Jianfeng Chen, Leyu Tao
Due to the ageing of the population, multimorbidity and polypharmacy has significantly increased, causing medication safety to become a global health concern. Traditionally, medication risk assessment methods based on the counts of medications or rule-based criteria are inadequate to perceive the dynamic, individual and longitudinal nature of medication-related risks in older people. Artificial intelligence (AI) has shown promise in retrospective predictive modelling for precision medication and medication risk stratification, although prospective evidence demonstrating clinical effectiveness remains scarce. This narrative review presents the current evidence on AI-based polypharmacy risk prediction approaches in older people with a focus on machine learning, deep learning, natural language processing, graph neural network, large language models, and longitudinal trajectory modelling. We highlight some of the major data sources, predictive algorithms, clinical decision support systems and real-world applications, stressing mostly on patient-centred medication management, nurse-led interventions, transitional care and continuity of care. Furthermore, the review highlights significant challenges specifically related to model interpretability, data quality, external validation, ethical governance, privacy protection and clinical implementation which currently limits the translation of AI into routine practice. Future research needs to focus on longitudinal dynamic risk prediction and multimodal data integration. The validation of Esplanade AI by multiple centers as well as collaboration among various fields, will support better medication safety and health outcomes for Older Adults. Harmonization of contemporary AI technologies with geriatric drug use through a public health lens will provide possible guideposts for precision prevention, healthy ageing and sustainable health care delivery.