Structural biases in digital health intervention and health equity research: a bibliometric knowledge graph analysis (2015-2025).

OBJECTIVE: This study systematically analyzed the knowledge structure, evolution patterns, and structural biases in the intersection of digital health interventions, patient engagement/adherence, and health equity from 2015 to 2025. METHODS: We searched Scopus, Web of Science Core Collection, and PubMed. We included 4,677 original articles or systematic reviews. We used Bibliometrix, SciExplorer, and CiteSpace for bibliometric and knowledge graph analysis. RESULTS: The annual number of publications increased from 75 to 1,317 (R2 = 0.9955). The average annual growth rate was 33.18%. However, author keywords were highly dispersed (9,302 vs. 4,971). Geographical distribution showed "one superpower and multiple strong countries." The United States accounted for more than 54% of the top ten countries' frequencies. China ranked fifth but did not enter the core collaboration network. JMIR series journals published 58.8% of the top ten journals' total articles. Keyword clustering identified ten themes, including telemedicine and chronic disease management. "Health equity" did not become an independent high-frequency cluster. The author collaboration network showed a multi-center structure with methodologists, theory builders, and tool developers as cores. CONCLUSION: This field faces "knowledge disorder under exponential growth." There are systematic biases in geography, theme, and power structure. We recommend building a core literature database, creating interdisciplinary journals, supporting reverse innovation led by low- and middle-income countries, and reforming academic evaluation systems to bridge the gap between research and policy translation.