Domain-specific analytical language modeling--the chief complaint as a case study

Area of Science:

• Computational linguistics
• Medical informatics
• Natural language processing

Background:

• Electronic medical records (EMRs) contain significant free-text data, hindering automation.
• Developing structured data standards for diverse clinical information is challenging.
• Automating the analysis of unstructured medical text is crucial for EMR advancement.

Purpose of the Study:

• To improve automation in EMRs by computationally understanding emergency department chief complaints.
• To develop and evaluate a domain-specific analytical model for chief complaint classification.
• To enable applications like automatic syndromic surveillance using classified chief complaints.

Main Methods:

• Applied domain-specific analytical modeling for computational understanding of chief complaints.
• Developed an algorithm for automatic classification of emergency department chief complaints.
• Incorporated approximate matching to handle typographic variations in text.

Main Results:

• The algorithm achieved accurate classification correctness in a multi-hospital setting.
• Approximate matching did not negatively impact classification correctness.
• Approximate matching significantly increased classification completeness for chief complaints.

Conclusions:

• Computational understanding of emergency department chief complaints is feasible and effective.
• The developed algorithm enhances EMR automation through accurate and complete text classification.
• This approach supports advanced applications like automated public health surveillance.