Detecting Topic and Sentiment Trends in Physician Rating Websites: Analysis of Online Reviews Using 3-Wave Datasets

Area of Science:

• Computational Social Science
• Health Informatics
• Natural Language Processing

Background:

• Physician rating websites (PRWs) offer insights into patient experiences and healthcare concerns.
• Analyzing online patient reviews (PORs) provides a dynamic understanding of public sentiment and evolving health topics.
• The study leverages a large dataset of PORs from U.S. platforms spanning 2018-2020.

Purpose of the Study:

• To investigate and analyze prevalent patient concerns and sentiment dynamics within physician online reviews.
• To identify shifts in healthcare topics and associated sentiments over time, particularly influenced by the COVID-19 pandemic.
• To evaluate the effectiveness of computational methods in analyzing large-scale text data from PRWs.

Main Methods:

• Collected text data from four U.S.-based PRWs across 2018, 2019, and 2020.
• Employed dynamic topic modeling to identify key healthcare themes.
• Utilized a hybrid approach of aspect-based sentiment analysis to analyze topic-specific emotions and social networks.

Main Results:

• Identified 30 dominant topics, with Stage III (2020) significantly differing due to COVID-19.
• Prevalent keywords included treatment, patient questions, communication, hospital environment, symptoms, and COVID-19 related terms.
• Provider service quality and treatment cost were dominant in Stages I & II, shifting to COVID-19 information and countermeasures in Stage III.
• Sentiment analysis revealed a shift from positive (joy) in Stage I to negative (disgust) in Stage II, and initial anger transitioning to trust during Stage III (COVID-19).
• The proposed computational method demonstrated superior performance compared to conventional machine learning models.

Conclusions:

• Computational techniques are vital for analyzing large text corpora and complementing traditional social science research.
• Dynamic topic and sentiment analysis of online patient reviews offer valuable insights into public health concerns and emotional responses.
• The study highlights the impact of major events like COVID-19 on public discourse regarding healthcare and provider performance.