Evaluation of Available Cognitive Tools Used to Measure Mild Cognitive Decline: A Scoping Review

Cognitive decline is a broad syndrome ranging from non-pathological/age-associated cognitive decline to pathological dementia. Mild cognitive impairment MCI) is defined as the stage of cognition that falls between normal ageing and dementia. Studies have found that early lifestyle interventions for MCI may delay its pathological progression. Hence, this review aims to determine the most efficient cognitive tools to discriminate mild cognitive decline in its early stages. After a systematic search of five online databases, a total of 52 different cognitive tools were identified. The performance of each tool was assessed by its psychometric properties, administration time and delivery method. The Montreal Cognitive Assessment (MoCA, n = 15), the Mini-Mental State Examination (MMSE, n = 14) and the Clock Drawing Test (CDT, n = 4) were most frequently cited in the literature. The preferable tools with all-round performance are the Six-item Cognitive Impairment Test (6CIT), MoCA (with the cut-offs of ≤24/22/19/15.5), MMSE (with the cut-off of ≤26) and the Hong Kong Brief Cognitive Test (HKBC). In addition, SAGE is recommended for a self-completed survey setting whilst a 4-point CDT is quick and easy to be added into other cognitive assessments. However, most tools were affected by age and education levels. Furthermore, optimal cut-off points need to be cautiously chosen while screening for MCI among different populations.


Introduction
Dementia is currently recognised as a global health priority, and is one of the major causes of disability amongst older adults [1,2]. Globally, there are 50 million people diagnosed with dementia, with a disease burden of AUD 1.4 trillion annually [1,2]. As the population continues to age, the worldwide prevalence of dementia is predicted to triple to 152 million people within the next three decades [3]. This will result in further costs for governments, communities, families and individuals. In addition, the medical, psychological and emotional impact on those with dementia and to caregivers/families is significant and detrimentally affects their quality of life [1].
Cognitive decline is a broad syndrome ranging from non-pathological/age-associated cognitive decline to pathological mild cognitive impairment, and further progression to dementia [4]. Mild cognitive impairment (MCI) is a term used to identify the stage of cognition that falls between normal ageing and dementia, defined as slight but measurable cognitive decline without the loss of functional ability [5][6][7]. Therefore, cognitive decline is recognised to occur through a mild and subtle manner onto a more comprehensive presentation; and its changes form a continuum [4]. Different from dementia, people with MCI can perform daily living activities independently with minimal aids or assistance [5].
instead of systematic review was chosen in order to include all the relevant information available and tools cited in the literature and to identify any gaps for future studies.

Protocol and Registration
This protocol was developed using the methodological framework for scoping reviews proposed by Arksey and O'Malley (2005) [19] and further refined by using the Preferred Reporting Items for Systematic reviews and Meta-Analyses extension for Scoping Reviews (PRISMA-ScR) Checklist [20]. The protocol for this review was registered with the Open Science Framework: https://osf.io/tb3gc/ (accessed in 1 June 2020).

Eligibility Criteria
To be included in this review, papers need to be focused on the evaluation of screening and/or diagnostic performance of cognitive tools used to measure mild cognitive decline. Peer-reviewed journal papers were included if they were: in English language, assessed general healthy adult humans (>45 years, without any diagnosed health conditions or diseases) and evaluated the psychometric performance (i.e., specificity, sensitivity, validity, reliability) of cognitive tools. All quantitative study designs were eligible for inclusion. However, reviews and grey literature were excluded. Papers were excluded if they did not meet the above specified criteria or they focused on interventions rather than performance of cognitive tools. Tools that are not easily administered or are invasive (such as imaging tools or biomarkers) were also excluded. Moreover, papers published before 2015 were excluded to provide an up-to-date review on current literature. All papers had to be easily available to the research team at the time of the study, as time was limited due to the nature of the embedded honours program of the principal researcher.

Information Sources and Search
Comprehensive literature searches for potentially relevant articles up until April 2020 were conducted in the following online databases: CINAHL (Ebsco), MEDLINE (Ovid), EMBASE (Ovid), PsycINFO (Ovid) and Cochrane. The search strategies were developed with the assistance of an experienced research librarian. The search strategy contained population, intervention and outcome terms. Searches were limited to adults aged 45 years and above as this is the age range in which mild cognitive decline presents [9]. The articles with publication dates before 2015 were excluded to provide an up-to-date review. The final search strategy for MEDLINE can be found in Supplementary Table S1. Similar search strategies were used while conducting searches in other identified databases. The final search results were exported into the EndNote X9 [21] referencing software. After removing the duplicates, the results were uploaded onto the online systematic review management system Covidence [22] for article screening purpose.

Selection of Sources of Evidence
After removing duplicates from EndNote X9 [21] and Covidence [22], 32,681 publications were available for screening ( Figure 1). Prior to screening, 3 reviewers (CTC, KS and AM) conducted screening trials and discussions on two occasions to increase consistency among reviewers. During the screening trials, CTC, KS and AM double screened 10 articles independently before discussions. After the mutual agreement of screening trial results, abstracts and titles of potentially relevant articles were single screened by CTC, KS or AM in Covidence [22]. Full-text screening and discussions as above were conducted again prior to data extraction. Relevant full-text articles (n = 444) were single screened by CTC, KS or AM against the inclusion criteria, with the reason for exclusion recorded. All included full-text papers (n = 49) underwent data extraction.

Data Charting Process and Data Items
CTC designed a standardised data-charting form (a customised spreadsheet) under supervision to chart data from eligible studies and to determine the appropriate variables to extract. The included variables in the spreadsheet were study characteristics (author, year, country of origin), characteristics of tools (name of the tool, the version of tool, range of the scores/points, cut-off point to detect mild cognitive decline, administration method and the duration of administration), study design, study population (age, %female, education level), settings, the psychometric performance of tools (including sensitivity, specificity, reliability and validity in detecting mild cognitive decline), factors that may affect the performance of the cognitive tool and the comparison standard(s) in the validation studies.
CTC charted the data in the data charting form under supervision. LMW checked the extracted data. AM hand-search the information if there was missing data in the spreadsheet. KS double-checked 10% of the extracted data. Reviewers iteratively updated the data-charting form before synthesising the results.

Synthesis of Results
By using the standardised data-charting form, all results were summarised and synthesised after discussions with all reviewers. By using the Preferred Reporting Items for

Data Charting Process and Data Items
CTC designed a standardised data-charting form (a customised spreadsheet) under supervision to chart data from eligible studies and to determine the appropriate variables to extract. The included variables in the spreadsheet were study characteristics (author, year, country of origin), characteristics of tools (name of the tool, the version of tool, range of the scores/points, cut-off point to detect mild cognitive decline, administration method and the duration of administration), study design, study population (age, %female, education level), settings, the psychometric performance of tools (including sensitivity, specificity, reliability and validity in detecting mild cognitive decline), factors that may affect the performance of the cognitive tool and the comparison standard(s) in the validation studies.
CTC charted the data in the data charting form under supervision. LMW checked the extracted data. AM hand-search the information if there was missing data in the spreadsheet. KS double-checked 10% of the extracted data. Reviewers iteratively updated the data-charting form before synthesising the results.

Synthesis of Results
By using the standardised data-charting form, all results were summarised and synthesised after discussions with all reviewers. By using the Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) flowchart, reviewers documented the screening methods and recorded the quantity of included and excluded studies in this review ( Figure 1). Additionally, by using the coding system, reviewers counted the frequency that each tool cited in included papers to catalogue which tool had the most frequent research done on its performance.
Regarding the psychometric properties, validity was charted as the Sensitivity (Sn), Specificity (Sp), Area Under the Curve (AUC), Positive Predictive Value (PPV) and Negative Predictive Value (NPV). Sn is the ability of a tool to correctly classify an individual as having 'mild cognitive decline', whereas Sp is the ability of a tool to correctly classify an individual as 'without mild cognitive decline' [23]. AUC is an overall measurement of validity performance of a screening/diagnostic test [13]. PPV is the percentage of patients with a positive test who actually have 'mild cognitive decline'; whereas NPV is the percentage of patients with a negative test who actually do not have 'mild cognitive decline' [23]. All the above properties were charted as percentages, with the closeness to 100% being higher respective validity. Reliability of a tool was identified based on its performance on all reliability tests used in the included studies. Interpretation of the above properties is presented in Table 1. By referencing with other validity studies, reviewers interpreted the psychometric properties based on the criteria developed by researchers' consensus [13,24]. To be classed as good, the cognitive tool has to achieve the below criteria: good to excellent validity, good reliability, short administration time of ≤15 min whilst being able to be self-administered or conducted by non-health care professionals [14]. Hence, reviewers assessed the performance of cognitive tools using the above appraisal format. Lastly, a narrative synthesis of results was developed to assess and evaluate the characteristics and psychometric properties of each of the identified cognitive tools based on the data charting form and the criteria (Table 1).

Study Selection
In total, 46,015 articles published in the five-year period (2015 to April 2020) were retrieved. After removing duplicate articles, 32,681 articles were screened in Covidence [22], with another 395 articles excluded due to inappropriate outcomes (n = 137), inappropriate study purpose (n = 104), inappropriate population (n = 84), papers which were unable to be retrieved (n = 25), not tools of interest (n = 23), inappropriate study design (n = 17) and duplicated articles (n = 5). After evaluating the full text, 49 articles met inclusion criteria and were included in this review.

Cognitive Tools for Mild Cognitive Decline
A total of 52 different cognitive tools used to detect cognitive decline were catalogued and assessed in this review ( Table 3).

Discussion
This scoping review collates a comprehensive list of brief cognitive tools used to measure mild cognitive decline in healthy elderly populations. To achieve effective screening outcomes, the brief cognitive tools are required to have good to excellent psychometric properties, short administration time and can be self-administered or administered by non-health care professionals [14,24].
Among these ideal screening tools, HKBC has the highest validity and reliability in identifying the earliest stages of subtle cognitive decline [27]. However, it was only validated in Hong Kong with a limited number of studies, and might not be generalisable among other populations.
MMSE is the most recognised brief cognitive tool which is frequently used in measuring cognitive impairment in clinical, research and community settings [75]. However, as supported by multiple systematic reviews and meta-analysis, MoCA can detect the subtle changes in cognitive capacity better than MMSE [14,75,76]. Studies proposed that there are several features in MoCA's design that can potentially explain its superior sensitivity in MCI detection [77]. As compared to MMSE, MoCA's assessment tasks includes more words, fewer learning trials, and a longer delay before the memory recall test [77]. MCI participants can be mildly impaired in their executive functions, complex visuospatial processing and the higher-level language abilities [77]. Thus, MoCA with more diverse and demanding tasks can better distinguish the changes in the above components than MMSE [77].
Even so, both MoCA and MMSE are recommended as the widely generalisable cognitive tools with all-round performance. They have been adapted and validated in different versions to minimise the effect of language and culture on their psychometric performance. Both tools can be administered by trained or untrained personnel in multiple health care settings such as hospital, primary care and the community. However, not all cut-off points provide high psychometric performances in screening mild cognitive decline. Different cut-off scores have also been published when the tests are modified to suit the local culture [74]. Hence, optimal cut-off points need to be carefully chosen while interpreting these results. Nonetheless, the presence of educational bias remains a concern while administering MoCA and MMSE and this was supported by a systematic review by Roshaslina Rosli et al. [74]. The impact of education may result in inappropriate referral due to the overestimation of the prevalence of mild cognitive decline [74]. To address this issue, MoCA-B is an modified version of MoCA which was designed to be less dependant on literacy levels [32]. Additional studies in this area may be beneficial for future use and development of the tools. Alternatively, Visual Cognitive Assessment Test (VCAT) is not affected by languages or cultural background, overcoming the common barriers for most cognitive tools including MoCA and MMSE [33,35]. It is designed to be a visual-based cognitive tool to reduce the language demands [35]. Only the instructions, but not the test components require translation [35]. Based on our appraisal, the only criteria resulting in its exclusion from the 'good cognitive tool' category was the slightly lengthy administration time (15 to 20 min) for a brief cognitive tool [33].
To detect mild cognitive decline in surveys, self-completed tools such as the Dementia Screening Interview (AD8), SAGE, the Everyday Memory Questionnaire (EMQ), the Cognitive Change Questionnaire (CCQ), HK-VMT and Test Your Memory (TYM) can be suitable. Among these self-administered tools, SAGE has the best validity and reliability and is also validated to be conduct via electronic devices [60]. From our review, there are some very brief cognitive tools which required less than 5 min to deliver. 6CIT is the preferable very brief cognitive tool with versatile properties [42]. However, it was only validated against MMSE which is not a true gold standard in diagnosing MCI [42]. A 4-point CDT only requires less than 2 min to conduct [50]. Its only limitation is the fair to good validity while screening MCI. Thus, CDT may be beneficial to use in combination with other screening tools without adding a significant amount of administration time. In addition, a short-form Brief Cognitive Assessment Tool (BCAT) is also valid and reliable to be conducted by professional personnel within 3 to 4 min [58].
Interestingly, the level of psychometric performance can be different while screening different types of MCI. There are generally two subtypes of MCI, which are amnestic MCI (a-MCI) and non-amnestic MCI (na-MCI) [78,79]. Research has shown that there are structural differences in brain tissues among different MCI subtypes and these pathological changes affect different cognitive components [80]. Thus, people with a-MCI have impaired memory whereas na-MCI affects people's thinking skills other than memory [78,79]. Hence, cognitive tests which assessed different domains may have different performance in identifying each MCI subtype. For instance, Short Test of Mental Status (STMS) has higher validity in discriminating na-MCI as compared to a-MCI which is potentially due to its assessment properties of having a larger domain in assessing memory rather than other cognitive skills [61,81]. Therefore, future studies are recommended to further validate the MCI screening tools' performance in discriminating different subtypes of MCI. Additional studies were also required to further validate the cut-off points and psychometric performance of the included brief cognitive tools in this review.
The limited available studies and data among included articles remains the biggest limitation to our review. The exclusion of studies before 2015, grey literature and non-English studies may limit some of the information relevant to this review. To make this review more feasible within the honours program limitation, the optional critical appraisal of study quality was not conducted in this review. Despite these limitations, this is a thorough scoping review and has collated a large number of studies from the previous 5 years. Studies from various countries were included, which allowed us to catalogue the brief cognitive tools used in worldwide populations and across a variety of settings. Substantial work was undertaken to evaluate each of the tools used in measuring mild cognitive decline.

Conclusions
Based on our review, there were 52 different tools available to discriminate mild cognitive decline among healthy elderly populations. 6CIT [42], MoCA (with the cut-offs of ≤24/22/19/15.5) [28,32,34,35,44,46,49,60], MMSE (with the cut-off of ≤26) [26][27][28]50,72] and HKBC [27] are good at discriminating the subtle cognitive changes as a result of MCI. They have versatile performance in terms of their psychometric properties, administration time and delivery methods. In addition, MoCA and MMSE have been modified into various versions to be generalisable in multiple populations. To detect subtle cognitive changes in surveys, SAGE is recommended, and it can also be administered digitally. A 4-point CDT is quick and easy to be added into other cognitive screening tests while assessing MCI. However, suitable cut-off points need to be further studied to validate performance as a mild cognitive decline screening test.
The lack of thorough evaluation of cognitive tools in identifying MCI appears to be a challenge among clinical and research settings. The aim of this review was to catalogue and assess the tools used to evaluate mild cognitive decline among healthy elderly populations, and to identify gaps in the literature which might guide future research in this area. This review advocates additional research being needed to recommend the best MCI cognitive screening tools among different populations and environments.