Modeling the evolution of virtual reality in nursing education: a BERTopic-based analysis of research trends and future directions

This study employs BERTopic, an advanced natural language processing (NLP) technique, to systematically analyze the thematic evolution and research hotspots of virtual reality (VR) applications in nursing education from 2008 to 2025. Using a corpus of 683 peer-reviewed articles from Web of Science, we applied BERTopic's transformer-based embedding and hierarchical clustering pipeline to identify latent topics, quantify their temporal trends, and visualize inter-topic relationships through uniform manifold approximation and projection (UMAP) dimensionality reduction. Three dominant research streams emerged: (1) technical applications, (2) humanistic skill development, and (3) specialized high-stakes training. The COVID-19 pandemic accelerated VR adoption, with publications surging by 95% in 2020. Topics evolution revealed a shift from feasibility studies (pre-2018) to outcome optimization (post-2020), particularly in AI-integrated virtual patients and haptic feedback systems. Instructors can leverage topic prominence data to prioritize VR curricular integration, while policymakers should address disparities in cultural adaptability research (only 12% of studies involved non-Western contexts). Notably, this study applies dynamic topic modeling in nursing education research, offering a data-driven framework for tracking technological adoption and predicting future trends.