AI-Based Conversational Social Support for Older Adults With Heart Failure

This three-year study aims to investigate gender differences in social frailty and evaluate the effects of an artificial intelligence (AI)-based conversational social-support intervention among older adults with chronic heart failure (CHF). The study plans to enroll 300-500 adults aged 60 years or older from four Chang Gung Memorial Hospital branches in Taiwan. In the first phase, researchers will assess participants' clinical characteristics, sarcopenia, physical activity, and physical, psychological, and social frailty. The study will examine gender differences and identify factors associated with social frailty. In the second phase, 174 participants are planned to be assigned to three groups: an AI-based conversational social-support group, a community social participation group, and a control group. The AI group will engage in daily 15-30-minute app-based conversations, while the community group will be encouraged to participate in regular social activities. The control group will maintain routine activities. The third phase will assess longer-term changes in social frailty, emotional well-being, physical activity, and health outcomes, with follow-up assessments at 1, 3, 6, and 12 months after the intervention.

Inclusion Criteria: 1. Participants aged 60 years or older. 2. Diagnosis of chronic heart failure (CHF). 3. Ability to communicate effectively in Mandarin or Taiwanese. Exclusion Criteria: 1. Unstable heart failure. 2. Severe comorbidities, such as end-stage renal disease, advanced hepatic dysfunction, or terminal malignancies. 3. Severe cognitive deficits that impede the ability to provide informed consent or participate in study interventions. 4. Uncorrected visual or hearing impairments that hinder effective communication or interaction with the AI-based conversational social-support intervention. 5. Active psychiatric conditions, including major depression or psychosis, that could interfere with study participation or adherence.