Which assessment tools best distinguish between mild cognitive impairment and dementia? Lessons from a Slovak memory clinic cohort Authors: Petr Novak, Stanislav Katina, Petra Brandoburova, Michaela Jezberova, Veronika Reznakova, Jozef Hanes, Frantisek Jurcaga, Peter Koson, Michal Novak, Linus Jönsson, Norbert Zilka Published: 2026 ABSTRACT AbstractBackgroundThe utility of various cognitive assessment tools for distinguishing between mild impairment and dementia, as well as for determining cutoff values for specific populations, continues to be the subject of extensive research. Here, we assessed the utility and feasibility of these tools in the first Slovak memory clinic cohort.MethodsWe enrolled a Slovak memory clinic cohort of patients with MCI and dementia (MCI, n = 84; dementia, n = 55). The participants were characterized using a range of cognitive assessment tools—Auditory Verbal Learning (AVLT), Category and Letter Fluency (CFT, LFT), Digit Span (DSF, DSB), Digit-Symbol Coding (DS-C), Frontal Assessment Battery (FAB), MMSE, MoCA, Rey Osterrieth Complex Figure (ROCF), and Trail Making Test (TMTA, TMTB); clinical assessments—Amsterdam Instrumental Activities of Daily Living (A-IADL) and 5-level EuroQol questionnaire with 5 dimensions (EQ-5D-5 L); and scales for anxiety, dependency, depression, dignity, and MRI volumetry. The ability of the various assessments to distinguish between MCI and dementia was evaluated.ResultsOver the course of three years, at a single memory clinic, it was feasible to enrol and evaluate a total of 150 participants, 139 of whom fulfilled the definition of either MCI or nonvascular dementia. Of the employed cognitive and clinical assessment tools, the best differentiation between MCI and dementia was observed for the AVLT and A-IADL. The DSF and DSB tests did not reveal differences between the populations. No differences were observed in education, vital signs, or anthropometric measurements. Participants with dementia had greater degrees of brain atrophy in the hippocampi and frontal, parietal, and temporal cortex; lower total brain volumes; and greater ventricular dilation.ConclusionsThis study confirms the utility of a range of cognitive assessment tools and scales for differentiating between MCI and dementia but reveals that some commonly employed tools, such as the DSF and DSB, may not be sensitive to these differences. This study highlights the importance of accurate assessment of the ability to perform activities of daily living and supports the development of objective, ecologically valid assessments of IADL.Supplementary InformationThe online version contains supplementary material available at https://doi.org/10.1186/s12877-026-07801-3.Keywords: Alzheimer’s disease, mild cognitive impairment, neuropsychology, observational study Introduction Alzheimer’s disease (AD) is a complex neurodegenerative disorder characterized by pathological deposits of amyloid [1] and tau proteins [2] accompanied by neuroinflammation [3] followed by neuronal loss, brain atrophy, and loss of cognitive function. As pathophysiological changes become more pronounced and proteinaceous lesions in the brain continue to accumulate, AD progresses from a mostly silent preclinical period through a stage of mild cognitive impairment (MCI), ultimately culminating in dementia [4, 5]. While memory impairment is the most common initial symptom of AD [6], based on the distribution of pathology, the disease may also present with language impairment [7], dysexecutive symptoms [8], or visual impairment [9]. Over the course of AD, memory, attention, orientation, language, judgment, psychomotor speed, and visuomotor skills can be affected as pathological lesions spread [10]. Over one hundred years since the description of AD [11] and over forty years since the disorder drew widespread interest from the scientific community, a consensus on how best to track the progression of AD has yet to emerge—understandably—because there is a wide range of possible approaches to test a given domain of cognition. A wide array of cognitive tools is thus being employed in memory clinics and clinical research, some by virtue of their seniority [12] and some as a result of recent optimization and harmonization efforts [13, 14]. Another important question is how best to define the milestone of a patient transitioning from mild cognitive impairment to dementia, which is important, e.g., for the allocation of caregiving resources [15]. All of the above are particularly relevant for the Slovak Republic. Overall, the Slovak population is characterized by a high prevalence of risk factors for AD and dementia, including high rates of obesity, heart disease and diabetes [16]. As such, the population is of particular interest for AD research. The cohort utilized in the present study was established at the MEMORY Centre n.o. In Slovakia, the diagnosis of Alzheimer’s disease is still frequently made only once the patient reaches the dementia stage; given the paucity of dedicated dementia facilities, specialist memory centres often see a wide spectrum of disease severities across the patient population. Patients also revisit memory centres over the course of their disease, as the alternatives available are limited. These factors make the established cohort suitable for both longitudinal observation and cross-sectional analyses; efforts are being made to extend the cohort across multiple memory centres throughout Slovakia, similar to efforts in the neighbouring Czech Republic [17]. In the present analysis, we sought to characterize a cohort of patients with mild cognitive impairment or dementia using a diverse array of assessment tools and evaluate the utility of these assessment tools in discriminating between patients with mild cognitive impairment and those with dementia. The study also served as a pilot study on the utility and feasibility of various assessments for use in the planned nation-spanning cohort. Methods Cognitive and functional assessments An overview of the employed cognitive and clinical assessment tools is provided in Table 1. Table 1Overview of cognitive and clinical assessment toolsAssessment ToolDescriptionDirectionScore rangeAmsterdam IADLTotal score and subdomainsActivities of daily livingLower = better0–4(individual domains; total score is an average of domains)Auditory Verbal Learning Test - Trials 1–5 (total sum) (A5)MemoryHigher = better0–75Auditory Verbal Learning Test - Immediate recall after interference set (A6)Memory and resistance to interferenceHigher = better0–15Auditory Verbal Learning Test - Delayed recall after 30 min (A30)Delayed recallHigher = better0–15Beck’s Depression InventoryDepression questionnaire (patient-facing)Lower = better0–63Category Fluency Test (animals, vegetables)Executive function, fluency, languageHigher = better≥ 0DEM-QOLQuality of life, patient-facingLower = better0–84 (item sum)0–3 (overall)Digit Span ForwardImmediate recallHigher = better≥ 0Digit Span Backwardplanning, decision makingworking memoryHigher = better≥ 0Digit Symbol SubstitutionAttention, processing speed, executive functionHigher = better≥ 0EQ-5D-5 LSubitemsPatient-facing questionnaire about Mobility, Selfcare, Usual activitiesPain/Discomfort, Anxiety/DepressionLower = better1–5EQ-5D-5 LVisual scalePatient’s subjective estimation of their subjective assessment of their own healthHigher = better0–100Frontal Assessment BatteryFrontal functionHigher = better0–18Generalized Anxiety DisorderAnxiety questionnaire (patient-facing)Lower = better0–21Geriatric Depression ScaleDepression questionnaire (patient-facing)Lower = better0–15Letter Fluency TestExecutive function, fluency, languageHigher = better≥ 0MMSETotal score and subdomainsSubdomains: Orientation to time, Orientation to place, Registration, Attention and calculation, Recall, Language, Complex commandsHigher = bettersingle digit for domains0–30 totalMoCATotal scoreSubdomains: visuospatial, executive, naming, attention, language, abstraction, delayed recall, and orientationHigher = better0–30ROCF – copyingPerceptual-motor, visual perception, constructionalHigher = better0–36ROCF – reproduction (3 min delay)Short-term memoryHigher = better0–36ROCF – reproduction (30 min delay)Delayed recallHigher = better0–36Trail Making Test AAttention (sustained, selective, divided), processing speedLower = better≥ 0Trail Making Test BFeedback use, flexibilityLower = better≥ 0 Overview of cognitive and clinical assessment tools Amsterdam IADL Total score and subdomains 0–4 (individual domains; total score is an average of domains) 0–84 (item sum) 0–3 (overall) planning, decision making working memory EQ-5D-5 L Subitems Patient-facing questionnaire about Mobility, Selfcare, Usual activities Pain/Discomfort, Anxiety/Depression EQ-5D-5 L Visual scale MMSE Total score and subdomains single digit for domains 0–30 total MoCA Total score The Mini-Mental State Examination (MMSE) and Montreal Cognitive Assessment (MoCA) are both screening assessments with scores ranging from 0 to 30, with lower scores indicating greater impairment. The MMSE is designed with the dementia stages of AD in mind, with items covering orientation to time and place, registration, attention and calculation, recall, language, and complex commands [12]. The MoCA is designed to include more items sensitive to mild cognitive impairment and covers the visuospatial, executive, naming, attention, language, abstraction, delayed recall, and orientation domains of cognition [18]. The Category Fluency Test (CFT) requires the participant to name as many words fitting into a certain category as possible within 60 s [19]. The categories used for this study were ‘animals’ and ‘vegetables’. The Letter Fluency Test (LFT) requires the participant to say as many words starting with a certain letter as possible within 60 s [19]. The letters used for the Slovak version were K, P, and S. The Rey–Osterrieth Complex Figure (ROCF) test is a text of praxis (copying a complex design) and visual memory (delayed recall and reproduction of said design) [20]. The Digit Span Tests (DST) evaluate working memory by asking the subject to recall sequences of numbers in their original order (forward) or in reverse order (backward) [21]. The Auditory Verbal Learning Test (AVLT) assesses short-term and long-term memory, utilizing 15-item word lists, with five learning trials, an interference set, and delayed recall [22]. The Digit Symbol Substitution Test (DSST) assesses working memory, processing speed, and judgment; the subject is required to substitute specific symbols for a sequence of numbers, as many as possible within 90 s [23]. The Frontal Assessment Battery (FAB) is a brief test of frontal function that scores 6 domains 0–3 [24]. For all of the above, higher scores indicate better function. The Trail Making Tests (TMTs) are timed visual attention and task switching assessments requiring the subject to connect a sequence of numbers (TMT-A) or a sequence of alternating letters and numbers (TMT-B) [25]. Lower scores indicate better function. The Beck Depression Inventory [26], Dependence Scale [27], Generalized Anxiety Disorder (GAD) [28], and Geriatric Depression Scale (GDS) [29] are questionnaires designed to assess depression, dependence, and anxiety, respectively; higher scores indicate greater severity. The study employed the 30-item version of the Amsterdam IADL (A-IADL) scale [30]; higher A-IADL scores indicate worse function. The total score is the average of all scorable items for a given patient. The EQ-5D-5 L is a standardized measure of health-related quality of life; higher scores indicate worse quality of life [31]. MRI Structural and morphological magnetic resonance imaging (MRI) scans of the brain were acquired on an Ingenia Elition X 3.0T unit (Philips Medical Systems, Eindhoven, Netherlands) using a 16-channel head coil. The 3D T1-weighted TFE (turbo field echo) sequence was used for volumetric assessments. Volumetric evaluation was performed using Icobrain MR 5.13.0 software (Icometrix, Wilsele, Belgium). Biomarker analysis Concentrations of neurofilament light chain (NFL) were measured by digital ELISA using a Simoa™ NF-Light Advantage Kit (Quanterix, Boston, USA). Prior to analysis, the samples were blinded, and all the assays were performed on an HD-X Analyzer (Quanterix Corp., MA, USA). Briefly, frozen plasma samples were melted and centrifuged for 10 min at 25 °C and 20,000 × g, after which the supernatants were transferred to a Simoa sample plate together with calibrators for NFL (parts of the kits). Each measurement was performed in duplicate according to the manufacturer’s recommendations. Concentrations were calculated using Simoa™ HD-X instrument software. Statistical analysis Data processing and statistical analyses were performed in the R programming environment version 4.3.2 (R Core Team [32]). Discrete variables For each task with categorical scores – such as MMSE items (scored 0–8), A-IADL tasks (0–4), or EQ-5D-5 L items [1–4] – we compared the distribution of patient responses between MCI and AD using a chi-square test. This allowed us to determine whether patients with MCI and AD showed statistically different patterns in how frequently they selected each score category. Formally: the null hypothesis of equal score distributions (i.e., homogeneity of score probabilities across response categories) between AD and MCI was tested using the chi-square test at a significance level of 0.05 [33]. The alternative hypothesis was that at least one score level for a given task differs in probability between the diagnoses. P values were calculated via Monte Carlo simulation with 10,000 repetitions to avoid issues with an asymptotic chi-square distribution when the cell counts (scores) were too small. If the absolute frequency of some level was equal to zero for both AD and MCI patients, this level was excluded from the analyses. For each item in the MMSE, A-IADL, and EQ-5D-5 L assessments, we calculated the number and percentage of patients in each diagnostic group (MCI and AD) who had each possible response score. These response distributions were summarized for both groups, including the total counts per group. Continuous variables For continuous variables, such as demographic variables, vital signs, MRI regions of interest (ROIs), biomarkers, and total scores of cognitive and functional assessments, the data were separately analysed for each diagnosis, and the data were Winsorised using the Tukey interquartile range (IQR) [34] due to the presence of outliers on both sides of the distribution; i.e., values smaller than the first quartile minus 1.5 times the IQR and greater than the third quartile plus 1.5 times the IQR were substituted for these limits. The distributions were visualized using boxplots with jittered data and the mean for each diagnosis. The null hypothesis that the mean difference between AD and MCI diagnoses is equal to zero was tested against the alternative hypothesis that the mean difference between AD and MCI diagnoses is not equal to zero by two-sample Student’s t test with Welch degrees of freedom for mean difference [35] at a significance level equal to 0.05. The results are presented numerically as the mean difference, standard deviation of the mean difference, t-statistic, degree of freedom, p value, and lower and upper bounds of the Wald two-sided 95% confidence interval (CI) for the mean difference. The mean differences and the CIs of the mean differences were also visualized graphically. To assess the diagnostic ability of all the variables for distinguishing between the two diagnoses, receiver operating characteristic (ROC) analysis was used (Pepe 2003). The results were summarized numerically by the area under the curve (AUC), lower and upper bounds of the Wald two-sided 95% CI for an AUC, cutoff point, optimal sensitivity, and specificity with the respective lower and upper bounds of the Wald two-sided 95% CI calculated using logit transformation. If the optimal sensitivity or specificity is equal to zero or one, the corresponding 95% CI is not reported. The ROCs were also visualized graphically. Correlations We assessed associations between plasma neurofilament light chain (NfL) and selected brain regions of interest (ROIs) using Pearson correlation coefficients. First, within each diagnostic group (AD and MCI) we computed Pearson correlations between NfL and each ROI and tested the null hypothesis of no association versus an alternative of any association using a one-sample Fisher z-test. Second, we compared the Pearson correlation coefficients between diagnoses by calculating the difference in these correlation coefficients (AD minus MCI) and tested the null hypothesis of no difference in association between AD and MCI versus any difference in association between AD and MCI using a two-sample Fisher z-test. Results Population A total of 150 patients with mild cognitive impairment (MCI) (n = 84), dementia (n = 55) or subjective cognitive decline (n = 11) were enrolled between March 2019 and February 2022. The enrolled population was mostly female, with a majority at the MCI or mild-to-moderate dementia stage of cognitive impairment. Patients were mostly retired, with age being the primary reason for retirement (see Table 2). Table 2Basic characteristics of the enrolled populationCharacteristic n percentSexMale4624.47Female10455.32Patient statusSubjective cognitive decline115.85Mild cognitive impairment8444.68Dementia5529.26Primary CNS disorderAlzheimer’s Dementia5127.13Not reported9952.66Employment statusNot Known115.85Retired12264.89Employed (Full Time)105.32Employed (part-time)21.06Unemployed52.66Reason for retirementNot Reported2312.23Age12264.89Other Health Issue52.66Patient living arrangementNot Reported115.85Own home12667.02A relative’s/friend’s home73.72Institution (e.g., care facility)63.19 Basic characteristics of the enrolled population Comparison of baseline characteristics Compared to MCI patients, dementia patients were older; no statistically significant differences were observed in vital signs or anthropometric measurements; on average, the enrolled population was hypertensive and overweight (Fig. 1). Demography and anthropometric measurements. The box indicates the median and quartiles; the whiskers indicate the range. The red dots indicate the arithmetic mean. * = p < 0.05; ** = p < 0.01; *** = p < 0.001 Brain volumetry Appreciable differences in brain area volumes were observed between dementia patients and MCI patients. Dementia patients displayed more pronounced brain atrophy, having, on average, a 13.2% smaller hippocampal volume, a 5.7% smaller frontal cortex volume, a 9.5% smaller temporal cortex volume, a 4.7% smaller parietal cortex volume, and a 35.1% greater ventricular volume in dilated ventricles than did their MCI counterparts (Fig. 2). MRI volumetry. The box indicates the median and quartiles; the whiskers indicate the range. The red dots indicate the arithmetic mean. * = p < 0.05; ** = p < 0.01; *** = p < 0.001. The occipital cortex volume was available from n = 22 only; hence, the observed difference was not statistically significant Cognitive tests and clinical scales and their ability to distinguish between MCI and dementia A majority of the utilized tests distinguished between population means of MCI and dementia populations, with the Montreal Cognitive Assessment (MoCA), Rey‒Osterrieth Complex Figure test (ROCF), and Auditory Verbal Learning Test (AVLT) showing somewhat greater mean differences between populations than the Mini-Mental State Examination (MMSE) (see Fig. 3; detailed results are listed in Table 3). It is worth noting that some domains of the MMSE (language and registration) and the Digit Span tests did not distinguish between population means of MCI and dementia population means. The most pronounced difference between the population means was observed for the AVLT (t-statistic: -7.99) and A-IADL (t-statistic: 7.81). Patients with dementia also displayed higher levels of dependence, while patients with MCI scored higher on measures of anxiety and depression (see Fig. 4). Cognitive tests. The box indicates the median and quartiles; the whiskers indicate the range. The red dots indicate the arithmetic mean. * = p < 0.05; ** = p < 0.01; *** = p < 0.001. For tests with few discrete outcomes, such as the Digit Span, multiple data points are overlaid Table 3Comparison of baseline characteristics between MCI patients and dementia patientsComparisons with p > 0.05 are grayed out. The further t-statistics deviate from 0, the more intense color coding was used; green was used for higher values in dementia patients, and red was used for higher values in MCI patients. df = degrees of freedom, LB = lower boundary of the 95% confidence interval, UB = upper boundary of the 95% confidence interval Comparison of baseline characteristics between MCI patients and dementia patients Comparisons with p > 0.05 are grayed out. The further t-statistics deviate from 0, the more intense color coding was used; green was used for higher values in dementia patients, and red was used for higher values in MCI patients. df = degrees of freedom, LB = lower boundary of the 95% confidence interval, UB = upper boundary of the 95% confidence interval Other assessments. The box indicates the median and quartiles; the whiskers indicate the range. The red dots indicate the arithmetic mean. * = p < 0.05; ** = p < 0.01; *** = p < 0.001. For tests with few discrete outcomes, such as the Digit Span, multiple data points are overlaid. For the AIADL, only the total score is shown. See the supplementary material for a full tabulation To further characterize the ability of various assessments to distinguish between dementia and MCI, we calculated responder-operator characteristics (area under the curve (AUC), optimal sensitivity, optimal specificity, asymmetry, and cutoff; see Table 4. Table 4Comparison between MCI patients and dementia patients: responder operator curvesTPF true positive fraction, FPF false positive fraction. The area under the curve (AUC) is colour coded, with red representing lower and green representing higher values Comparison between MCI patients and dementia patients: responder operator curves TPF true positive fraction, FPF false positive fraction. The area under the curve (AUC) is colour coded, with red representing lower and green representing higher values Correlation analyses Plasma NfL was strongly intercorrelated with cognitive impairment and brain atrophy. Salient associations at p < 0.05 or r > 0.3 with a range of disease indicators. Higher plasma NfL levels were significantly associated with ventricular expansion (p < 0.0001 and r = 0.7415 in MCI; p = 0.03740 and r = 0.5376). Smaller hippocampal volume and total brain volume were associated (r > 0.3) with higher plasma NfL in both DEM and MCI; smaller frontal/temporal/parietal cortex volumes were associated with NfL in DEM patients. Negative correlations between NfL and MMSE Orientation to Place (p = 0.0003, r = -0.7619) and total MMSE score (r = -0.3298), as well as total MoCA score (r = -0.3696) in DEM patients suggest an association between elevated NfL and general cognitive impairment. Trail Making Tests A and B showed correlations with NfL in DEM patients (r = -0.3298 and − 0.5707 respectively), indicating slower processing speed and impaired executive function among individuals with higher NfL. Category fluency animals and vegetables were similarly associated with NfL levels in MCI patients (r = -0.3794, p = 0.05550, and r -0.5292, p = 0.00470 respectively). The A-IADL score was significantly negatively correlated with NfL in the MCI group (r = 0.4475, p = 0.04100). The fact that the associations in MCI were discrepant with DEM for some markers, e.g., EQ-5D-5 L indicates that the association of NfL with various AD markers may not be linear across the entire AD continuum. As expected, memory-focused tests such as the AVLT associated with hippocampal volume especially in the MCI population, with delayed recall after 30 min showing the strongest association (r = 0.4230, p < 0.05). Category fluency was most strongly associated with total brain volume and ventricular expansion (r = 0.4121 and − 0.4171, p < 0.05). A correlation heatmap is provided in Supplement 2. A selection of correlation analyses of NfL vs. other variables is provided in Supplement 3, and details on all correlation analyses in Supplement 4. The ability to perform individual IADLs differed between the MCI and dementia groups, with the dementia group exhibiting worse performance in all activities. Regardless, for most activities, in both groups there were patients at both extremes of the ability spectrum – those able to perform the activity without problems or not at all. A difference in IADL difficulty was apparent, with all participants with MCI being able to perform simpler activities, such as using household appliances or preparing coffee, while more complex activities, such as using the GPS or filling in official forms, were beyond the ability of many participants with MCI and all participants with dementia (see Fig. 5). Of the individual IADLs, the ability to use household appliances or the ATM differentiated best between MCI and dementia (see Table 5). Differences in the ability to perform individual AIADL tasks between patients with MCI and patients with dementia. A score of 0 indicates no impairment, and a score of 4 indicates impairment preventing performance of the given IADL Table 5Comparison of results for subdomains of the Amsterdam IADL, MMSE, and EQ-5D-5L between MCI patients and dementia patients. The statistics are color-coded, with green values corresponding to high differences and red values corresponding to low differences. For all IADL and MMSE items, the differences were in favor of those of the MCI group P values < 0.05 are highlighted in green Comparison of results for subdomains of the Amsterdam IADL, MMSE, and EQ-5D-5L between MCI patients and dementia patients. The statistics are color-coded, with green values corresponding to high differences and red values corresponding to low differences. For all IADL and MMSE items, the differences were in favor of those of the MCI group P values < 0.05 are highlighted in green Discussion The recruited cohort was a fairly representative population of individuals with MCI/mild dementia aged on average 74.3 and 78.8 years, respectively, who displayed impairments in memory and orientation. Importantly, as this is the first thoroughly characterized cohort enrolled in Slovakia, it showcases that MCI and dementia populations can be enrolled in Slovakia, and the resulting cohorts are fundamentally comparable with those enrolled in Western Europe. To classify a patient as having MCI or dementia, the neuropsychologists used the entirety of the available information, avoiding linking the decision to a single cutoff on a single test. As expected, the delayed reproduction and recall tests (Auditory Verbal Learning and ROCF) were quite sensitive at differentiating between means of MCI and dementia. One of the best mean differentiators appeared to be the ability to perform activities of daily living – understandably so, considering that a major distinguishing factor between MCI and dementia is the ability to live independently. Nevertheless, it is informative to see that simply asking the primary caregiver a question about the patient’s ability to, e.g., use household appliances or an ATM may have very high value for initial screening. This highlights the need to establish interpretable, up-to-date IADL scales and optimize their item content, as was done for the Amsterdam IADL [30, 36]. Similarly, for drug research, IADL scales are valuable for supporting the clinical relevance of drug effects [37]. A major limitation of IADL assessment is that patients themselves are poor reporters, and the scales rely on informants (often the patient’s relatives and caregivers). Thus, the development of objectively scorable tasks to evaluate competence in activities of daily living should be considered; such tasks would need to be feasible in a memory clinic setting. Several of the tasks that discriminate between MCI patients and AD patients with greater accuracy (such as using the ATM) are theoretically suitable for implementation as test paradigms, e.g., on a tablet. As expected for a cohort with a mostly amnestic phenotype [38], various tests of executive function (e.g., Letter and Category Fluency Tests) did not distinguish between the MCI and AD populations as strongly as memory tests or activities of daily living but still showed noticeable (and statistically significant) differences between the means of the two populations. This illustrates the fact that dementia becomes more generalized as Alzheimer’s pathophysiological changes leave the confines of the medial temporal lobe and spread throughout the brain [39, 40]. This finding is also in line with the (admittedly slight, ~ 9%) mean difference between the groups in frontal lobe volume and with published studies [41]. Curiously, although they are a part of recent batteries such as the cUDS [13], the Digit Span Tests (which aim to measure immediate recall and working memory) did not differentiate means between dementia and MCI in the present cohort; this highlights the need to continuously evaluate the performance of cognitive assessment tools. The observed mean differences in brain area volumes corresponded to approximately 2–3 years of atrophy of the hippocampus and 4–5 years of atrophy of the ventricles (estimated based on reported atrophy rates) [42, 43]. The mean self-reported depression and anxiety rates were lower in the dementia population, highlighting the fact that patients at the dementia stage suffer from anosognosia, while patients with MCI are painfully aware of their impairment [44] and often respond with depression or anxiety [45]. The empirical cutoffs derived from the study population aligned well with those reported in the literature, e.g., the MoCA cutoff for the two populations was established at 19.5, and the MMSE cutoff was established at 22.5 [46]. The A-IADL cutoff derived from the study (at 1.81) is in line with the intended scoring system; scores of 1 indicate that a given activity is ‘slightly more difficult’ to perform (corresponding conceptually to mild impairment), while scores of 2–4 indicate moderate or severe difficulty or inability to perform a given activity [30, 36]. The data support the notion that the MMSE is not strictly necessary in studies featuring the MoCA due to lower sensitivity and redundancy; it was included in this study for comparability purposes. While NfL was collected only in a subset of study participants, the results highlight its usefulness for studies in AD across the MCI-dementia continuum. Conclusions The study evaluated the utility of a range of assessment tools for differentiating between MCI and dementia. Activities of daily living assessments stood out as an important differentiator, indicating both the need to have good IADL assessment tools in the neuropsychologist’s arsenal and highlighting the utility of IADL assessment in initial screening. Limitations The study population was not characterized via ground-truth AD biomarkers (CSF or PET). The method used for classifying a participant as MCI or dementia patient was a subjective assessment by the examining neuropsychologist based on a discussion with the patient and caregiver (where available), in combination with the MMSE and MoCA scales, and was not based on a single cutoff point on any single test; moreover, while this evaluation is thorough and comprehensive, it cannot eliminate personal bias. As the decision is based on cognitive ability and on questions pertaining to the ability to perform activities of daily living, this will influence the score distribution of assessments measuring similar constructs to a certain degree. The remaining assessments (Amsterdam IADL, the various cognitive tests of the battery) were used for the first time in a Slovak dementia population, and were not used in the diagnostic decision-making; also, the tests (other than MoCA and MMSE) were conducted after the MCI/dementia classification was made, minimizing the risk of circularity. The majority of patients in the dementia group had mild dementia, limiting the ability to compare results for patients with mild versus moderate dementia. Supplementary Information Supplementary Material 1. Supplementary Material 2. Supplementary Material 3. Supplementary Material 4. Acknowledgements In the loving memory of Michaela Novakova, who contributed significantly to this work, but was taken from us too soon by the cruel vagaries of fate.This work received support from grants: JPND grant “ADDITION: Alzheimer’s disease data-driven insights on individual outcomes of importance” and APVV grant 20-0447.The Slovak MMSE was used by the study team without permission, however this has now been rectified with PAR. The MMSE is a copyrighted instrument and may not be used or reproduced in whole or in part, in any form or language, or by any means without written permission of PAR (www.parinc.com). Authors’ contributions Petr Novak designed the study, interpreted the results, and compiled the final manuscript.Stanislav Katina wrote the section on statistical methods, performed the statistical analysis, and interpreted the results.Petra Brandoburova was responsible for the neuropsychological assessments.Veronika Reznakova was responsible for psychiatric assessments, anthropometric measurements, and various clinical assessments.Michaela Jezberova performed the MRI analyses.Jozef Hanes performed biomarker assessments.Peter Koson and Frantisek Jurcaga were responsible for the management of the clinical evaluations and clinical MRI evaluation.Norbert Zilka orchestrated the study design and oversaw both the clinical assessments and the analyses of biomarkers.Linus Jönsson and Michal Novak contributed to the study design, oversaw the study, contributed to manuscript writing and reviewed the manuscript for intellectual content. Funding This work received funding from grants JPND grant “ADDITION: Alzheimer’s disease data-driven insights on individual outcomes of importance” and APVV grant 20–0447. Data availability Data and material are available from the corresponding authors upon reasonable request. Declarations Ethics approval and consent to participate The study was approved by the ethics committee of the Unit of the Director of the Office of the Bratislava Self-Governing Region (approval number 04838/2019/HF). The study was conducted in accordance with the Declaration of Helsinki. All participants provided written informed consent to participate. Consent for publication All participants have consented to publication of results. Competing interests The authors declare no competing interests. Footnotes Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Contributor Information Petr Novak, Email: petr.novak@savba.sk. Norbert Zilka, Email: norbert.zilka@savba.sk. References Associated Data Supplementary Materials Supplementary Material 1. Supplementary Material 2. Supplementary Material 3. Supplementary Material 4. Data Availability Statement Data and material are available from the corresponding authors upon reasonable request.