JMIR Publications

ABSTRACT

Background:

Traditional cognitive screening methods are episodic and may overlook subtle early changes. Wearable and mobile health (mHealth) technologies enable continuous monitoring of activity, sleep, and circadian rhythms, generating digital biomarkers that, combined with artificial intelligence (AI), may support scalable early detection and preventive strategies for cognitive decline. Evidence remains fragmented across devices, populations, and analytic approaches.

Objective:

This review synthesizes recent evidence on wearable devices for early detection and prevention of cognitive impairment and dementia, focusing on device types, cognitive assessments, analytic methods, and prevention relevance.

Methods:

PubMed, Scopus, ACM Digital Library, and SpringerLink were searched for studies published between January 2020 and April 2025. Eligible studies used wearable or mobile devices to collect behavioral or physiological data linked to cognitive decline. Two co-authors independently screened and extracted study characteristics, devices, monitored variables, cognitive measures, analytic techniques, and prevention-related outcomes.

Results:

A total of 40 studies were included. Most (35/40, 87.5%) used research-grade actigraphy, 6/40 (15.0%) employed commercial wearables, 3/40 (7.5%) used ad-hoc prototypes, and 2/40 (5.0%) multimodal research devices. Cognitive assessments most often relied on the Mini-Mental State Examination (17/40, 42.5%), followed by the Clinical Dementia Rating (4/40, 10.0%) and the Montreal Cognitive Assessment (4/40, 10.0%). Analytic approaches were mainly statistical (29/40, 72.5%), with fewer applying machine learning (6/40, 15.0%) or deep learning (5/40, 12.5%). One half (20/40, 50.0%) directly addressed early detection or prevention through longitudinal risk quantification, predictive modeling, or by linking wearable-derived behaviors to modifiable factors.

Conclusions:

Wearable and mobile devices hold promise for contributing to the early detection of cognitive decline and for informing preventive strategies. Current evidence, however, is preliminary and limited by small samples, short monitoring periods, heterogeneous methods, and scarce external validation. While early findings highlight potential behavioral markers and predictive models, these should not yet be considered clinically ready. Future work should emphasize larger prospective studies, integration of consumer and research-grade devices, and the use of explainable AI within standardized protocols to advance toward scalable tools for dementia risk detection and prevention.