Development and feasibility testing of an AI-powered chatbot for early detection of caregiver burden: protocol for a mixed methods feasibility study.

Abstract

Listen

Translate article

Caregivers of patients with end-stage kidney disease (ESKD) face significant challenges that contribute to caregiver burden, negatively impacting their physical, psychological, social, and financial well-being. With the growing prevalence of chronic diseases and an aging population, there is an urgent need for accessible and scalable solutions to detect and address caregiver burden. Artificial Intelligence (AI) chatbots using natural language processing (NLP) have shown promise in providing mental health support and monitoring through natural conversations. This study will contribute to research and clinical practice by: (1) validating a novel approach for early detection of caregiver burden through NLP, (2) analyzing the feasibility of AI-powered chatbots for continuous caregiver monitoring, and (3) informing the development of scalable, accessible tools to identify at-risk caregivers. This protocol for the mixed methods aims to evaluate the feasibility, acceptability, and preliminary effectiveness of BOTANIC (Burden Observation and Timely Aid for Navigating Informal Caregiving), an AI-powered chatbot for early detection of caregiver burden. A single-center validation study will be conducted at Alexandra Hospital, Singapore. Twenty primary caregivers of ESKD patients will be recruited to use BOTANIC for 12 weeks. BOTANIC, developed using Python and open-source libraries, will integrate with Telegram and utilize advanced NLP techniques to analyze caregiver conversations and detect signs of burden. The NLP algorithm will analyze conversations to generate burden scores at baseline and at 12 weeks. Participants will also complete baseline and 12-week assessments using validated questionnaires including the Zarit Burden Interview (ZBI), Patient Health Questionnaire-9 (PHQ-9), and Generalized Anxiety Disorder-7 (GAD-7). Primary outcomes include concordance between caregiver burden levels detected by the NLP algorithm and validated assessment scores at both timepoints. Secondary outcomes include user engagement metrics and system satisfaction. Semi-structured interviews will explore participants' experiences with the chatbot. Quantitative data will be analyzed using descriptive statistics and appropriate statistical tests such as paired t-tests or Wilcoxon signed-rank tests, while qualitative data will undergo thematic analysis. The study has been approved by the NHG Domain Specific Review Board. Findings will be published in peer-reviewed journals, presented at conferences, and used to inform the development of larger-scale trials of AI-powered caregiver support interventions.