Abstract
Background
Wearable sensor technology shows promise for neuropsychiatric symptoms (NPS) identification/monitoring in persons living with dementia (PLWD). Accelerometry measures acceleration of body segments and physical activity. This study explores the diagnostic test accuracy (DTA) of accelerometry to identify the following 6 NPS associated with motor activity: aberrant motor behavior (AMB), aggression, agitation, anxiety, apathy, and wandering.
Methods
As part of a larger systematic review, we conducted an extensive literature search in nine health science and engineering databases. Meta‐analysis involved converting correlations to Fisher's Z scores, calculating 95% confidence intervals using the R Studio Metacor package, and assessing fixed effects, DerSimonian and Laird's random effects models, and heterogeneity.
Results
Out of 12,853 identified records, 84 reports were retained for analysis. Eight reports that provided 12 datasets assessed the DTA of accelerometry data for identifying/monitoring apathy (n=4), AMB (n=1), anxiety (n=2) agitation (n=5), aggression (n=2), and wandering (n=2). Studies included one single‐blinded trial, one non‐randomized trial, and ten observational studies. A random‐effect meta‐analysis yielded a pooled correlation across studies of r=0.61 (0.50; 0.70), heterogeneity I2=68%. When the wandering NPS subgroup was eliminated from analysis, heterogeneity fell considerably to I2=36%, with a pooled correlation of r=0.55 (0.47; 0.62).
Conclusions
These findings suggest that accelerometry consistently provides good to moderate DTA for identifying/ monitoring NPS associated with motor behavior in PLWD. Future research should prioritize standardizing the reporting of measurement procedures, signal thresholds, outcome measures, devices, and reference standards.