EHR Text Categorization for Enhanced Patient-Based Document Navigation.

Area of Science:

• Medical Informatics
• Natural Language Processing
• Computational Linguistics

Background:

• Patient diagnosis lists often contain lengthy ICD-10 codes and free-text descriptions.
• Physician-generated free-text descriptions can overwrite standardized ICD-10 topics, leading to redundancy.
• Current diagnosis list formats present challenges for efficient data management and analysis in healthcare.

Purpose of the Study:

• To develop an accurate method for grouping patient diagnosis lists.
• To reduce redundancy and compress lengthy diagnosis lists.
• To explore the potential of content-based categorization for hospital applications.

Main Methods:

• Utilized a combination of natural language processing (NLP) techniques.
• Employed hierarchical clustering for grouping diagnosis lists.
• Evaluated the approach using an F-measure metric.

Main Results:

• Achieved accurate grouping of diagnosis lists with an overall F-measure of 0.87.
• Demonstrated significant compression of initial diagnosis lists, up to 89%.
• Successfully addressed redundancy in patient diagnosis records.

Conclusions:

• The NLP and hierarchical clustering approach offers an effective solution for organizing complex patient diagnoses.
• This method has the potential for large-scale implementation in hospital information systems.
• Further research should address identified pitfalls and challenges for broader application.