How Physician Reviews Affect Online Consultation Demand: An Innovative Small Language Model with Fine-Tuning
Bin Zhang, Haijing Hao, Yongcheng Zhan, Jiang Wu
Information Systems Research · Fri
Abstract
Large language models (LLMs) are advancing rapidly but remain expensive to train and deploy, especially in specialized domains such as healthcare. To improve efficiency and reduce costs, we develop a small language model (SLM) that can be fine-tuned for effective sentiment analysis in this context. Our SLM operationalizes a quality evaluation framework for healthcare service to extract providers’ service quality scores from online physician reviews, and then we estimate the relationship between the quality scores and the online consultation demand. Using a panel data set from one of China’s largest online physician review and consultation platforms, we show that higher service-quality scores are associated with greater consultation demand. Our research contributes in three ways. Methodologically, we develop an SLM (Doc-BERT) tailored to the healthcare context and demonstrate that a task-aligned, domain-tuned SLM can outperform a wide range of existing NLP approaches, including recent general LLMs, for multidimensional sentiment analysis of physician reviews. Theoretically, we adapt and refine a healthcare service quality assessment framework at provider level that has been widely used in healthcare practice but rarely applied in information systems research. Empirically, we identify specific dimensions of service quality that most strongly predict online consultation demand, offering actionable guidance for healthcare professionals and administrators seeking to optimize services and increase uptake. History: Indranil Bardhan, Senior Editor; Atanu Lahiri, Associate Editor. Funding: This work was supported by the Wuhan University, National Natural Science Foundation of China [Grant 72232006]. The views expressed in this article are those of the authors and do not necessarily reflect the views of the funding agency. Supplemental Material: The online appendices are available at https://doi.org/10.1287/isre.2024.1183 .
Keywords
healthcare providers’ service quality, online consultation demand, natural language processing, deep learning, sentiment analysis, small language model, large language model, online physician review
Enable JavaScript for citation tools and additional paper interactions.