Healthbots for conducting clinical screening and remote monitoring with patient mood assessment: A scoping review.

BACKGROUND: Patient mood assessment is key in managing chronic diseases but is often overlooked. Although conversational agents enhance telemonitoring and engagement, few healthbots incorporate automated mood analysis into routine clinical workflows or hybrid care. The rise of multimodal and large language models presents new opportunities to embed emotional assessment into daily healthcare interactions. OBJECTIVE: This scoping review aims to identify existing AI-based healthbots that combine clinical screening and remote monitoring with mood assessment. Its secondary objectives are to (1) describe their technological architectures and AI methods, (2) examine validation and evaluation strategies, and (3) identify current research gaps. METHODS: Following the Arksey and O'Malley framework and PRISMA-ScR guidelines, a comprehensive search was conducted across seven databases (ACM Digital Library, Embase, IEEE Xplore, PubMed, Scopus, SpringerLink, Web of Science), covering the period from January 2020 to December 2024. Studies were included if they presented empirical evidence of AI-based clinical screening with mood assessment. Ten studies met the inclusion criteria after screening and deduplication. Data were charted and synthesized based on key dimensions, including technological features, validation methods, and limitations. RESULTS: Ten studies, mostly in mental health, used multimodal inputs (voice, facial expressions, text) with CNNs, LSTMs, NLP, and LLMs via web or mobile platforms. Some achieved high accuracy on public data and in cross-validation, but few conducted external or longitudinal validation in clinical settings. Integration with EHRs and standards was rarely reported. Limitations included small, homogeneous samples, limited generalizability, insufficient explainability, and privacy concerns. CONCLUSIONS: AI-driven healthbots with mood assessment show promise but are still immature. While they can recognize emotions, few are validated in real clinical settings or integrated into workflows. Broader adoption depends on long-term validation, explainable, bias-aware algorithms, EHR interoperability, and ethical standards. Progress requires collaboration among technical, clinical, and policy experts to ensure the safe and equitable use.