Automating Large-scale Health Care Service Feedback Analysis: Sentiment Analysis and Topic Modeling Study

Area of Science:

• Health Informatics
• Natural Language Processing
• Data Visualization

Background:

• Patient feedback is crucial for assessing healthcare quality and lived experiences.
• Free-text feedback offers unbiased insights but is challenging to analyze quantitatively.
• The National Health Service (NHS) in England collects patient feedback via a web service.

Purpose of the Study:

• To develop a novel data analysis and interactive visualization pipeline for NHS patient feedback.
• To facilitate the interrogation of and provide unique insights into patient experiences.
• To create an accessible web application for public use.

Main Methods:

• Utilized natural language processing (NLP) and machine learning for text analysis of free-text reviews.
• Employed a three-iteration design science research methodology: sentiment analysis, topic modeling, and nested topic modeling.
• Developed an interactive web application with geographic data visualization.

Main Results:

• Analyzed 51,845 reviews across 2030 clinical services in England (Oct 2017-Sep 2019).
• Identified dominant topics, sentiment-specific topics, and trends in topic/sentiment occurrence.
• Found that highly polarized sentiment reviews were more frequent.

Conclusions:

• Contemporary NLP techniques effectively characterized unstructured patient feedback data.
• An efficient pipeline combined with a web application enables accessible automated analysis and dissemination.
• This work advances the generation and visualization of actionable insights from patient reviews.