Abstract: 【Purpose/significance】Based on patient generated text in online health communities, intelligence analysis is conducted to identify key factors that affect the negative tendency of critical chronic disease patients to comment. This provides important reference for targeted improvement of satisfaction for critical chronic disease patients, improvement of online and offline medical service levels for critical chronic disease, and alleviation of doctor-patient conflicts【. Method/process】Based on the review data of major chronic dis⁃ ease patients in the online health community of Good Doctor, a basic dictionary is constructed and the SOPMI algorithm is used to ex⁃ pand the sentiment analysis method of the sentiment dictionary. The BERTopic method is used to analyze the thematic features of negative reviews of major chronic disease patients.【 Result/conclusion】The key dimensions that affect the negative tendency of critical chronic disease patients to comment on are: the treatment effectiveness of medical services, the quality of doctor-patient communica⁃ tion, professional skills of doctors, medical ethics and personal characteristics, and the maintenance of normalized doctor-patient inter⁃ action. Corresponding strategies and suggestions are proposed based on the key dimensions.【 Innovation/limitation】This article intro⁃ duces text mining technology to the field of online healthcare, and uses deep learning models to explore key factors that affect patient satisfaction based on negative comment data classified from major chronic disease patients. Providing a data science research para⁃ digm for identifying key influencing factors of negative tendencies in critical chronic disease patients′ comments.