Topic Modeling Analysis of Diabetes-Related Health Information during the Coronavirus Disease Pandemic

Abstract

This study aimed to provide diabetes-related health information by analyzing queries posted in the diabetes-related online community required during the COVID-19 pandemic. A total of 9156 queries from the diabetes-related online community, dated between 1 December 2019 and 3 May 2022, were used in the study. The collected data were preprocessed for bidirectional encoder representation from transformer topic modeling analysis. Topics were extracted using the class-based term frequency–inverse document frequency for nouns and verbs. From the extracted verbs, words with common definitions were subject to substitution and unification processes, which enabled the identification of multifrequent verb categories by noun topics. The following nine noun topics were extracted, in this order: dietary management, drug management, gestational and childhood diabetes, management of diabetic complications, use and cost of medical treatment, blood glucose management, exercise treatment, COVID-19 vaccine and complications, and diabetes in older adults. The top three verb categories by noun topics were permission, method, and possibility. This study provided baseline data that can be used by clinical nurses to deliver diabetes-related education and management based on information sought by patients.