Evolution of Digital Health Research in Heart Failure: Bibliometric Analysis and Topic Modeling Study (Preprint)

BACKGROUND Digital health research in heart failure (HF) has expanded rapidly, encompassing telemedicine, remote monitoring, patient self-management, and artificial intelligence (AI)–enabled applications. However, publication growth alone may not indicate whether technological expansion is accompanied by the accumulation of clinically evaluative evidence, and research areas with similar growth trajectories may differ substantially in their evidence structures and functional roles. A multidimensional assessment is therefore needed to distinguish these heterogeneous patterns and clarify how emerging technologies are being integrated into digital HF care. OBJECTIVE This study aimed to characterize the evolution of digital health research in HF by integrating thematic growth, evidence structure and knowledge impact, semantic evolution, knowledge overlap, and cross-database reproducibility, with particular attention to how telemedicine, patient self-management, remote monitoring, and AI-enabled research have evolved and whether rapid technological expansion has been accompanied by a corresponding accumulation of clinically evaluative evidence. METHODS Publications on digital health and HF from January 1, 2018, to June 21, 2026, were retrieved from the Web of Science Core Collection (WoSCC), with PubMed used to assess cross-database reproducibility. After screening, 3094 WoSCC records were retained, of which 3063 with usable abstracts were included in topic modeling. Latent Dirichlet allocation was used for topic modeling, with the number of topics selected based on model diagnostics and semantic interpretability and topic labels independently evaluated by 2 researchers. Structural topic modeling and nonnegative matrix factorization were used to assess the robustness of the identified thematic structure. Temporal growth, research-type composition, high-impact publication concentration, within-topic semantic continuity, and cross-topic knowledge overlap were evaluated. Sensitivity analyses examined alternative topic-assignment thresholds and time windows, and cross-database reproducibility was assessed using the PubMed corpus. RESULTS Five major topics were identified: Telemedicine and Remote Care Services, Digital Patient Empowerment and Self-Management, AI-Driven Precision Management, Data-Driven Risk Identification, and Device-Mediated Remote Monitoring. From 2018 to 2025, Data-Driven Risk Identification and AI-Driven Precision Management showed the fastest growth, with compound annual growth rates of 59.27% and 45.49%, respectively, whereas Digital Patient Empowerment and Self-Management and Device-Mediated Remote Monitoring grew more slowly (9.59% and 9.89%, respectively). Evidence structures differed substantially across topics: algorithm/model studies accounted for 50.7% of AI-Driven Precision Management research, whereas randomized controlled trials and systematic reviews/meta-analyses accounted for 19.2% and 22.8%, respectively, of Digital Patient Empowerment and Self-Management research. The concentration of high-impact publications also differed across topics (χ²=21.28, P=.002), ranging from 7.31% for Telemedicine and Remote Care Services to 2.03% for Device-Mediated Remote Monitoring. Cross-model analyses, sensitivity analyses, and cross-database comparisons with PubMed broadly supported the robustness of the main thematic patterns. CONCLUSIONS Digital health research in HF showed heterogeneous patterns of thematic evolution. Rapid growth in AI-driven and data-driven research was not accompanied by a comparable accumulation of clinically evaluative evidence, whereas the more slowly growing patient empowerment and self-management topic contained a greater proportion of randomized controlled trials and evidence syntheses. Publication growth alone may therefore be insufficient to characterize research development. Future research should prioritize rigorous clinical evaluation and examine how rapidly expanding digital technologies can be integrated into clinically meaningful, patient-centered HF care pathways.