Nearly 20% of adults over 50 are estimated to be living with mild cognitive impairment. Yet more than half of individuals with dementia never received a formal cognitive evaluation during the early or mild stages of their disease.
For primary care providers, more effective cognitive screening means identifying at-risk patients earlier and more efficiently, in a way that fits into a routine care visit. For specialists, it means receiving referrals that are more likely to reflect genuine cognitive impairment, supporting more efficient triage, as well as access to more sensitive tools for assessing and tracking cognitive decline over time.
A new paper by researchers at Western University, published in the Journal of Alzheimer's Disease, takes a two-part approach to analyzing the problem of cognitive screening. First, a rapid review examined the current landscape of cognitive screening tools and their limitations. The paper then presented a study on the development and validation of the two-task Creyos digital cognitive screener, demonstrating that it can detect cognitive impairment linked to Alzheimer's disease with 100% sensitivity and 86% specificity in approximately six minutes, without the need for clinical supervision.
The team reviewed 258 full-text articles, categorizing cognitive assessment tools into four groups: paper-and-pencil performance-based screeners, digital measures, informant-reported measures, and full cognitive batteries.
Of the 258 articles reviewed, the Mini-Mental State Examination (MMSE) appeared in 164 and the Montreal Cognitive Assessment (MoCA) in 93. This concentration has reinforced their status as default reference points despite limited evidence that either is superior to other available screeners.
The review identified several limitations of the MMSE and MoCA that are directly relevant to clinical practice:
Beyond the limitations identified in the rapid review, separate research has shown that both the MoCA and the MMSE can show improved scores on repeated administration due to familiarity with the test format rather than genuine cognitive change. This reduces their suitability for tracking cognitive decline over time.
At the 2025 Alzheimer's Association International Conference (AAIC), emerging clinical perspectives signaled growing limitations of traditional tools in primary care, alongside increased support for digital cognitive assessments as a scalable approach to cognitive detection and triage.
The rapid review confirmed that digital platforms offer advantages over paper-and-pencil approaches: richer performance data (including response times, attempt counts, and error types), adaptive difficulty, and automated scoring that reduces bias in interpretation. Notably, older adults with limited technological experience have been shown to use unsupervised digital devices proficiently, producing data comparable to in-clinic assessments.
Despite these advantages, some digital screeners lack established normative databases. The review found that most currently available tools, including digital screeners, rely on small normative datasets composed primarily of adults over 65, or group all individuals above a certain age into a single category. Normative data derived solely from older adults cannot be reliably applied to middle-aged individuals, and without representative norms, cognitive impairment may be over- or under-identified.
Following the review of the current landscape of cognitive screening tools, the paper next described the development and validation of a two-task digital cognitive screener included in the Creyos cognitive assessment platform.
A machine learning model was trained on data from 4,715 individuals: 2,859 healthy adults from the Creyos normative database and 1,856 patients from neurology clinics specializing in dementia and age-related cognitive decline. The model evaluated all possible combinations of cognitive tasks and performance features to determine which pairing best differentiated individuals with cognitive impairment from those without.
Two tasks were identified as the most effective for distinguishing cognitively impaired from unimpaired individuals: Number Ladder, which measures visuospatial working memory, and Feature Match, which measures attention, along with a reaction-time feature from each task.
Both cognitive domains are well-established in the literature as early indicators of decline along the Alzheimer's disease continuum and have been consistently associated with mild cognitive impairment and dementia.
To evaluate the screener's performance, the model was applied to a sample of 9,350 adults aged 21 to 97, drawn from two published datasets of the general population, as well as a smaller cohort of patients independently diagnosed with Alzheimer's disease.
The proportion of individuals positive on the screener—that is, flagged as potentially cognitively impaired—increased with age, with positive results rising significantly after age 50 in a pattern consistent with the nearly 20% prevalence of mild cognitive impairment.
| Age Group (Years) | Proportion Flagged as At-Risk for Cognitive Impairment |
|---|---|
| 20–30 | 2.6% |
| 30–40 | 4.3% |
| 40–50 | 6.7% |
| 50–60 | 12.3% |
| 60–70 | 20.5% |
| 70+ | 38.5% |
Gender, education level, and study source had no significant effect on outcomes, meaning the screener performed consistently across demographic groups. For clinicians, this means a single tool can be applied across a varied patient panel.
To test sensitivity to Alzheimer’s-related cognitive impairment, the screener was applied to a small clinical cohort: 14 patients independently diagnosed with Alzheimer's disease by a qualified neurologist. All 14 were positive on the screener, for a sensitivity of 100%. In a case-matched control group of 14 individuals matched for age, gender, and education, only two were flagged, yielding a specificity of 86%.
One patient in the Alzheimer's disease cohort scored above the standard MMSE cut-off and would not have been flagged by that tool alone. The Creyos screener correctly identified this patient as cognitively impaired. In a clinical context, that gap is the difference between a patient being referred for further evaluation and being told their cognition is within normal limits.
A remaining question was whether other health-related disorders associated with cognitive deficits would also trigger positive screener outcomes. To investigate, individuals with self-reported hypertension, concussion history, cardiac problems, depression, and anxiety, among other conditions, were compared to matched controls on measures of cognitive performance and screener outcome.
Several of these groups showed lower cognitive performance in certain domains. However, none were associated with an increased rate of positive screener results. The Alzheimer's disease group, by contrast, was both cognitively impaired and significantly more likely to be flagged by the two-task screener.
These findings suggest the screener is identifying patterns of cognitive impairment associated with dementia rather than generalized difficulties from other conditions. For clinicians weighing referral decisions, that distinction means more confidence that a positive result reflects something worth investigating, not a false signal from an unrelated condition.
The screener is not intended to replace comprehensive neuropsychological evaluation, but rather to serve as an accessible first-line tool that informs subsequent testing and intervention.
For clinicians, the practical result is a screener that can be completed in approximately six minutes. A care team member initiates the assessment, and the patient completes it independently, with no clinical oversight required during administration. Scoring is fully automated, producing a clear binary result that flags patients who may benefit from further evaluation, without interpretation variability or the need for specialized training.
When a positive result indicates the need for further testing, the broader Creyos cognitive assessment platform offers a 12-task battery. Unlike traditional cognitive tests that rely on fixed cut-off scores, the platform assesses cognitive function on a continuum and compares patient performance against a normative database of approximately 85,000 healthy individuals, including adults under 65. The platform also uses adaptive difficulty that adjusts to each patient's performance, minimizing ceiling effects. Together, these features support more detailed characterization of cognitive performance and the ability to monitor changes over time.
A large-scale longitudinal study is now underway to evaluate the screener's long-term predictive value. The study is tracking cognitive trajectories in cognitively normal adults, individuals with subjective cognitive decline, mild cognitive impairment, and dementia over 12 months, assessing sensitivity across the cognitive continuum and performance in relation to daily functioning and overall health. This study will also examine how the screener works alongside a longer assessment to provide both screening and follow-up neuropsychological testing in a single clinical workflow. The Creyos platform's non-repeating problem sets and adaptive difficulty are designed to support screening, assessment, and longitudinal monitoring without the practice effects that limit traditional repeat assessments.
The validation of the screener is available in the Journal of Alzheimer's Disease. Clinicians interested in how this screener fits into a screening or monitoring workflow can learn more about the Creyos platform.
Reviewed by Mike Battista, Director of Science & Research at Creyos
Mike Battista specializes in brain health, cognition, and neuropsychological testing. He received his PhD in personality and measurement psychology at Western University in 2010 and has been doing fun and useful stuff in the intersection between science and technology ever since.