Text Classification Model Explainability for Keyword Extraction - Towards Keyword-Based Summarization of Nursing Care...

Area of Science:

• Health Informatics
• Natural Language Processing
• Machine Learning

Background:

• Healthcare professionals require efficient methods to review patient information.
• Nursing entries in electronic health records (EHRs) contain critical patient data.
• Automated summarization tools can improve clinical workflow efficiency.

Purpose of the Study:

• To develop and evaluate a keyword-based text summarization method for nursing entries in EHRs.
• To extract keywords and phrases that intuitively represent the content of multiple nursing notes.
• To assess the performance of the proposed method against a baseline approach.

Main Methods:

• A keyword-based text summarization method leveraging machine learning model explainability was developed.
• The method was applied to generate keyword summaries from 40 patients' EHR nursing entries.
• Performance was compared to a baseline method using word embeddings and PageRank.

Main Results:

• The proposed method successfully generated representative keyword summaries from nursing entries.
• Manual evaluation by domain experts indicated the new method outperformed the baseline approach.
• The keyword extraction method provides an intuitive overview of patient care episodes.

Conclusions:

• Keyword-based summarization is feasible for nursing entries in EHRs.
• The developed machine learning explainability-based method offers superior performance compared to traditional techniques.
• This approach has the potential to enhance clinical decision-making by providing rapid access to patient information.