Abstract
Background
Mild cognitive impairment (MCI) represents an intermediate stage between normal cognition and dementia, offering a critical window for early detection and intervention. Although a wide range of digital cognitive assessments have been developed, few have been adapted or validated for use in Vietnamese populations
Methods
The study was conducted in three phases: development of the TNmindtest based on episodic memory and attention cognitive domain, test-retest reliability assessment among 35 participants aged 30–50 years with subjective memory complaints, and validity evaluation in a separate sample of 80 participants, classified as MCI or non-MCI based on Montreal Cognitive Assessment (MoCA) scores.
Results
The TNmindtest demonstrated good test–retest reliability (ICC = 0.76, 95% CI: 0.59–0.88) over a mean interval of 8.9 days, with a modest learning effect of approximately 7%. Discriminative validity was supported by significantly lower TNmindtest scores in the MCI group (74.0 [60.0–80.0]) compared to the non-MCI group (80.0 [74.0–87.0], p = 0.003). ROC analysis yielded an area under the curve (AUC) of 0.70, with a sensitivity of 62.9% and a specificity of 66.7% at the optimal cut-off of 75%. TNmindtest scores were modestly correlated with MoCA total scores (r = 0.31, p = 0.0051), and showed significant associations with delayed recall and the combined memory-plus-attention subtotal.
Conclusions
The TNmindtest demonstrates acceptable performance as a culturally adapted, self-administered digital tool for MCI screening among Vietnamese adults. Its brief testing time, automated scoring, and potential for remote use support its feasibility for broader application, although further refinement and validation are warranted.
Introduction
Mild cognitive impairment (MCI) is an intermediate stage between normal aging and dementia, with a subtle and variable clinical presentation. People with MCI exhibit deficits in one or more cognitive domains, most commonly memory, attention, or executive function, while still maintaining their independence in daily activities. 1 Two main subtypes are recognized: amnestic MCI (aMCI), which primarily affects episodic memory and carries a higher risk of progression to Alzheimer’s disease, and non-amnestic MCI (naMCI), which involves other domains such as language or visuospatial processing.1,2
Globally, the burden of mild cognitive impairment has increased substantially over the past two decades and is now recognized as a major public health concern. Meta-analytic evidence suggests that approximately 10-20% of adults aged 65 years and older meet diagnostic criteria for MCI in community-based populations. 3 Importantly, growing evidence indicates that MCI is not restricted to older adults. Population-based studies have reported that measurable cognitive impairment can already be detected in midlife, with prevalence estimates ranging from approximately 6-10% among individuals aged 50-59 years and increasing steadily with age. 4 Longitudinal cohort studies further suggest that cognitive decline may begin 10-15 years before the clinical diagnosis of dementia, highlighting the importance of identifying early cognitive changes during midlife.5,6 Several modifiable midlife risk factors, including hypertension, diabetes, obesity, cardiovascular disease, and chronic psychosocial stress, have been consistently associated with an increased risk of cognitive impairment before the age of 65, particularly in rapidly urbanizing populations.7,8
Despite this growing global burden, epidemiological data on cognitive impairment remain limited in many low- and middle-income countries (LMICs), where rapid demographic aging is occurring alongside increasing cardiometabolic risk factors.7,9 Vietnam is experiencing one of the fastest aging populations in Southeast Asia, yet population-based evidence on cognitive health remains scarce. 10 Available community-based studies suggest that cognitive concerns may already be common among middle-aged and older adults. For example, a study conducted in Da Nang among 600 adults aged 55 years and older reported that 64% of participants perceived their memory as poor and 39% indicated that memory problems interfered with daily functioning. 11 These findings highlight the potential burden of cognitive problems in Vietnamese communities and underscore the need for accessible and scalable tools to facilitate early detection of cognitive impairment in community settings.
Traditionally, cognitive screening has relied on paper-based instruments such as the Mini-Mental State Examination (MMSE) and the Montreal Cognitive Assessment (MoCA), which remain widely used for global cognitive assessment. 6 Although electronic versions of these tools have been introduced, simple digitization does not fully resolve implementation barriers. For instance, the MoCA still requires standardized administration, professional supervision, and dedicated time for completion, and formal training or certification is recommended for most health professionals administering the test. 12 Therefore, even in digital form, these instruments remain dependent on healthcare personnel and may be difficult to scale for large-scale community screening.
Recent advances in neurotechnology and digital health have enabled the development of online cognitive assessment platforms for rapid and remote testing, such as The Brain on Track, 13 CompBased-CAT, 14 NIH Toolbox 15 and BrainCheck. 16 These platforms have demonstrated promising psychometric properties and offer advantages including automated scoring, cloud-based data storage, and remote accessibility. However, most of these systems were developed and normed in high-income, Western countries and are primarily designed for English-speaking populations. In addition, many commercial platforms require subscription fees or licensing costs, which may limit accessibility in low- and middle-income settings and hinder widespread community deployment. 17 Beyond economic considerations, performance on computerized cognitive tests may be influenced by educational attainment, cultural background, and digital literacy. 18 Variability in familiarity with electronic devices, internet access, and user-nterface interaction may introduce construct-irrelevant variance, particularly among older adults and individuals with limited technological exposure. 19 Moreover, cognitive constructs and normative performance distributions are shaped by sociocultural context. 20 Cut-off scores derived from Western populations may not be directly transferable to other settings without locally established normative data, raising concerns about diagnostic misclassification.
Collectively, these limitations suggest that existing digital platforms, while valuable in their original contexts, may not be readily applicable for large-scale community screening in diverse cultural and resource-limited environments. This underscores the need for culturally grounded, context-specific digital cognitive assessment tools that are accessible, scalable, and supported by locally derived normative standards.
Collectively, these limitations suggest that existing digital cognitive assessment platforms, while valuable in their original contexts, may not be readily applicable for large-scale community screening in diverse cultural and resource-limited settings. This highlights the need for culturally appropriate, accessible, and scalable digital screening tools supported by locally derived normative data. To address this gap, we developed the Time Now mind test (TNmindtest
Materials and methods
Study design
This study was conducted from December 2024 to August 2025 in three sequential phases: development, reliability testing, and validity assessment of the TNmindtest, a web-based tool for early detection of memory impairment in adults. All participants provided written informed consent prior to enrollment.
Phase 1 – Tool development (december 2024 – february 2025)
The development of the TNmindtest followed a structured three-step framework, including identification of cognitive domains relevant to mild cognitive impairment, construction of the screening tool, and expert review with iterative refinement.
Step 1: Identification of cognitive domains relevant to MCI
Mild cognitive impairment involves measurable deficits across several cognitive domains, including memory, attention, executive function, language, and visuospatial abilities. 21 Among these domains, episodic memory dysfunction is considered one of the earliest and most sensitive indicators of early cognitive decline, particularly in individuals at risk of developing Alzheimer’s disease. 21 Episodic memory refers to the ability to encode, retain, and retrieve newly learned information, and tasks assessing encoding and recognition processes are widely used in cognitive screening. 22
In addition to episodic memory deficits, impairments in working memory and executive processes have also been observed in individuals with MCI.23,24 Working memory supports the short-term storage and manipulation of information required for complex cognitive activities such as planning, reasoning, and problem solving. 23 Therefore, the conceptual design of the TNmindtest incorporated tasks targeting both episodic memory performance and working-memory-related processes.
Step 2: Construction of the TNmindtest
Based on these conceptual considerations, the TNmindtest was developed as a web-based, self-administered cognitive screening platform designed for remote use in community settings. The system was hosted on a secure cloud-based server with a dedicated administrative portal for data storage and study management. The platform supports automated scoring and real-time data recording, allowing participant responses to be securely stored and monitored through an integrated database.
Participants accessed the test through the public interface (https://tnmindtest.com) and registered using a unique mobile phone number, which served as a secure identifier to prevent duplicate entries and protect personal data confidentiality. The platform was accessible through standard web browsers without requiring software installation and was optimized for both desktop and mobile devices, including smartphones and tablets.
The TNmindtest follows a three-stage paradigm-encoding, interference, and recognition designed to simulate real-world memory demands by combining passive viewing, distraction, and active retrieval under controlled timing conditions.
Encoding phase
Participants are sequentially presented with 15 images from 15 distinct semantic categories (e.g., animals, tools, food). For each category, a pool of visually similar images is prepared, and one image is randomly selected at each test administration to reduce memorization effects across repeated testing while maintaining comparable task difficulty across sessions. Each image is displayed for 3 seconds, and participants are instructed to observe and memorize the images.
Interference phase
Immediately after the encoding stage, participants are exposed to a 20-second video clip unrelated to the memorized images. To ensure active engagement, participants are required to answer a multiple-choice question regarding a specific detail in the video at the end of the clip.
Recognition phase
A total of 45 images are presented sequentially, including 15 original target images and 30 foil images. Each foil image closely matches the semantic category and visual complexity of its corresponding target. Each image appears for 3 seconds, during which participants must indicate whether it was presented during the encoding phase.
The primary TNmindtest score was defined as the percentage of correctly recognized target images (range: 0–100%). To reduce unnecessary testing time, the recognition module incorporated a predefined stopping rule: once a participant correctly identified all 15 target images, the test was automatically terminated. As a result, responses to all 45 items were not consistently obtained across participants, precluding the calculation of signal detection measures and response bias indices.
Step 3: Expert review and iterative refinement
The preliminary prototype was reviewed by an expert panel of 20 specialists, including neurologists, neuropsychologists, and digital application design experts. The panel evaluated the prototype for content validity, language clarity, and cultural appropriateness for Vietnamese users. Based on expert feedback, several refinements were implemented. The initial prototype included 10 semantic categories, which was expanded to 15 categories to improve discriminative capacity. The display duration for each image in the recognition phase was standardized to 3 seconds to reduce random responding. The interference module was also revised from an initial sand-art animation to a dynamic video with visual and audio elements to increase attentional engagement.
The final version of the TNmindtest was subsequently standardized for web-based administration. An overview of the TNmindtest architecture and representative task interface is presented in Figure 1. Overview and user interface of the TNmindtest platform
Phase 2 - Reliability evaluation (march–may 2025)
Test-retest reliability was examined in a community sample of adults aged 30–50 years with subjective memory complaints. Participants completed the TNmindtest twice under consistent conditions, with an average retest interval of 8.9 days. For each administration, one image per category was randomly selected from the available pool, resulting in conceptually equivalent but non-identical test versions across sessions. This approach was implemented to minimize stimulus-specific practice effects while maintaining comparable task difficulty. Potential confounders, including device type, time of day, sleep quality, medication use, and relevant medical history, were recorded. Data were entered into standardized Excel case report forms and independently verified by two researchers.
Phase 3 – Validity assessment (june – august 2025)
In the third phase, the discriminative validity of the TNmindtest was evaluated in detecting individuals with mild cognitive impairment (MCI). A separate sample of community-dwelling adults aged 30–60 years with subjective cognitive complaints was recruited. All participants completed the TNmindtest, the Montreal Cognitive Assessment (MoCA), and the Instrumental Activities of Daily Living (IADL) scale.
Participants were subsequently classified into probable MCI and non-MCI groups using an operational definition adapted from the DSM-5 criteria for mild neurocognitive disorder. 25 Probable MCI was defined by: (1) self-reported concern regarding cognitive decline; (2) objective cognitive impairment, defined as impairment in at least one cognitive domain, operationalized as a subdomain score below expected normative performance on the MoCA; (3) preserved functional independence, confirmed by the IADL assessment; and (4) absence of clinical dementia. In addition, the diagnosis of MCI was reviewed and confirmed by a neurologist specialized in dementia. Participants with no evidence of cognitive impairment on the MoCA and preserved functional independence were classified as cognitively normal (non-MCI). This classification approach enabled the identification of individuals with probable early cognitive decline while maintaining feasibility for community-based screening studies.
The Vietnamese MoCA has demonstrated acceptable diagnostic performance in previous studies. For example, a validation study among Vietnamese American older adults reported a sensitivity of 84.2% and specificity of 69.2% at a cutoff score of <26 26. The Vietnamese MoCA was administered prior to the TNmindtest in all cases, with both tests completed within a single 20-minute session. 26 To reduce potential bias, participants and evaluators were blinded to diagnostic classification. MoCA scoring and interpretation were conducted by neurologists trained in the standardized administration of the MoCA.
Participants
A total of 52 individuals aged 30–50 years with subjective memory complaints were initially recruited through community pharmacies. For the validity phase, an additional 80 adults aged 30–60 years were analysed and classified into two groups: non-MCI (n = 45) and MCI (n = 35) based on the criteria outlined above. Exclusion criteria included prior history of stroke, traumatic brain injury, psychiatric illness (e.g., schizophrenia, bipolar disorder), or current use of psychotropic medications.
Data collection
Demographic and clinical data were collected using a structured questionnaire. The information included age, sex, education level, occupation, marital status, technology use, living area, and contact details. Health history, such as neurological, psychiatric, and vascular risk-related medical conditions (hypertension, diabetes, dyslipidemia), was also documented. Participants self-reported environmental conditions during testing (quiet vs. distracting), test timing (morning, afternoon, evening), and sleep quality before testing. For reliability evaluation, data were collected at two time points: Test Session 1 and Test Session 2. The time interval, total test duration, average response time, and device type (smartphone, tablet, or computer) were recorded for each session. Information on stimulant or medication use (caffeine, alcohol, psychotropics) before each test was collected to identify potential confounding factors. TNmindtest scores were recorded in both phases, and MoCA scores were collected only in the validity phase. All data were checked for completeness and consistency before analysis.
Statistical analysis
All statistical analyses were conducted using RStudio version 2025.9.1.401. Statistical significance was set at p < 0.05. Test–retest reliability of the TNmindtest was examined using the Intraclass Correlation Coefficient (ICC) based on a two-way mixed-effects model with absolute agreement, along with Pearson’s correlation coefficient. Agreement between test sessions was evaluated using Bland–Altman analysis to visualize the mean differences and limits of agreement. To assess discriminative validity, between-group comparisons (MCI vs. non-MCI) of TNmindtest scores were conducted using the non-parametric Mann–Whitney U test. The diagnostic accuracy of the TNmindtest was evaluated using receiver operating characteristic (ROC) curve analysis. The area under the curve (AUC) was computed to quantify test performance. Sensitivity and specificity were estimated at the optimal cut-off point determined using Youden’s index.
Correlations between TNmindtest and MoCA total and domain-specific MoCA subscores were assessed using Spearman coefficients, with Pearson analyses as sensitivity checks; p values were adjusted using the Benjamini-Hochberg method, and a multivariable linear regression adjusted for age and education was additionally performed.
Results
Participant characteristics
Figure 2 illustrates the study flowchart. A total of 115 participants were enrolled across two study phases. Thirty-five individuals participated in the test–retest reliability phase (Phase 2), while 80 individuals were included in the validity phase (Phase 3) and were classified into two groups based on their MoCA performance: non-MCI (n = 45) and MCI (n = 35). Study flow chart
Baseline characteristics of participants (phase 2).
Baseline characteristics of participants by cognitive status (phase 3).
Note. Data are presented as median [Q1, Q3] for continuous variables and n (%) for categorical variables. P-values were calculated using the Mann–Whitney U test for continuous variables and the χ2 test or Fisher’s exact test for categorical variables, as appropriate.
Test-retest reliability
A total of 52 individuals aged 30–50 years with subjective memory complaints were initially recruited through community pharmacies for the test-retest reliability assessment. Following screening, 17 cases were excluded due to incomplete baseline information (n = 9) or failure to return for the retest session within the required timeframe (n = 8). Consequently, a total of 35 participants completed the TNmindtest twice to evaluate test–retest reliability. The average interval between sessions was 8.9 days. Median scores increased from 80.0 [74.0–86.5] at Time 1 to 87.0 [79.0–90.5] at Time 2, with a median difference of 7.0 points [4.2–9.8.]. The intraclass correlation coefficient (ICC, single measures) using a two-way mixed-effects model (absolute agreement) was 0.76 (95% CI: 0.59–0.88), indicating good reliability according to the Koo & Li interpretation scale.
27
A Bland–Altman analysis revealed a mean difference of −7.03 points between the two sessions, with 95% limits of agreement ranging from −22.92 to +8.86. This suggests a slight learning effect, though the differences remained within clinically acceptable bounds. Additionally, no significant influence was observed from potential confounders such as sleep quality, use of psychoactive substances, time of day, or testing device used across sessions (Figure 3). Bland-Altman plot of score differences between Time 1 and Time 2
Validity of the TNmindtest compared to MoCA
To evaluate construct validity, TNmindtest scores were compared between participants classified as having mild cognitive impairment (MCI) and those with normal cognitive function (non-MCI), using the MoCA. The median TNmindtest score was significantly lower in the MCI group (74.0 [60.0–80.0]) compared to the non-MCI group (80.0 [74.0–87.0]; p = 0.003). Similarly, MoCA scores were markedly lower in the MCI group (24.0 [23.0–25.0]) than in the non-MCI group (28.0 [27.0–29.0]; p < 0.001). To assess the discriminative performance of TNmindtest in detecting MCI, Receiver Operating Characteristic (ROC) curve analysis was performed (Figure 4). ROC Curve Comparison between TNmindtest and MoCA
In the ROC curve analysis, the area under the curve (AUC) for TNmindtest was 0.70, while the AUC for MoCA was 0.95, confirming the superior diagnostic performance of the MoCA. At the optimal cut-off score of 75%, the TNmindtest achieved a sensitivity of 62.9% and a specificity of 66.7% in detecting MCI.
The relationship between MoCA and TNmindtest scores was further examined via Pearson correlation analysis. As shown in Figure 5, the correlation coefficient was r = 0.31 (p = 0.005). The regression plot with a 95% confidence interval is also presented. Correlation between MoCA and TNmindtest
To further assess construct validity, we examined correlations between TNmindtest and both the total MoCA score and conceptually relevant MoCA subcomponents. TNmindtest was modestly correlated with MoCA total (Spearman rho = 0.279, p = 0.012) and delayed recall (rho = 0.293, p = 0.008). The strongest association was observed for the combined MoCA memory-plus-attention subtotal (rho = 0.327, p = 0.003). By contrast, the MoCA attention subtotal showed only a borderline association (rho = 0.213, p = 0.058), and individual attention items were not significantly correlated with TNmindtest. These findings remained directionally consistent in Pearson analyses. After correction for multiple comparisons, the associations with MoCA total, delayed recall, and the combined memory-plus-attention subtotal remained significant (Figure 6). Correlation between Tnmindtest scores and MoCA subdomains (delayed recall and memory-attention subtotal scores)
Discussion
Our study has demonstrated the development and validation of the TNmindtest for the Vietnamese population. The test has been designed for ease of self-administration, with a focus on the cognitive domain of attention and memory. TNmindtest exhibits good test-retest reliability and modest diagnostic value, which may be effectively utilized in triage or preliminary screening settings.
The demographic and clinical profiles across both study phases indicated a relatively homogeneous and well-balanced sample. Participants were primarily of working age (median 37-42 years), highly educated (median 16 years of formal education), and frequent users of digital devices. These characteristics are advantageous in the early validation phase of a digital cognitive screening tool, as they reduce confounding factors such as limited digital literacy or age-related sensory impairments that could influence performance. The high proportion of technology-familiar individuals in both the test–retest (88.6%) and validation (98.8%) phases further supports the reliability of the TNmindtest results, as performance was unlikely to be biased by unfamiliarity with digital interfaces. This aligns with findings from prior studies, such as evaluations of the BrainCheck battery and other computerized assessments, where digital literacy was positively associated with test usability, completion rates, and user satisfaction.28,29
Although participants with MCI were older and had slightly lower education levels than the non-MCI group, these differences were not statistically significant (p = 0.201 and p = 0.085, respectively), supporting demographic comparability. Consistent with previous literature,19,30 age and education are known to modulate cognitive performance, particularly in tasks involving memory and processing speed. Furthermore, the prevalence of medical conditions potentially affecting cognition, such as hypertension, diabetes, and prior neurological injury, was similar between groups, minimizing the risk of clinical confounding. It is also worth noting that all participants resided in urban areas, which may have enhanced the feasibility of the digital test through better access to devices and stable testing environments. However, this urban sample profile may limit generalizability, as discussed further in the section on limitations.
The intraclass correlation coefficient (ICC) of 0.76 (95% CI: 0.59–0.88) observed for the TNmindtest indicates moderate to good test-retest reliability, exceeding the commonly accepted threshold of 0.75 for clinical tools. 27 Although it does not reach the “excellent” level (≥0.90), this result is consistent with similar digital cognitive assessments, such as the Mindmore battery (ICC ≥ 0.70). 31 In comparison, paper-based tools like the Montreal Cognitive Assessment (MoCA) typically report higher ICC values, ranging from 0.85 to 0.92, when administered by trained personnel.32,33 The Mini-Mental State Examination (MMSE) demonstrates a wider reliability range (0.60–0.93), influenced by testing conditions, intervals, and populations. 34 The interval between test administrations in our study was approximately 9 days, which is considered short but appropriate for preliminary reliability testing. Shorter intervals reduce the likelihood of actual cognitive change but increase susceptibility to practice effects. While no standardized optimal interval exists, many cognitive assessment studies use 1-4 weeks to balance temporal stability and minimize memory or familiarity effects. 35
A learning effect was observed, with participants demonstrating an approximately 7% improvement on the second administration. This magnitude is consistent with prior reports from repeated cognitive testing, where short retest intervals are associated with performance gains due to familiarity with test procedures and interface. 36 Similar effects have been documented in both digital and traditional tools, such as RBANS. 37 Practice effects are a known limitation of repeated cognitive testing and are often attributed to improved familiarity with test structure, response format, or digital interface. 38 The use of fixed stimuli in TNmindtest may have contributed to this effect, highlighting the need for randomized or alternate test forms in future versions. Compared with other mobile cognitive tests, such as SWAY, 39 which showed minimal practice effects over similar intervals, TNmindtest exhibited a moderate learning effect. While this level of improvement is within an acceptable range, it should be considered when interpreting repeated measures, particularly in longitudinal or clinical monitoring contexts.
Our study provides preliminary evidence for the validity of TNmindtest as a digital cognitive screening tool for detecting mild cognitive impairment, using the Montreal Cognitive Assessment (MoCA) as the reference standard. While TNmindtest successfully distinguished between individuals classified as MCI and non-MCI by MoCA, its diagnostic performance, as reflected by an area under the receiver operating characteristic curve (AUC = 0.70), was modest relative to MoCA’s performance, which has been consistently reported with AUC values around 0.84 in meta-analyses across diverse populations.40,41 According to established diagnostic accuracy guidelines, an AUC between 0.70 and 0.80 is considered acceptable, although not strong enough for standalone clinical decision-making. 27 This suggests that TNmindtest may be used in triage or preliminary screening settings, particularly in communities where face-to-face assessments, such as MoCA, are constrained by financial or personnel limitations. Notably, digital tools such as BrainCheck (AUC = 0.79), 28 MemTrax (AUC = 0.799–0.839),30,42 and Neurotrack (AUC = 0.76) 43 have demonstrated similar or slightly better diagnostic performance, supporting their role in scalable, first-line cognitive screening models.
TNmindtest demonstrated a moderate correlation with the total MoCA score, indicating that it captures a meaningful aspect of overall cognitive performance while not fully overlapping with a multidomain global cognitive screening instrument. This level of association supports convergent validity at the level of general cognitive function and suggests that TNmindtest may be suitable as a pragmatic screening tool for identifying individuals with potential cognitive impairment who require further evaluation. However, TNmindtest should not be considered a measure of global cognition equivalent to the MoCA. 44
Notably, stronger associations were observed with delayed recall and the combined memory–attention subtotal, providing more direct evidence of domain-specific construct validity. These findings indicate that TNmindtest more closely reflects the cognitive domains it was specifically designed to assess, particularly episodic memory performance. This pattern is clinically relevant, as impairment in delayed recall is a core feature of amnestic mild cognitive impairment and is strongly associated with early Alzheimer’s disease–related cognitive decline. 21 Previous studies have demonstrated that memory-focused indices derived from the MoCA, such as delayed recall and composite memory scores, show improved sensitivity in identifying individuals within the amnestic MCI spectrum. 45 In this context, the observed correlation profile suggests that TNmindtest may be particularly suitable for screening individuals with memory-predominant cognitive impairment consistent with amnestic MCI.
In contrast, weaker correlations with individual MoCA attention items should not be interpreted as evidence against the construct validity of TNmindtest. Rather, they may reflect the limited granularity of the attention subtests included in the MoCA, which primarily consist of brief and structured tasks (e.g., digit span, vigilance, and serial subtraction). 44 These measures may not adequately capture the more complex and dynamic attentional and working memory demands embedded in TNmindtest. Collectively, these findings support the positioning of TNmindtest as a domain-sensitive cognitive screening tool with particular relevance for detecting memory-dominant cognitive impairment, while maintaining utility in broader cognitive screening contexts.
In summary, TNmindtest shows promise as a practical and scalable digital cognitive screening tool, particularly in settings where access to conventional assessments is limited. Its ease of administration and domain-sensitive profile support its potential role in early identification and triage of individuals at risk for cognitive impairment. Further studies incorporating larger and more diverse populations, as well as biomarker-informed classification, are warranted to establish its clinical applicability and long-term utility.
Strengths and limitations
This study presents a systematically developed and culturally adapted digital cognitive screening tool for the Vietnamese population. The web-based, self-administered design with automated scoring enhances scalability and reduces examiner-related bias. The multi-phase approach, incorporating both reliability and validity assessments, together with the use of randomized stimuli across administrations, further strengthens the internal and construct validity of the findings.
However, several limitations should be considered. The sample was relatively small and consisted predominantly of urban, well-educated individuals with high digital literacy, which may limit generalizability to broader populations. The short test–retest interval may have contributed to practice effects despite efforts to mitigate them through randomized stimuli. In addition, reliance on the MoCA as the sole reference standard may have introduced misclassification bias, as comprehensive neuropsychological evaluation was not performed. Finally, the cross-sectional design precludes assessment of longitudinal cognitive changes.
Conclusions
TNmindtest shows promise as a brief, self-administered digital tool for the early detection of mild cognitive impairment. While initial findings support its feasibility and acceptable psychometric performance, its clinical utility remains preliminary. Further validation in larger and more diverse populations, along with comparison against comprehensive diagnostic standards, is required to establish its role as a scalable cognitive screening solution in Vietnam and similar settings.
Footnotes
Acknowledgments
The authors appreciate the staff who assisted with participant recruitment, data collection, and logistical support. We also thank all participants for their valuable time and engagement in this study.
Ethical considerations
The study protocol was approved by the Institutional Review Board of the University of Medicine and Pharmacy at Ho Chi Minh City (Approval No.136/2025/HĐ-DHYD).
Consent to participate
All participants were provided with detailed information regarding the study’s objectives and procedures, and written informed consent was obtained prior to enrollment.
Author contributions
Tran Cong Thang: Conceptualization, Methodology, Supervision, Project administration, Writing - review & editing. Quynh Phuong Vo: Conceptualization, Methodology, Investigation, Writing -original draft. Vinh-Khang Nguyen: Conceptualization, Methodology, Formal analysis, Writing-review & editing. Duc Nhut Le: Investigation, Technical support. Long Bao Trinh: Investigation, Technical support. Ly Thi Ngoc Nhi: Investigation, Technical support.
Funding
The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This study received partial financial support from Mayoly Pharma France, which covered costs related to the development of the TNmindtest application. The funding organization had no role in the study design, data collection, analysis, interpretation, or manuscript preparation.
Declaration of conflicting interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Data Availability Statement
The data that support the findings of this study are not publicly available due to institutional policy but are available from the corresponding author upon reasonable request.