Digital Markers for Behavioral Symptoms in Dementia Care
Original Title: Dementia Care Research and Psychosocial Factors
Journal: Alzheimer's & dementia : the journal of the Alzheimer's Association
Overview
Overview
This study investigates the utility of Real-Time Location Systems (RTLS) in inpatient dementia care units to monitor behavioral and psychological symptoms. While these systems are primarily deployed for safety and nurse calls, the research explores their potential to provide longitudinal insights into resident health. The Space-Time Indices for Clinical Support project utilized data from 47 participants with a mean Mini-Mental State Examination score of 5 out of 30. Over an average of seven weeks per participant, location data was used to build machine learning models for detecting motor agitation and identifying rest-activity rhythms. Results showed that models could distinguish motor agitation from normal activities with an Area Under the Receiver Operating Characteristic curve (AUROC) of 0.81. Furthermore, the analysis identified six distinct digital phenotypes for rest-activity, ranging from well-regulated circadian rhythms to severe rhythm disturbances.
Novelty
Novelty
The primary innovation lies in repurposing existing infrastructure within care facilities to derive clinical markers without requiring additional wearable sensors that residents might find intrusive. Unlike traditional methods relying on periodic manual observations, this approach utilizes continuous spatio-temporal data to quantify behavior. The study employed SHAP explainability analysis to determine that 17 of the top 20 model features for agitation detection were derived directly from the location system, with movement speed and total distance serving as the key predictors. Additionally, the application of unsupervised deep learning to identify six specific rest-activity profiles represents a shift toward data-driven phenotyping. This allows for the categorization of residents into groups such as those with high night activity or significant day-to-day instability, providing a nuanced understanding of behavioral patterns often difficult to capture through standard clinical scales.
Potential Clinical / Research Applications
Potential Clinical / Research Applications
These findings suggest that location data can be integrated into clinical decision support systems to help staff monitor the effectiveness of behavioral interventions or medication changes in real-time. Clinicians can use the identified rest-activity phenotypes to tailor individualized care plans, such as adjusting light exposure for residents in the high night activity cluster. In research settings, these digital markers provide objective, continuous endpoints for clinical trials, reducing the burden on caregivers to complete subjective rating scales. Future studies could expand this work by validating these markers across diverse long-term care environments and exploring how these patterns correlate with long-term cognitive decline.