Large Language Models in Digital Public Health: Data Foundations, Trustworthy Pretraining, and Real-World Applications

The domain of digital public health is rapidly evolving with the emergence of large language models (LLMs), which are poised to revolutionize disease surveillance, risk prediction, and personalized public health interventions. LLMs, when trained on comprehensive multimodal datasets, including electronic health records, imaging, structured laboratory data, social media, and crowdsourced information, hold promise for advancing context-aware systems and supporting real-time public health decision making. However, the deployment of medical LLMs continues to face substantial challenges around data fragmentation, limited interoperability between datasets, unaddressed biases, privacy concerns, and a shortage of standardized benchmarking and evaluation frameworks. Contemporary research has highlighted persistent gaps not only in building reliable and ethical pretraining resources but also in the seamless integration of LLMs into diverse, real-world digital health applications.This Research Topic aims to catalyze breakthroughs in the foundation, training, and clinical and public health validation of large language models for digital and public health scenarios. It will bring together researchers, clinicians, AI specialists, public health practitioners, and policymakers to address central questions around the construction of rich, representative, and interoperable datasets; development of transparent, privacy-preserving, and ethically robust models; and the integration of LLMs into context-aware platforms for infectious disease modelling, health outcome prediction, and participatory digital interventions. Special emphasis is placed on understanding the societal and ethical dimensions, ensuring interdisciplinary readiness, and accelerating safe, equitable, and impactful deployment of LLMs for public health advancement.To gather further insights into advancing and integrating LLMs for digital public health, this Research Topic welcomes submissions that investigate, but are not limited to, the following themes:• Construction and harmonization of multimodal, context-rich pretraining datasets (EHRs, imaging, structured and public health surveillance data, social media, and crowdsourced sources)• Standards, pipelines, and real-time big data stream integration for disease modeling and health outcome prediction• Context-aware and ethically grounded AI frameworks tailored for public health needs• Digital interventions leveraging LLMs, such as serious games, persuasive technology, and personalized recommender systems,for prevention and behavior change• Methodologies for crowdsourcing, mobile health, and participatory systems integration into digital LLM pipelines• Strategies to address demographic and institutional biases, and ensure global and cross-lingual relevance• Privacy-preserving data sharing, secure federated training, and alignment with national and international disease surveillance systems• Benchmarking, clinical validation, and real-world deployment of LLMs in public health settings• Informatics training, interdisciplinary capacity building, and ethical considerations in the design and implementation of LLM-powered digital health solutionsWe acknowledge the initiation and support of this Research Topic by University College London’s RDR Centre of Digital Public Health in Emergencies (dPHE). We hereby state publicly that UCL’s RDR Centre of dPHE has had no editorial input in articles included in this Research Topic, thus ensuring that all aspects of this Research Topic are evaluated objectively, and unbiased by any specific policy or opinion of UCL’s RDR Centre of dPHE. Topic editors: Chunjia Han, Xiang Wu, Yuan Yang, Aisha Aldosery Source: https://www.frontiersin.org/research-topics/77975/large-language-models-in-digital-public-health-data-foundations-trustworthy-pretraining-and-real-world-applications