Knowledge structure and thematic evolution of host response-oriented sepsis research: a multi-database bibliometric and LDA topic modeling study.

Objective: To characterize the knowledge structure, thematic domains, and temporal evolution of host response-oriented sepsis research and to summarize its major research themes, organizational patterns, and temporal shifts. Methods: Publications related to host response, immune phenotypes, endotypes, and multimarker stratification in adult sepsis were retrieved from Web of Science Core Collection, Scopus, and PubMed. A total of 5,839 records were identified. After exclusion of two records with missing titles, 5,837 records entered a two-stage deduplication workflow based on DOI matching followed by metadata-adjudicated normalized-title matching, yielding 2,974 unique publications. Titles, abstracts, author keywords, and index keywords were concatenated to construct the textual corpus. Bibliometric analysis and latent Dirichlet allocation topic modeling were used to identify thematic domains; each publication was assigned to its dominant topic based on the maximum document-topic posterior probability from the final eight-topic (K = 8) model. Topic-specific annual distributions from 2000 to 2025 were analyzed to characterize temporal evolution. Results: The annual publication output increased steadily after 2000 and entered a marked expansion phase after 2016. The literature involved broad international participation and was published across immunology, critical care, infectious disease, and translational medicine journals. Across K = 4 to 12 solutions evaluated by coherence, perplexity, seed stability, and minimum topic size, an eight-topic model offered the best balance, and eight thematic domains were identified. The largest domains were inflammatory and innate-immune signaling (n=768, 25.8%), clinical management and precision medicine (n=573, 19.3%), and organ dysfunction, endothelial injury, and coagulation (n=392, 13.2%). Temporal analysis based on annual publication counts showed that inflammatory signaling and clinical management remained foundational, whereas transcriptomics, diagnostic and prognostic biomarkers, and ICU outcome themes showed apparent recent increases at the macro-thematic level. Preprocessing sensitivity analyses indicated that macro-level themes were interpretable but preprocessing-dependent. Conclusion: Host response-oriented sepsis research has become organized around several distinct but interconnected thematic domains. Its publication activity shows a descriptive shift from traditional inflammation-centered investigation toward molecular characterization, precision stratification, and individualized management. Transcriptomics, biomarkers, prognostic stratification, and early diagnosis may remain important directions for future research.