Assistive technologies on safety monitoring for community-dwelling persons living with dementia
Year: 2026
Type: article
Study design: Observational Study
Abstract: Individuals with dementia face safety-related incidents at home. However, research predicting such incidents using assistive technologies is limited. This observational study explored the feasibility of a machine learning approach using penalized logistic regression to predict safety-related incidents among community-dwelling persons living with dementia. From July 2021 to October 2022 in South Korea, 67 dyads of persons with dementia and their caregivers provided surveys, a 14-day safety diary, Actiwatch2 actigraphy, and smart motion sensor data. As the target was whether a dyad had experienced any incident during the window, the analysis was conducted at the participant level, with each dyad contributing one observation. A penalized logistic regression model was trained using cross-validation grouped by participants. Discrimination was summarized by the area under the curve (AUC) with bootstrap 95% confidence intervals (CI). A clinical baseline model predicted which dyads experienced an incident (AUC 0.726, 95% CI 0.591–0.845); adding sensor summaries, sleep (actigraphy) summaries, or both did not improve discrimination (all AUC ≤ 0.703). Neuropsychiatric symptoms and the number of medications were the strongest predictors, whereas technology-derived features contributed little. These exploratory findings suggest potential feasibility rather than clinical readiness, and require confirmation in larger, properly powered, externally validated studies.
Source: Scientific Reports
Institutions: Yonsei University, Korea University, Gachon University
Language: English
Subfield: Psychiatry and Mental health
Field: Medicine
Domain: Health Sciences
Open Access status: gold
Estimated APC paid: $2,690
Awards: G1457387214, G2693913580, G4172731568
OpenAlex ID: W7234733146
