Extracting Temporal Relationships in EHR: Application to COVID-19 Patients

Area of Science:

• Data Mining
• Database Systems
• Temporal Data Analysis

Background:

• Association rules are a key data mining technique.
• Temporal Association Rules (TAR) extend this to time-dependent relationships.
• Existing methods lack TAR extraction within Online Analytical Processing (OLAP) multidimensional systems.

Purpose of the Study:

• To adapt Temporal Association Rules (TAR) for multidimensional data structures.
• To introduce a novel method, COGtARE, for efficient TAR extraction in OLAP systems.
• To address the complexity and scalability challenges in temporal data mining.

Main Methods:

• Identifying the transaction-defining dimension in multidimensional models.
• Developing techniques to find time-relative correlations across other dimensions.
• Extending a prior approach to reduce the complexity of extracted association rules.
• Applying the COGtARE method to a real-world dataset of COVID-19 patients.

Main Results:

• Demonstrated the feasibility of extracting TAR from multidimensional OLAP structures.
• The COGtARE method effectively identifies temporal correlations within patient data.
• The approach successfully reduces the complexity of the resulting temporal association rules.
• Successful application to COVID-19 patient data showcases practical utility.

Conclusions:

• The proposed COGtARE method enables effective temporal association rule mining in multidimensional OLAP systems.
• This research bridges a gap in applying TAR to complex data structures.
• The method shows promise for analyzing time-dependent patterns in large datasets, such as patient records.