Unsupervised Learning Identifies Sleep Disturbance Subtypes Among Dementia Caregivers
Introduction
Providing long-term care for individuals with dementia is an arduous role that frequently results in severe sleep disruption. While sleep problems are known to affect two-thirds of caregivers, new research leverages unsupervised machine learning and wearable technology to identify distinct subtypes of sleep disturbance, allowing for more targeted clinical interventions.
Methodology
Researchers analyzed 14 days of sleep data collected via the Oura Ring wearable device from 143 dementia caregivers. Using K-means clustering, the study modeled multiple parameters simultaneously, including total sleep time (TST), sleep efficiency (SE), wake after sleep onset (WASO), and sleep onset latency (SOL).
Findings
The analysis revealed three distinct subtypes: Optimal Sleep (53.1%), Disturbed Onset & Maintenance (29.4%), and Insufficient Sleep (17.5%). Notably, the Disturbed Onset cluster showed the highest prevalence of comorbidities like hypertension, while the Insufficient Sleep cluster was predominantly male and reported the lowest levels of support availability.
Impact
This work underscores the value of building a risk-profiling framework that integrates sleep subtypes with caregiver characteristics and social support data. Such precision health approaches can inform the development of more effective, tailored sleep interventions for those in high-stress caregiving roles.
This research was powered by the Centralive Platform.
Authors: Eunbee Angela Kim, PhD, RN; Jiuchen Zhang, PhD; Amir M. Rahmani, PhD; Sanghyuk Shin, PhD; Adeline Nyamathi, PhD, ANP, FAAN; and Jung-Ah Lee, PhD, RN, FGSA, FAAN, FADLN.
Read the full research article here: Nursing Research Journal
Join our network for deeper insights: https://newsletter.centralive.health/signup
