An AI-driven multi-criteria decision-support system for elderly care personalization

The global rise in life expectancy has led to an increased prevalence of multimorbidity and chronic health conditions among older adults, significantly impacting their quality of life. This growing demand for long-term care presents substantial challenges to the sustainability of health and social support systems. Advancements in Artificial Intelligence (AI) and the integration of digital technologies in clinical practice offer promising avenues for enhancing personalized healthcare. In this context, the present study aims to develop a decision-support system to optimize the personalization of care for older adults, thereby promoting healthier and more sustainable ageing. The proposed system integrates machine learning (ML) techniques with a knowledge-based framework to assist healthcare professionals in selecting appropriate monitoring devices (e.g., fall risk, pressure, arrhythmias, weight) and support services (e.g., walking groups, communication activities, recreational programs, motor-functional rehabilitation, nutritional counselling, and transport) tailored to individual needs. Encouraging results were obtained by three ML methods (Logistic regression, Random Forest and XGBoost) validated on 121 real users, providing accurate and reliable classifications of users’ needs. A structured validation consisting in methodological, quantitative sensitivity analysis and clinical validation of the system confirmed the robustness of such a hybrid approach compared to purely ML-based system, pure rule-based system and a traditional MCDA, and its relevance among clinicians. As a future development, it is planned to consider the user’s opinion obtained from usability and user experience tests to optimize the personalization of the assignment and improve the user’s engagement. Moreover, AI techniques based on large language models could automate the devices classification by extracting unstructured data from product documentation (e.g., devices’ leaflets and brochures).