BackgroundGrowing evidence suggests that dementia risk is influenced not only by genetic factors but also by social and environmental determinants. Understanding these modifiable factors is critical for informing prevention strategies.ObjectiveTo synthesize existing evidence from systematic reviews on the associations between social and environmental determinants and the risk of dementia, including Alzheimer's disease, vascular dementia, and frontotemporal dementia.MethodsAn umbrella review was conducted by systematically searching five major databases for systematic reviews published between 2004 and 2024. Eligible reviews examined the relationship between at least one social or environmental determinant and dementia outcomes.ResultsThe review found strong associations between environmental exposures and increased dementia risk. Exposure to fine particulate matter (PM) was consistently linked to elevated dementia risk, with estimates ranging from 3% to 226% per 10 μg/m increase. Occupational exposures to toxic metals, pesticides, and electromagnetic fields were also associated with higher neurodegeneration risk. Conversely, protective environmental factors included residential greenness and walkable neighborhoods. Among social determinants, higher education, socioeconomic status, and social engagement were found to promote cognitive resilience. In contrast, social disadvantage and limited access to healthcare contributed to increased dementia risk, likely through cumulative psychosocial stress.ConclusionsThis umbrella review underscores the significant role of social and environmental determinants in dementia risk. Targeted public health policies aimed at reducing environmental hazards and addressing social inequalities are essential for mitigating dementia risk and promoting cognitive health at the population level.

Alzheimer's disease (AD) is a growing problem in our society and the most common form of dementia. This neurodegenerative disease is characterized by neuroinflammation and the accumulation of amyloid-β (Aβ) and tau. Previous studies have found associations between the oral microbiome and AD. This review aims to elucidate the role of the oral microbiome in AD, through neuroinflammation, and reviews the relationship between AD and bacteria and fungi. Studies have found bacteria (e.g., ) and fungi (e.g., ) in postmortem AD brains. Moreover, mice models have shown that oral microbes are able to cross the blood-brain barrier (BBB), and were correlated with activated microglia, neuroinflammation, and Aβ load. This review introduces a mechanistic framework that describes how oral microbes cause an inflammatory response resulting in AD pathology. Specifically, oral dysbiosis causes oral pathogens to disseminate into the bloodstream, this triggers an inflammatory response, subsequently activating microglia, ultimately resulting in AD pathology. This process can follow two pathways: First, there is a direct response of the immune system in the brain to oral pathogens that migrate through the bloodstream and cross the BBB, which causes neuroinflammation and activates microglia, leading to AD pathology. Second, an early-life systemic inflammation causes microglia to get into a "hyperactive" state, in which they respond in an exaggerated way to normal stimuli triggering immune responses throughout a person's life that result in AD pathology. This mechanistic framework provides new line of thought for future research on the question of causality of AD.

BACKGROUND AND OBJECTIVES: Despite the availability of evidence-based antipsychotic reduction strategies, implementation of these strategies in long-term care (LTC) often stalls. Understanding how change occurs in real-world practice requires perspectives from interdisciplinary team members. This study interviewed interdisciplinary team members, a family caregiver, and a person living with dementia in LTC in Vancouver, Canada, to explore how they operationalize antipsychotic reduction in LTC homes and examine their perceived barriers and facilitators to implementation. RESEARCH DESIGN AND METHODS: Semi-structured interviews were conducted with 20 participants: 18 interdisciplinary healthcare providers, one dementia advocate, and one family caregiver. Data were analyzed using reflexive thematic analysis. To preserve inductive insights, the Consolidated Framework for Implementation Research (CFIR) was applied post hoc to the emergent themes. We examined the extent to which the findings were aligned with the CFIR, thereby deepening the analysis. RESULTS: Identified strategies included person-centred care planning, non-pharmacological interventions, medication review with behaviour monitoring, and education. Barriers were: (1) challenging work environments, (2) safety concerns and unconscious dismissive attitudes, and (3) communication gaps. Facilitators included: (1) supportive leadership and frontline champions, (2) team communication, and (3) persistence. Each factor aligns with different CFIR constructs and domains to varying degrees. DISCUSSION AND IMPLICATIONS: The main contribution of this study is that, drawing on CRIF, it found barriers to antipsychotic reduction in LTC are not merely individual but also systemic. Sustained improvement depends on policies that enable and resource effective interdisciplinary teamwork. This includes adequate staffing, education, team communication, and culture change.

BACKGROUND AND OBJECTIVES: Dementia's rising prevalence places an immense burden on caregivers. Knowledge Graphs (KGs) and Large Language Model (LLM)-augmented KGs are emerging AI approaches that organize complex dementia care knowledge and enable personalized, context-aware support, yet this field remains nascent. We aimed to map and synthesize research on KGs and LLM-augmented KGs in dementia caregiving, identifying system types, applications, outcomes, challenges, and ethical considerations. RESEARCH DESIGN AND METHODS: Following the JBI framework, a comprehensive search was conducted across six academic databases (PubMed, Scopus, Web of Science, IEEE Xplore, PsycINFO, CINAHL) and grey literature. Eligibility criteria included studies detailing the design, development, or evaluation of KGs or LLM-augmented KGs for dementia caregiving. RESULTS: Twelve articles representing 11 unique studies met the inclusion criteria. All 11 studies used KG or ontology components; eight were KG-only systems, often supporting personalized meal planning, care plan recommendations, knowledge management, robotic assistance, or virtual assistants. Three studies described LLM-augmented KGs (3/11), primarily using retrieval-augmented generation to enhance conversational AI for caregivers or persons with dementia. Reported benefits included improved usability, personalized support, more accurate or relevant recommendations, and potential improvements in quality of life and independence. Key challenges involved technical complexity, KG maintenance, data quality, limited real-world evaluation, and underdeveloped ethical analysis. DISCUSSION AND IMPLICATIONS: Integrating KGs with LLMs for dementia caregiving is a promising yet nascent interdisciplinary field. While early systems demonstrate potential, significant gaps remain in clinical validation, comprehensive ethical guidelines development, and responses to caregivers' diverse and evolving needs.

BACKGROUND: Multiple sclerosis (MS) is a chronic inflammatory disease of the central nervous system (CNS) characterized by relapses and progressive disability. The Expanded Disability Status Scale (EDSS), used to quantify disability, is based on single, discretely assessed and potentially inaccurate patient-estimated walking ability, whereas digital health technologies (DHTs) enable continuous activity monitoring and more objective assessment of real-world functional performance. METHODS: In this prospective observational study conducted at two German centers, patients with relapsing-remitting MS (RRMS) underwent clinical assessments at baseline (V1) and study completion (V2). Walking distance and step counts were measured using a measuring wheel and pedometer, while continuous physical activity was assessed via smartwatch-derived metrics. RESULTS: Sixteen patients with RRMS were included (median age 57.5 years [interquartile range (IQR) 49.25-63.25]; median EDSS 4.5 [IQR 3.5-6]). Patient-estimated walking distance at V1 showed moderate correlation with clinically measured distance (Spearman's ρ = 0.60, p = 0.013), with 12 of 16 patients misjudging distances relative to EDSS thresholds. Walking distance showed intra-individual variability between V1 and V2 (median absolute difference: 113.6 m). Median daily walking distance (ρ = -0.61, p = 0.0123), step count (ρ = -0.64, p = 0.0082), and peak steps (ρ = -0.69, p = 0.0032) correlated negatively with EDSS. CONCLUSION: Patient-estimated maximum walking distance demonstrated moderate agreement with clinical performance and frequently crossed EDSS thresholds, while clinical assessments varied substantially within individuals over the short study duration, underscoring the limitations of single evaluations. In contrast, smartwatch-derived metrics aligned with clinical measures, reflected EDSS scores, and captured real-world mobility.

BACKGROUND: Dementia affects over 50 million individuals globally, predominantly due to Alzheimer's disease (AD). Effective early detection and intervention remain clinical challenges, as there is a lack of unified, portable solutions to assess multiple biomarkers. METHODS: We evaluated Evoked Potential Tomography (EPT), an EEG-based method using a novel visual evoked potential protocol. An automated pipeline for EEG preprocessing, ERP extraction, feature selection, optimization, and regression modeling was developed to estimate key AD biomarkers: PET-amyloid standardized uptake value ratio (SUVR), CSF phosphorylated tau (p-tau181), Free and Cued Selective Reminding Test (FCSRT), and Mini-Mental State Examination (MMSE) scores. RESULTS: Regression models using ERP features from dementia participants demonstrated strong correlations (r = 0.8-0.94, p < 0.01) between predicted and true PET-amyloid SUVR, p-tau181, FCSRT, and MMSE values. In an independent external cohort, PET-amyloid SUVR predictions remained significantly associated with true values (r = 0.60, p < 0.01). DISCUSSION: Despite limitations, these preliminary results support EPT's potential as a sensitive and non-invasive method for estimating AD-related biomarkers in a clinically enriched AD cohort. Further validation studies are ongoing.

OBJECTIVE: To analyze the temporal trends of both Alzheimer's disease and Diabetes mellitus-related mortality in adults aged > 45years in the United States between 1999 and 2023, and to evaluate the changes in mortality patterns over time. BACKGROUND: Alzheimer's Disease and Diabetes Mellitus are two different diseases that have diverse underlying pathophysiology, but they often coexist, having common pathways. There is a high prevalence of concurrence between these two conditions, yet their combined mortality trend is underexplored. METHODS: We utilize mortality data from the CDC Wide-Ranging Online Data for Epidemiologic Research (WONDER). Individuals aged > 45 were included who had both Alzheimer's disease (G30) and Diabetes mellitus(E10-14). Age-adjusted mortality rates (AAMRs) and crude mortality rates (CMRs) per 100,000 were calculated and were standardized to the 2000 U.S population. Joint point regression models were used to identify the temporal variations and to calculate Annual Percentage Change (APC) and Average Annual Percentage Change (AAPC) with 95% confidence intervals. RESULTS: Overall, a total of 224,082 deaths occurred in patients of both Alzheimer's disease and Diabetes mellitus, in the age group ≥ 45 years, from 1999 to 2023. There is an upward trajectory noted from 2.82 in 1999 to 4.42 in 2023, with the highest incidence between 2017 and 2020, followed by a decline. Mortality rose in both sexes, with a persistently higher rate in females. The mortality rise from 1999 to 2023 in middle-aged people (45-64 years), and there was a rise in the trend of around 41% among adults ≥ 65 years. White individuals show higher deaths (78.9%), yet higher AAMR is observed in Black and Hispanic populations, showing racial disparities. Regionally, the West shows the highest AAMR, while non-metropolitan areas show higher mortality than metropolitan areas. CONCLUSION: The trend of mortality in individuals with both Alzheimer's disease and Diabetes Mellitus has increased in the past two decades, but there is a sharp rise observed after 2020 that may show the impact of the COVID-19 pandemic. These findings emphasized public health strategies.

White matter hyperintensities are ubiquitous in magnetic resonance imaging scans of older adults and reflect multiple pathological causes, particularly cardiovascular and metabolic disease risk factors. Elevated white matter hyperintensity burden is a predictor of dementia, but the relationship between white matter hyperintensity progression and cognitive change in typical aging remains unclear, arguably due to the scarcity of longitudinal evidence. Here, we estimated white matter hyperintensity volume in 67 participants of the Detroit Aging Brain Study, aged 50-77 at baseline, over four occasions spanning 7 years. The white matter hyperintensity volume increased over time, especially around the frontal horns of the lateral ventricles, and accelerated with advanced age. Greater baseline white matter hyperintensity volume predicted faster progression across all regions. Men had a greater white matter hyperintensity burden than women, independent of age, but did not differ in progression rate. Of all examined cognitive abilities, only perceptual speed slowing was associated with faster white matter hyperintensity progression in parietal and occipital lobes. Thus, not only does white matter hyperintensity volume increase over time in normative aging, but the change is linked to declines in a quintessential age-sensitive cognitive skill.

Isolated rapid eye movement sleep behavior disorder (iRBD) is a major prodromal marker of α-synucleinopathies, often preceding the clinical onset of Parkinson's disease, dementia with Lewy bodies, or multiple system atrophy. While wrist-worn actimeters hold significant potential for detecting RBD in large-scale screening efforts by capturing abnormal nocturnal movements, they require a reliable and efficient analysis pipeline. This study presents ActiTect, a fully automated, open-source machine learning tool to identify RBD from actigraphy recordings. To ensure generalizability across heterogeneous acquisition settings, our pipeline includes robust preprocessing and automated sleep-wake detection to harmonize multi-device data and extract physiologically interpretable motion features. Model development was conducted on a cohort of 78 individuals, yielding strong discrimination under nested cross-validation (AUROC = 0.95). Generalization was confirmed on a blinded local test set (n = 31, AUROC = 0.86) and two independent external cohorts (n = 113, AUROC = 0.84; n = 57, AUROC = 0.94). To assess robustness, leave-one-dataset-out cross-validation across cohorts demonstrated consistent performance (AUROC range = 0.84-0.89). Complementary stability analysis showed that predictive features remained reproducible across datasets, supporting the pooled multi-center pre-trained model for broader deployment. As an open-source, easy-to-use tool, ActiTect promotes adoption, independent validation, and collaborative improvements, thereby advancing generalizable wearable-based RBD detection.

We demonstrate how Large Language Models (LLMs) accelerate biomedical data harmonization through automated Common Data Element (CDE) generation. We processed 31 datasets including clinical taxonomies and research data dictionaries through OpenAI's Generative Pre-trained Transformer - 4 (API Model gpt-4-0613), generating comprehensive metadata for each element using a template-based system. Subject-matter experts validated outputs, finding 94% of generated metadata fields required no revision overall, with an unweighted accuracy of 83.8%, unweighted, for semi-structured sources. Dramatically faster than manual approaches. Our system uses ElasticSearch with weighted field matching to identify semantic equivalences between variables, avoiding duplicate CDEs while building a standardized repository. Testing with Alzheimer's Disease Neuroimaging Initiative (ADNI) and Global Parkinson's Genetic Program (GP2) datasets showed 32.4% of previously unseen headers successfully mapped to our CDEs, with interoperability scores averaging 53.8/100 based on matching, completeness, and compliance metrics. This approach automates the most tedious aspects of data integration, reducing barriers to cross-study collaboration in biomedical research.

Brain disorders characterized by progressive neurodegeneration, such as Alzheimer's disease (AD) and frontotemporal dementia (FTD), represent an increasing medical and societal challenge. While genome‑wide studies have uncovered numerous susceptibility loci, these efforts have largely focused on common variants and leave a substantial portion of genetic liability unresolved. Variants of low frequency, often associated with stronger biological effects, remain insufficiently characterized, particularly in heterogeneous populations. Genetically isolated populations offer an effective strategy to overcome these limitations. Finland, shaped by historical demographic events, harbors a distinctive spectrum of enriched rare variants that can facilitate gene discovery. The FinnGen initiative capitalizes on this setting by combining extensive genotyping with nationwide health registry data through a coordinated network of Finnish biobanks. With half a million participants analyzed, FinnGen supports highly powered analyses across a broad array of clinical outcomes and registry data. Recent comprehensive analyses have reported thousands of significant genotype-phenotype associations, including novel protein‑altering variants. Importantly, the FinnGen cohort structure favors older individuals and hospital‑derived samples, increasing representation of brain disorders, such as AD and idiopathic normal pressure hydrocephalus (iNPH), a disorder frequently accompanied by AD‑like pathological features. In this expert review, we summarize FinnGen‑based investigations relevant to neurodegenerative diseases and iNPH, highlighting insights into genetic susceptibility, disease overlap, and protective factors, and discuss how integration with recall studies as well as biomarker and clinical data accelerates translational applications in brain disorders.

Brain age prediction has gained significant attention due to its strong correlation with neurological and cognitive disorders. The discrepancy between an individual's chronological age and their predicted brain age-known as the Brain Age Gap-has been linked to conditions such as schizophrenia, Alzheimer's disease, cognitive decline, and lifestyle factors like stress and poor health. A positive Brain Age Gap is often associated with accelerated aging and neurodegeneration, highlighting the need for precise and reliable estimation methods. In this study, we propose a novel deep learning model that incorporates time-distributed, convolutional and bidirectional LSTM layers for brain age estimation. Using MRI data from the OpenBHB dataset, processed through Voxel-Based Morphometry (VBM), our model undergoes rigorous preprocessing, including outlier detection, data augmentation, and MRI slice selection, to enhance learning efficiency. The model is optimized with the Adam optimizer with a scheduled learning-rate decay and evaluated using Mean Absolute Error (MAE) and [Formula: see text] Score. Experimental results demonstrate that our model achieves an MAE of 3.1573 years, outperforming previous methods and improving brain age prediction accuracy. These findings underscore the importance of advances in deep learning and data preprocessing in enhancing brain age estimation.

Neuroinflammation is a hallmark of Parkinson's disease (PD), a progressive neurodegenerative disorder characterized by the accumulation of α-synuclein and the death of dopaminergic neurons in the substantia nigra. Mutations in GBA are a common risk factor for PD, which can lead to lipid metabolism dysfunction, autophagy/lysosomal dysregulation, as well as the disruption of other cellular functions. In this study, we investigated the impact of the GBA-N370S mutation and astrocytic reactivity on α-synuclein pathology and neurotoxicity. To investigate the impact of reactive astrocytes on Parkinson's disease pathology, we employed iPSC-derived midbrain astrocyte and dopaminergic neuron co-cultures from control and GBA-N370S donors, as well as primary mouse midbrain astrocyte cultures and transcriptomic assays to examine the response of astrocytes to Tumor Necrosis Factor-α (TNFα) and Interferon-γ (IFNγ). We show that upon inflammatory stimuli astrocytes become reactive, leading to extensive transcriptional changes. RNAseq and experimental validation revealed that calcium transport and homeostasis were severely dysregulated, and functional studies confirmed that GBA-N370S astrocytes exhibited increased calcium release when treated with cytokines. We further explored the impact of inflammation on astrocytic neurosupport in an iPSC-derived dopaminergic neuron and astrocyte co-culture model finding that combined treatment of TNFα, IFNγ and α-synuclein pre-formed fibrils (PFFs) led to neurotoxic effects, suggesting that TNFα and IFNγ-activated astrocytes mediate α-synuclein PFF toxicity. Taken together, these data provide evidence of reduced neurosupport in both control and GBA-N370S iPSC-derived midbrain astrocytes exposed to inflammatory cytokines, suggesting a role for reactive astrocytes in PD pathology.

PURPOSE: Nurse leaders at our site identified a gap in evidence-based practice related to dementia-specific pain assessment. The purpose of this quality improvement project was to improve pain detection and management in patients with advanced dementia by implementing the observational Pain Assessment in Advanced Dementia (PAINAD) scale in the clinical electronic medical record. DESIGN: Our project team used a pre- poststudy design to evaluate practice changes in implementing the PAINAD scale. Nurses were notified of the change to documentation through hospital-wide newsletters, daily huddles, and unit managers. METHODS: Data was retroactively collected through chart review of patients with advanced dementia for the four weeks prior to implementation, and at 4- and 6-weeks after the implementation. Data collected included documentation of an observational pain assessment, treatment (pharmacologic or nonpharmacologic), whether pain was present, use of psychotropic medications, and documented improvement in pain at reassessment. We also surveyed nurses to understand their perceptions of the PAINAD. RESULTS: Improvements were observed in rates of observational pain assessment, analgesic or nonpharmacologic treatment of pain, and avoidance of psychotropic medications at 6-weeks post-implementation. However, no improvement was observed in the proportion of patients with reduced pain at reassessment after treatment. No changes were statistically significant. Overall, nurses were comfortable using the PAINAD and perceived it as useful. CONCLUSION: Implementing the PAINAD scale appeared to improve observational pain assessments and pain treatment at our facility. CLINICAL IMPLICATIONS: Implementation of the PAINAD scale across all clinical care units and expanded staff education are necessary to improve pain outcomes for patients with advanced dementia.

Neurodegenerative diseases share common pathophysiological mechanisms, including chronic neuroinflammation, glutamatergic excitotoxicity, oxidative stress, mitochondrial dysfunction, and disruptions in synaptic and lipid homeostasis. In this context, the endocannabinoid system has emerged as a key modulator of neuroimmune communication and neuronal survival. Within this system, Monoacylglycerol Lipase (MAGL) plays a central role by regulating the levels of the endocannabinoid 2-Arachidonoylglycerol (2-AG) while simultaneously contributing to the generation of arachidonic acid and pro-inflammatory eicosanoids. Pharmacological or genetic inhibition of MAGL increases 2-AG levels and concurrently reduces the biosynthesis of pro-inflammatory lipid mediators, thereby modulating microglial activation, astrocytic responses, and neuronal excitotoxicity. Preclinical studies in models of Alzheimer's disease, Parkinson's disease, multiple sclerosis, and amyotrophic lateral sclerosis consistently demonstrate that MAGL blockade attenuates neuroinflammation, preserves synaptic and neuronal integrity, improves motor and cognitive function, and, in some cases, delays disease progression. Although clinical evidence remains limited, the available data position MAGL as a metabolic convergence point between inflammation and neurodegeneration, suggesting that its modulation may represent a therapeutic strategy with disease-modifying potential.

Ischemic stroke induces prolonged T cell accumulation within injured brain tissue, yet it remains unclear whether these cells reflect nonspecific inflammatory persistence or organized adaptive immune responses. To define the clonal architecture of post-stroke T cells, we performed genomic DNA-based bulk T cell receptor (TCR) immunosequencing of CDR3α and CDR3β repertoires from infarcted brain and spleen during the chronic phase of experimental stroke across age and sex. TCRβ repertoires were further examined across three ischemic stroke models reproduced independently at sites in the United States and Europe. Chronic infarct tissue consistently exhibited oligoclonal T cell expansion across age, sex, stroke models, and laboratories; spleen and blood remained broadly polyclonal. Dominant clonotypes occupied a substantial fraction of the infarct repertoire, revealing a structured clonal architecture within the injured brain. Computational annotation identified recurrent sequence similarities to self-associated TCRs, including receptors linked to myelin, nuclear, and insulin-related antigens, although many expanded clonotypes lacked database matches. These annotations are presented as hypothesis-generating rather than evidence of antigen specificity. Together, these findings demonstrate that chronic ischemic brain injury is associated with a reproducible, infarct-associated clonal T cell signature whose conserved architecture is consistent with antigen-driven selection, although stochastic or cytokine-driven expansion cannot be excluded. The accompanying publicly available TCR repertoire dataset provides a clonotype-resolved reference resource for future investigations of antigen specificity and adaptive immune dynamics in chronic post-stroke neuroinflammation.

BACKGROUND: Accumulation of brain amyloid beta (Aβ), a key pathological hallmark of Alzheimer's disease (AD), begins decades before cognitive symptoms. Being able to predict the risk of Aβ accumulation, or the age at which Aβ exceeds a critical threshold, may enable intervention to delay or prevent onset of AD. METHODS: Using published genome-wide association studies (GWASs), we developed polygenic scores (PGS) for AD risk (PGS) and resilience (PGS), and tested whether these predicted (i) if an individual is an Aβ accumulator ('Accumulator Status'), and (ii) in accumulators, the age at which brain Aβ exceeds a 20 centiloid (CL) threshold ('Age at onset of Aβ'; AAO-Aβ) in 2175 participants (1158 with AAO-Aβ) from the Alzheimer's Dementia Onset and Progression in International Cohorts (ADOPIC) study. We also performed GWASs on these traits to develop phenotype-specific PGSs. FINDINGS: Higher genetic risk of AD predicted increased odds of Aβ accumulation (OR = 1.16; 95% CI = 1.05-1.29; p = 0.003) and younger AAO-Aβ (β = -1.32; SE = 0.31; p = 1.63 × 10). Higher genetic resilience to AD predicted later AAO-Aβ (β = 0.91; SE = 0.29; p = 0.002) but did not predict Aβ accumulation. These associations were independent of APOE ε4 status, the strongest genetic risk factor for AD. Phenotype-specific PGSs were not significantly associated with either trait. INTERPRETATION: Polygenic scores, alongside other risk factors, may help identify individuals at risk of accumulating Aβ, and predict the age at which this exceeds a critical threshold. This could provide a window for administering disease-modifying treatment or lifestyle interventions to prevent or delay the onset of AD. FUNDING: National Institutes of Health (R01-AG058676-01A1) and Australian National Health and Medical Research Council (GNT1161706; GNT2001320).

BACKGROUND: Delirium superimposed on dementia is associated with poor outcomes yet remains underdetected in home settings. Current detection relies on face-to-face clinical assessment (eg, the Confusion Assessment Method criteria), which is rarely applied outside hospitals. OBJECTIVE: This proof-of-concept study developed a theory-driven framework for detecting delirium-consistent anomalous patterns in home-dwelling people with dementia, using passive smart home sensor data. METHODS: The Technology Integrated Health Management dataset, an open access resource comprising a clinically derived cohort of older adults (aged 50 years) with a confirmed diagnosis of dementia or mild cognitive impairment, was used. The analysis included 13 patients who had at least 50% valid data for at least one 10-day analysis window, with data collected between April 1, 2019, and June 30, 2019. Individualized anomaly detection algorithms, including Isolation Forest and Long Short-Term Memory models, were applied to identify delirium-related anomalies within each participant. Predictor features consisted of theory-driven digital markers approximating key Confusion Assessment Method criteria, including agitation, disrupted sleep-wake cycles, and disorientation (indexed by activity entropy), along with clinically relevant indicators, such as physiological instability (early warning scores) and urinary tract infections. RESULTS: Using matched thresholds, the Isolation Forest identified 77 anomalies (anomaly rate: 15.65%), and the Long Short-Term Memory model identified 78 anomalies (anomaly rate: 15.85%), with anomalies typically occurring in short temporal clusters; agreement between methods ranged from 0% to 40% across individuals. Feature importance analyses indicated that activity entropy, sleep quality, and early warning scores were the most influential features, with stronger interfeature correlations observed during anomaly periods than during nonanomaly periods. CONCLUSIONS: This study demonstrates the technical feasibility of detecting delirium-related anomalies through passive smart home monitoring. While lacking ground truth validation, the approach shows promise for early intervention in community settings. Future validation studies with clinically confirmed delirium labels are essential.