Predicting and Early Detection of Delirium through Motion Patterns: A Narrative Review

Delirium is a common acute neuropsychiatric syndrome, and its early detection may improve clinical outcomes. This narrative review synthesized findings from 11 original studies and two systematic reviews that employed wearable sensors (actigraphy) to predict or detect delirium. In surgical, intensive care unit, and geriatric populations, delirium has consistently been associated with disrupted rest-activity rhythms, including lower daytime activity, increased nighttime activity, and fragmented sleep-wake cycles. Characteristic motor patterns also differed based on the motor subtype (hyperactive vs. hypoactive). Several studies have demonstrated that continuous wrist accelerometry can objectively detect the onset of delirium and classify motor subtypes. Notably, one machine learning model showed improved prediction accuracy, increasing from approximately 62% to 74% when motion features were included. Overall, continuous motion monitoring appears feasible and may serve as a promising non-invasive tool for early delirium detection and risk stratification. However, the findings remain heterogeneous, and motion-based algorithms alone show only moderate sensitivity. Further validation in larger and more diverse cohorts, as well as integration with clinical risk factors, is required before clinical implementation.