knowledge from routine patient data.
Area of Science:
- Health Informatics
- Medical Data Mining
- Biostatistics
Background:
- Electronic medical records (EMR) contain vast amounts of data with potential for discovering new medical knowledge.
- Identifying causal effects from statistical associations in EMR data is challenging due to interpretation difficulties.
- Current iterative statistical approaches for analyzing attribute pairs in EMR data are complex and hard to interpret.
Purpose of the Study:
- To improve the interpretation of statistical association matrices derived from EMR data using heat maps.
- To adapt heat maps for detecting causal mechanisms and associations within routine EMR data.
- To aid in identifying determinants of critical health events, such as imminent intensive care unit (ICU) readmission.
Main Methods:
- Utilized heat maps to visualize metric datasets, presenting measures of association between numerous attribute pairs clearly.
- Allocated distinct attribute sets to matrix dimensions based on plausible exposures and outcomes, simplifying visualization.
- Applied specific color schemes and transformations to incorporate attribute similarity and ensure finite distance metrics.
Main Results:
- Heat maps effectively indicated attribute associations and their relationships, with clustering procedures mirroring simulated statistical associations.
- Dendrograms aided in identifying homogeneous attribute associations within contiguous attribute sequences.
- Analysis of routine care data revealed plausible medical constellations, such as the association between breast cancer (ICD C50) and radiation therapy (8-52).
Conclusions:
- Heat maps of association measures are effective for visualizing patterns in routine EMR data.
- The adaptable method balances information display with graphical complexity, simplifying the detection of undiscovered causal effects.
- Pre-existing assumptions about plausible effects can guide the search scope, enhancing the discovery of causal effects in EMRs.
