Prediction of Dementia-related Agitation Using Multivariate Ambient Environmental Time-series Data

TL;DR

This study investigates whether ambient environmental data can predict dementia-related agitation, aiming to enable environmental modifications to prevent agitation episodes in persons with dementia.

Contribution

It introduces a predictive modeling approach using ambient environmental data to forecast dementia agitation, highlighting PWD-specific environmental triggers.

Findings

Ambient environment can predict upcoming agitation episodes.

Environmental triggers for agitation are specific to each PWD.

Predictive models show promising accuracy in agitation prediction.

Abstract

Dementia-related agitation causes high stress for dementia caregivers (CG) and to persons with dementia (PWD). Current clinical research suggests that dementia agitation can be affected or triggered by the ambient environment and other contextual factors. In this study, we evaluate this hypothesis through an analysis of ambient environmental data collected with a remote sensing system deployed in the homes of PWDs and their CGs. Furthermore, we determine if the occurrence of dementia-related agitation can be predicted from ambient environmental data, creating the potential for agitation to be prevented via the environmental alteration. These collected data are used to learn the environmental patterns using a predictive model approach. The agitation labels, used in model training, are provided by the CGs living with the PWDs. The results of the agitation prediction model evaluation suggest that ambient environment can be used as predictors for upcoming dementia-related agitation. We also observed that environmental triggers for agitation are PWD-specific. Future opportunities and techniques used to understand triggers for dementia agitation are also discussed.