This invention discloses a multimodal health management method and
system for TIPS postoperative care, specifically relating to the field of medical and health management technology. It involves constructing a postoperative multimodal follow-up dataset by collecting patient health consultation records,
dietary protein intake records,
oral medication adherence records, and doctor-
patient feedback records. Through consultation topic classification,
protein intake deviation identification,
medication adherence breakpoint detection, and feedback tendency determination, a set of behavioral events is extracted. Within the same follow-up period, co-occurrence associations of consultation, diet, medication, and feedback events are established to generate a risk trigger chain for
hepatic encephalopathy. Based on the risk trigger chain, the intervention type is determined, generating individualized intervention instructions such as dietary guidance, medication reminders, and health consultation responses. These intervention instructions are then pushed to both the patient and healthcare professionals, collecting execution and treatment feedback and writing it back into the dataset to form a closed-loop follow-up update. Finally, the updated data is used to correct the risk trigger chain and automatically generate intervention instructions for the next cycle.