To address the growing need for intelligent long-term care (LTC) solutions, we introduce SALUS, a modular framework that integrates digital twin technology, conversational AI, and wearable IoT devices. The system maintains a dual digital twin architecture: a physical twin updated in real time with physiological data from non-intrusive smartwatches, and a psychological twin constructed through structured interactions with a Large Language Model (LLM)-based virtual assistant. These models are continuously synchronized to provide a holistic, longitudinal view of the patient’s physical and cognitive-emotional health status. SALUS enables the delivery of validated psychological assessments through natural, multimodal conversations adapted to the cognitive needs of elderly users, while concurrently monitoring biometric trends such as sleep quality, heart rate variability, and physical activity. Predictive analytics and explainable AI components support the early detection of health deterioration and allow caregivers to make timely, evidence-based decisions. The system adheres to ethical AI principles, supports HL7 FHIR-based interoperability, and includes dynamic form management for clinical adaptability. Real-world scenarios demonstrate the framework’s versatility across home care, cognitive rehabilitation, and institutional settings. SALUS contributes to the shift toward data-driven, user-centered LTC models capable of anticipating and responding to individual health trajectories.