Implementation Outcomes for Agitation Detection Technologies in People with Dementia: A Systematic Review

Abstract

Background: Experiencing agitation can be particularly distressing for people with dementia and their caregivers. Using technologies to detect agitation can help monitor and intervene when agitation occurs, potentially reducing overall care and support needs. This systematic review aims to explore the implementation outcomes related to the use of agitation detection technologies in people with dementia. By adopting a taxonomy of implementation outcomes, this review seeks to provide insights valuable for the real-world adoption of such technologies for people with dementia. Methods: Searches were conducted in the following databases: SCOPUS, PubMed, PsychINFO, IEEEXplore, and CINAHL Plus. Included studies were required to have implemented, evaluated, or validated technology with the intention to detect agitation in people with dementia in real-time. Results: On 14 May 2024, 1697 records were identified, and 19 were included in the review. The median sample size was 10, and around two-thirds of the records (n = 12, 63%) used 'multimodal' technologies for detecting agitation. Over half of the records (n = 10, 53%) were reporting from two studies. Across technologies, there was evidence of acceptability and feasibility, though there was a general absence of primary data related to implementation outcomes. There were, however, a number of technical issues and limitations that affected the fidelity and appropriateness of the technology, albeit not unique to people with dementia. Conclusions: There is a need for more empirical data on this topic to maximise uptake and adoption. Future research needs to ensure that the voice of the person with dementia is integrated within the evaluation process.

Keywords: acceptability; agitation; biosensing techniques; cameras; detection; feasibility studies; implementation; monitors.

Figure 1

PRISMA flow diagram. * n = 539 ‘duplicate records identified’—some of the same duplicated reports were identified more than twice. ** using AS Review, reviewed n = 453 (10% + 50 irrelevant records since last relevant) of included records.