Diabetes: Bibliometric and Semantic Analysis with LDA Technology to Identify Current Trends in Research Directions

Background: The rapid growth of the diabetes-related literature makes it difficult to characterize the thematic structure of the field using conventional bibliometric indicators alone. This study aimed to combine bibliometric network mapping with probabilistic topic modeling to describe major research directions in diabetes and their citation visibility. Methods: We analyzed 211,637 Scopus-indexed publications classified as articles, reviews, short surveys, or conference papers and published between January 2015 and May 2025. Bibliometric mapping was performed using VOSviewer, and semantic analysis was conducted on publication titles using Latent Dirichlet Allocation (LDA). Model interpretation was supported by coherence score, topic diversity, perplexity, dominant-topic assignment, and expert validation of topic labels. Field-Weighted Citation Impact (FWCI) was used descriptively to compare citation visibility across thematic areas over time. Results: VOSviewer identified seven keyword co-occurrence clusters, while LDA produced six broad semantic topics: cardiovascular and metabolic risks, diabetes management and quality of life, therapy approaches, diabetes complications, gestational diabetes, and pathophysiology and modeling. The six-topic LDA model showed c_v = 0.392, topic diversity = 0.73, and perplexity = −7.18. Publication-share and FWCI trajectories indicated that cardiovascular and metabolic risk research remained the largest thematic area, whereas quality of life and disease control showed the most marked increase in citation visibility. Conclusions: The integrated bibliometric–semantic workflow identified the main thematic domains and citation-visibility patterns shaping contemporary diabetes research. The findings highlight the continued dominance of cardiometabolic comorbidity research, the growing visibility of quality-of-life and disease-control topics, and the visibility of digital and Artificial Intelligence (AI)-assisted approaches in complication screening and monitoring. This framework provides an interpretable basis for tracking research priorities and emerging directions in diabetes-related scientific literature.