Effectiveness and strategies of social prescribing in Korea using a machine learning topic modeling

Abstract Background Social prescribing, a non-medical approach linking individuals to community-based services to improve health and well-being, has expanded globally, including Korea. Despite their increasing adoption, there is limited systematic evidence evaluating the effectiveness, implementation strategies, and policy implications of social prescriptions in the Korean context. Methods This study conducted a scoping review of the literature related to social prescriptions in Korea. English- and Korean-language articles were retrieved from five databases (Google Scholar, Web of Science, PubMed, Scopus, and KCI) without time restrictions. The studies were screened and selected based on predefined criteria. Machine learning-based topic modeling (LDA, NMF, and BERTopic) was applied to extract the latent thematic structures from the included studies. The evaluation metrics included coherence score, perplexity, topic diversity, and topic balance-guided model selection. The NMF model was selected for final analysis because of its superior performance. Results Six key thematic categories were identified from 16 studies: (1) Mental Health, (2) evaluation, (3) program, (4) Social Issues, (5) COVID-19, and (6) international comparisons. Mental health and social isolation have emerged as major concerns, particularly in aging rural populations. Programs focusing on gardening, music, and digital platforms have been reported to be effective in improving psychological wellbeing and community engagement. The analysis also highlights the necessity of localized models tailored to Korea’s demographic and policy landscape. Conclusions This study emphasized the need for a comprehensive policy framework for social prescriptions in South Korea. The integration of digital technology for remote delivery, adaptation to rural health gaps, and benchmarking from established international models is recommended. This study demonstrates the utility of AI-driven text mining as an innovative approach for evidence synthesis and policy planning for public health.