Identifying ICU admission decision patterns in a '20-questions game' approach using network analysis

Area of Science:

• Medical Informatics
• Complex Systems Analysis
• Clinical Decision Support

Background:

• Intensive care unit (ICU) admission involves complex, non-linear decision-making with multiple influencing factors.
• Novel analytical techniques are needed to delineate the importance and interrelationships of these factors.
• Network analysis (NA), a graph theory-based method, shows potential for mapping these complex connections.

Purpose of the Study:

• To identify patterns in ICU decision-making for referred patients.
• To determine key factors, their distribution, connections, and relative importance in ICU admissions.
• To compare decision outcomes and case labels across different subgroups.

Main Methods:

• Network analysis (NA) was applied as a secondary analysis to existing data on ICU admission decision-making.
• The dataset was generated from a previous study utilizing a 20-questions game approach.
• Data were standardized and coded to a quaternary level for NA using the Gephi software package.

Main Results:

• The analysis generated 31 nodes and 964 edges, revealing critical factors.
• Properties of acute illness, illness progression, and comorbidities consistently emerged as most important.
• Different NA measures highlighted factors differentially, and modularity analysis identified novel subgroups distinct from traditional classifications.

Conclusions:

• Network analysis provides a comprehensive method for exploring the ICU admission decision process.
• This approach facilitates reflection on decision-making, potentially leading to improved outcomes and new decision support systems.
• Further research with larger datasets is recommended to fully establish NA's role in clinical decision-making.