Examining the Determinants of Patient Perception of Physician Review Helpfulness across Different Disease Severities: A Machine Learning Approach.

(1) . Patients are increasingly using physician online reviews (PORs) to learn about the quality of care. Patients benefit from the use of PORs and physicians need to be aware of how this evaluation affects their treatment decisions. The current work aims to investigate the influence of critical quantitative and qualitative factors on physician review helpfulness (RH). (2) . The data including 45,300 PORs across multiple disease types were scraped from . Grounded on the signaling theory, machine learning-based mixed methods approaches (i.e., text mining and econometric analyses) were performed to test study hypotheses and address the research questions. Machine learning algorithms were used to classify the data set with review- and service-related features through a confusion matrix. (3) . Regarding review-related signals, RH is primarily influenced by review readability, wordiness, and specific emotions (positive and negative). With regard to service-related signals, the results imply that service quality and popularity are critical to RH. Moreover, review wordiness, service quality, and popularity are better predictors for perceived RH for serious diseases than they are for mild diseases. (4) . The findings of the empirical investigation suggest that platform designers should design a recommendation system that reduces search time and cognitive processing costs in order to assist patients in making their treatment decisions. This study also discloses the point that reviews and service-related signals influence physician RH. Using the machine learning-based sentic computing framework, the findings advance our understanding of the important role of discrete emotions in determining perceived RH. Moreover, the research also contributes by comparing the effects of different signals on perceived RH across different disease types.