Purpose: This study aims to construct a human-computer interaction design model ‘SAFER’ for the problem of Alzheimer's disease patients’ cognitive decline that leads to wandering, to optimise the interaction experience between the product and the patient, the caregiver and the emergency service personnel, and to provide an innovative solution for the caregiving product. Methods: The cognitive characteristics and caregiving needs of patients were analyzed through literature review and questionnaires (n = 312, including family members of patients, caregivers and healthcare providers). User needs were classified using the KANO model, and functional weights were determined using the Analytic Hierarchy Process (AHP). The research process strictly followed ethical principles and protected participants’ privacy. Results: The study established the ‘SAFER’ HCI design model with five core dimensions: Simplicity, Adaptability, Feedback enhancement, Emergency handling, and Reliability, with quantified weights for each dimension. The interface scheme based on this model adopts a 3 × 3 grid layout, 64 × 64px minimum touch areas, and multimodal feedback mechanisms. User testing (n = 45) showed a task completion rate of 92.4%, a satisfaction score of 78.5, and a reduction in emergency response time to 15 s. Significance: This study provides a systematic theoretical framework and practical approach for designing caregiving products for people with Alzheimer's disease. It establishes a design mapping from cognitive characteristics to interface implementation and supports multi-party collaboration needs. The model offers a methodological reference for designing assistive products for people with cognitive disabilities, improving both caregiving efficiency and patients’ quality of life.