AI and Wearables for Early Detection of Cognitive Impairment and Dementia: Systematic Review (Preprint)
Bibliographic record
Abstract
BACKGROUND Traditional cognitive screening relies on episodic clinical assessments and may miss early changes preceding cognitive impairment and dementia. Wearable and mobile health technologies enable continuous monitoring of sleep, physical activity, and circadian rhythms, generating digital biomarkers that may support scalable early detection and prevention. However, current evidence remains fragmented across devices, analytic approaches, and cognitive outcomes. OBJECTIVE This study synthesizes and critically evaluates recent evidence on wearable devices for early detection and prevention of cognitive impairment and dementia, focusing on device categories, cognitive outcomes, analytic approaches, and prevention relevance. METHODS We searched PubMed, Scopus, ACM Digital Library, and SpringerLink for peer-reviewed studies published between January 2020 and December 1, 2025. Eligible studies included human participants with a mean age ≥50 years, continuous wearable-derived data collected for ≥24 hours, and validated cognitive outcomes; reviews, protocols, smartphone-only studies, and pharmacological interventions were excluded. Two reviewers independently screened studies, extracted data, and assessed risk of bias using the Appraisal Tool for Cross-Sectional Studies, Newcastle-Ottawa Scale, Cochrane Risk of Bias tool, and Quality Assessment of Diagnostic Accuracy Studies-2. Owing to substantial heterogeneity in devices, outcomes, and analytic methods, quantitative meta-analysis was not feasible; a structured narrative synthesis was conducted in accordance with PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) 2020 guidance. This study was not prospectively registered. RESULTS We included 49 studies, with sample sizes ranging from 14 to 91,948 participants (>200,000 total) and a median sample size of 145. Most used research-grade actigraphy (43/49, 87.8%), while fewer used commercial wearables (7/49, 14.3%). Cognitive outcomes most frequently relied on global screening instruments, including the Mini-Mental State Examination (18/49, 36.7%), followed by ICD-10 (International Statistical Classification of Diseases, Tenth Revision)–based clinical diagnoses (7/49, 14.3%) and the Montreal Cognitive Assessment (7/49, 14.3%). Analytic approaches were predominantly statistical (36/49, 73.5%), with fewer studies applying machine learning (7/49, 14.3%) or deep learning methods (6/49, 12.2%). Statistical analyses linked disrupted sleep, circadian rhythm fragmentation, and irregular activity patterns to worse cognitive outcomes, with modest-to-moderate effect sizes. Machine learning and deep learning approaches reported classification performance with area under the curve values between approximately 0.70 and 0.95. Approximately one-quarter of the studies (13/49, 26.5%) addressed early detection or prevention through longitudinal risk estimation or predictive modeling. Key limitations included small sample sizes, short monitoring durations, and limited external validation. CONCLUSIONS Wearable-derived behavioral markers show promise for early risk stratification. This review advances the field by shifting from descriptive associations toward a digital phenotyping framework evaluating artificial intelligence–driven prediction in the preclinical window. Unlike prior reviews focused on established dementia, it differentiates direct predictive evidence from indirect correlational findings and critically assesses methodological maturity. Continuous, passive monitoring may enable scalable detection of subtle behavioral changes, supporting earlier and more personalized risk reduction strategies. CLINICALTRIAL
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.011 | 0.047 |
| Meta-epidemiology (narrow) | 0.002 | 0.001 |
| Meta-epidemiology (broad) | 0.008 | 0.010 |
| Bibliometrics | 0.008 | 0.008 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.004 | 0.003 |
| Open science | 0.002 | 0.002 |
| Research integrity | 0.002 | 0.002 |
| Insufficient payload (model declined to judge) | 0.016 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".