JMIR Publications

ABSTRACT

Background:

Operational teams supporting people living with dementia in smart homes face growing workloads from routine tasks, technical troubleshooting, and responding to alerts. Calling participants is time-consuming, and without additional context, prioritising tasks is challenging. These calls also represent an opportunity to enhance monitoring by analysing speech as a potential marker of dementia progression. Scalable technological solutions are therefore needed to augment human monitoring while capturing clinically meaningful signals from everyday interactions.

Objective:

This study aims to develop, integrate and evaluate a scalable conversational service with embedded speech analysis that is used in smart homes within the UK Dementia Research Institute Care Research and Technology Centre. This system will provide support for routine tasks and facilitate daily communication with participants while reducing the workload of the operational teams.

Methods:

This study aims to develop, integrate and evaluate an automated conversational platform with embedded speech analysis in existing smart home systems for dementia. This is a novel study which adopts a staged, mixed-methods approach for early co-design, system development, technical integration, and real-world participant testing. The service will enable routine communication with households to help with managing simple queries and reduce the workload of operational teams. Initial deployment will focus on core functionalities, with features introduced iteratively based on user feedback and experience. Alongside feasibility outcomes such as adoption, retention, usability, and operational impact, the study will also support the development of speech analysis algorithms used to identify speech and linguistic markers associated with cognitive decline that can be used to inform clinical decision. Overall, this protocol is designed to guide the ethical, technical, and methodological development of a complex digital health intervention towards future large-scale evaluation.

Results:

We developed a prototype of the conversational service in a sandbox environment to simulate integration with smart home alerts and operational dashboards, ensuring automated interactions could be linked to existing workflows. Reviews of iterations with the clinical team, operational team and software team began in December 2025 and are ongoing. The recruitment and deployment will begin in the second quarter of 2026 through the UK Dementia Research Institute Care Research and Technology Centre. Data collection will run for at least 6 months, with new features being introduced based on participant and stakeholder feedback. We hypothesise that participants and operational teams will find the system acceptable, will engage with the automated conversations and will benefit from reduced workload supporting task prioritisation. Furthermore, speech biomarkers of cognitive health may support clinical and care decision-making.

Conclusions:

The outcomes of this study will improve the scalability of smart home support systems and, through voice analysis, provide novel insights into cognitive health, supporting timely clinical decision-making and enabling people living with dementia to remain at home for longer.