Abstract
OBJECTIVES: To develop, feasibility test, and evaluate an AI-enabled multimodal monitoring system to integrate safety outcomes, functional recovery and quality of life in elderly and stroke survivors in multiple care settings. METHODS: This was a 24-month mixed-methods study across 12 sites spanned from private homes, assisted living facilities, rehabilitation centers, to acute care transitions. The study was a quasi-experimental design, with three groups: integrated monitoring (n = 112), basic monitoring (n = 89), and usual care control (n = 86). 313 respondents were purposively selected using stratified purposive sampling. The multi modal system combines wearable sensors, ambient environmental control sensors, computer vision algorithms and voice analysis by adaptive fusion architectures. Key outcomes were safety events (falls, hospitalisations), functional independence, and cost-effectiveness. Semi-structured interviews were conducted with a total of 78 care-recipients, 93 caregivers, and 76 healthcare providers, until we reached thematic saturation. RESULTS: The integrated monitoring system achieved fall detection sensitivity of 94.8% (95% CI: 92.6-96.3%) and specificity of 96.2% (95% CI: 94.8-97.4%). Clinical outcomes showed significant improvements compared to usual care: 42% reduction in fall-related injuries (p < 0.001), 37% decrease in emergency department visits (p < 0.001), and 22% greater improvement in functional independence scores at 6 months (p < 0.001). Cost-effectiveness analysis revealed net savings of $15,311 per participant over 24 months. Qualitative analysis identified an enhanced sense of security (92% of participants), improved care coordination, reduced caregiver burden (33.3% reduction), and implementation challenges related to technology integration. CONCLUSION: The AI-enabled multimodal monitoring system demonstrates significant clinical, economic, and quality-of-life benefits for elderly individuals and stroke survivors. The integrated approach outperforms single-modality systems and usual care across multiple outcome domains. Implementation success depends on leadership engagement, technical infrastructure readiness, comprehensive staff training, and ongoing user involvement in system configuration.