A methodology for interactive mining and visual analysis of clinical event patterns using electronic health record...

Area of Science:

• Health Informatics
• Data Mining
• Clinical Research

Background:

• Patient medical conditions exhibit complex and unpredictable progression, even within defined care episodes.
• Variations in patient progression and outcomes necessitate methods to understand event patterns in electronic health data.
• Identifying event patterns correlating with outcomes is crucial for retrospective studies.

Purpose of the Study:

• To present a novel method for interactive pattern mining and visual analysis of retrospective clinical patient data.
• To support ad hoc visual exploration of patterns within electronic health records.
• To uncover event patterns impacting patient outcomes and their temporal associations.

Main Methods:

• Developed a method combining visual query capabilities for episode definition.
• Employed pattern mining techniques to discover significant intermediate events.
• Utilized interactive visualization to explore event patterns and their impact on outcomes over time.

Main Results:

• The methodology facilitates interactive specification of clinical episode definitions.
• Pattern mining identifies key intermediate events influencing patient trajectories.
• Interactive visualization reveals event patterns associated with differential outcomes and their evolution.

Conclusions:

• The presented approach enables visual exploration and pattern discovery in clinical data.
• This method aids in uncovering event sequences that correlate with patient outcomes.
• The prototype implementation demonstrates the potential for generating new clinical insights and hypotheses.