Currently submitted to: JMIR Aging
Date Submitted: Nov 13, 2026
Open Peer Review Period: Nov 15, 2026 - Jan 10, 2027
(currently open for review)
Warning: This is an author submission that is not peer-reviewed or edited. Preprints - unless they show as "accepted" - should not be relied on to guide clinical practice or health-related behavior and should not be reported in news media as established information.
Passive Multimodal Sensing of Delirium-Related Changes in Community Dwelling Older Adults With Dementia: Exploratory Feasibility Study
ABSTRACT
Background:
Delirium is common, morbid, and costly, with particularly severe consequences for older adults with dementia, and early detection may facilitate its reversal and improve outcomes. Passive home monitoring may detect behavioral and physiological changes that precede delirium-related clinical events. However, feasibility and clinical utility of such monitoring remain largely untested in community-dwelling adults with dementia.
Objective:
This exploratory feasibility study evaluated multimodal passive sensing in this population. We examined device data coverage, physiological and behavioral changes around clinical events, and supervised and unsupervised detection of delirium-related events.
Methods:
Older adults with dementia enrolled in the Advanced Leveraging of Electronic Remote Technology (ALERT) pilot study through a Program of All-Inclusive Care for the Elderly (PACE) organization in Massachusetts, USA, were monitored at home for up to 12 months with a Withings under-mattress sleep sensor and a Withings Pulse HR wrist-worn sensor. Caregivers monitored mental status with weekly administration of the Ultra-Brief 2-Item Screener (UB-2). Clinical events that could be related to delirium (e.g., infections, falls, emergency department visits) were abstracted from the electronic medical record. Clinical events were defined as a caregiver-reported mental status change or the person not seeming like themselves, failure of both the attention and disorganized-thinking UB-2 items, or an adjudicated medical-record event. Linear mixed-effects models compared passively sensed features between 1- to 7-day pre-event and post-event windows. Logistic regression, Support Vector Machine, and Naïve Bayes classifiers were evaluated with leave-one-participant-out (LOPO) cross-validation using the area under the receiver operating characteristic curve (AUROC), balanced accuracy (BA), and positive predictive value (PPV). Local Outlier Factor, Isolation Forest, and k-nearest-neighbor anomaly detection were evaluated within participants and with LOPO against clinical event days.
Results:
Among the seven participants, the sleep-mat provided greater mean daily data coverage than the wrist sensor (10.3 vs 2.4 hours/day). Average respiration rate differed significantly between pre- and post-event windows of 1 to 7 days in several comparisons (P<.05). At 2- to 6-day pre-event windows, Naïve Bayes achieved an AUROC of 0.618 to 0.665, BA of 0.594 to 0.634, and PPV of 0.37 to 0.48 across alert thresholds (base rate approximately 0.29). Within-person Local Outlier Factor anomaly detection performed better for some participants with last observation carried forward or linear interpolation imputation, but inconsistently across the cohort. In the LOPO evaluation, 60.0% to 83.3% of event days fell in participant-relative medium or high-anomaly bands.
Conclusions:
Extended passive home monitoring was feasible, particularly with the under-mattress sensor. Respiratory patterns emerged as potentially informative markers around clinical events. Supervised and unsupervised findings indicate detectable signals but do not establish clinical predictive validity. Larger studies are needed to determine whether passive sensing can support timely delirium detection, along with strategies to reduce adherence barriers for wrist-worn sensors.
Citation
Request queued. Please wait while the file is being generated. It may take some time.
Copyright
© The authors. All rights reserved. This is a privileged document currently under peer-review/community review (or an accepted/rejected manuscript). Authors have provided JMIR Publications with an exclusive license to publish this preprint on it's website for review and ahead-of-print citation purposes only. While the final peer-reviewed paper may be licensed under a cc-by license on publication, at this stage authors and publisher expressively prohibit redistribution of this draft paper other than for review purposes.