Population ageing is increasing the demand for complex care in nursing homes, where continuous medical attention is often limited due to the lack of on-site healthcare professionals. The primary objective of this study is to evaluate the feasibility of applying artificial intelligence to telemedicine in residential care, with a particular focus on developing and validating a predictive model for hospital referral. We conducted a multicentre observational study using a telemedicine system equipped with advanced diagnostic devices to manage acute events in 70 nursing homes across Spain. Clinical and socio-demographic data obtained during telemedicine consultations from 5,192 anonymized patients were used to train a deep learning model. Preliminary findings show that the neural network achieved an accuracy of 0.93 and an AUC of 0.90 in predicting hospital referrals, substantially outperforming traditional classifiers. These results provide early evidence of the potential of leveraging data generated through telemedicine for the development of AI-based predictive models in geriatric care. Future work will focus on expanding the dataset, refining the modelling pipeline, and integrating the predictive model into the telemedicine platform to support real-time clinical decision-making in nursing homes.