Examining between-website bias in online physician platforms: an empirical investigation based on patient reviews

Purpose This study aims at revealing discrepancies in characteristics and patient concerns among different online physician platforms. Design/methodology/approach An integrated Latent–Dirichlet Allocation (LDA), backpropagation neural network (BPNN) and importance-performance analysis (IPA) text-mining approach is proposed to compare both review ratings and contents of online physician reviews across information-driven and transaction-driven platforms. Findings Approximately 240,000 online reviews from different physician platforms were analyzed, and 4 insights were yielded. First, physicians’ review ratings are generally higher and less dispersed on transaction-driven platforms than information-driven ones. Second, transaction-driven platform reviews exhibit higher polarity, subjectivity, diversity and shorter lengths than those on information-driven platforms. Third, information-driven platforms offer the most readable content. Fourth, transaction-driven platform users prioritize efficiency attributes, while information-driven platform users gravitate towards interpersonal attributes. Practical implications These findings assist in understanding the users’ characteristics and implementing tailored business strategies. Originality/value This study proposed an LDA-BPNN-IPA integrated method for text mining and online reviews analysis, categorized online physician platforms into information-driven and transaction-driven platforms according to the corresponding characteristics, revealing characteristics discrepancies across different online physician platforms considering ratings and contents simultaneously and analyzing patients’ concerns and satisfaction conditions using an integrated LDA-IPA approach.