Randomised Controlled Trial of Artificial Intelligence-assisted Health Education

With the rapid advancement of biopharmaceutical technology, clinical trials have become the crucial bridge connecting new drugs from the laboratory to clinical application. Despite the increasing number of clinical trial projects being conducted, nearly all such projects face the common challenge of recruitment difficulties. Subject recruitment constitutes a pivotal stage in clinical trials; the ability to recruit a sufficient number of subjects meeting the trial requirements significantly impacts trial quality and also serves as a key factor influencing trial progress. Hematologic cancers constitute a highly heterogeneous group of malignant diseases originating in the haematopoietic organs and primarily affecting the haematopoietic system. They encompass acute and chronic leukaemias, malignant lymphomas, multiple myeloma, myelodysplastic syndromes, and related disorders. For patients facing treatment decisions, clinical trials represent not only a vital avenue for accessing cutting-edge therapies but also impose heightened demands on their capacity for informed decision-making. Conversational artificial intelligence (AI) based on large language models is rapidly advancing in health education and public health communication. Medical chatbots offer scalable and personalised advantages in delivering health information, promoting behavioural change, and enhancing patient engagement, providing a viable pathway for improving trial literacy and decision support. Accordingly, this study proposes to conduct a clinical trial literacy intervention using AI-powered chatbots among haematological malignancy patients. Through a randomised controlled trial (RCT), it aims to evaluate the impact of AI-assisted health education on patients' understanding of clinical trials and intention to participate. This research seeks to validate the application value of AI technology in health education and explore scalable AI-assisted health education intervention models.