Research design

All participants will be tested for hearing, vision and cognitive function (figure 1). Participants with age associated acquired hearing or vision impairment will complete adapted versions of MoCA (version 8.1) designed to account for hearing or vision impairment, respectively. Participants with normal sensory function will complete the standard version of the MoCA (version 8.1) as well as novel items for the hearing-independent and vision-independent MoCA versions. Adaptations to standard items will be designed to assess the same cognitive domain of the hearing-dependent or vision-dependent items. Adaptation of hearing-dependent items will be done by presenting written instructions and/or visual versions of items requiring spoken presentation of stimuli. Adaptation of vision-dependent items involves using spoken or tactile versions of items that are presented visually. Two novel scales will be compiled for individuals with hearing and vision impairment, respectively, using the combination of items with the optimal discriminative power to differentiate between normal cognitive function and cognitive impairment. An ideal solution would be one in which each hearing/vision sensitive item in the MoCA is replaced by a single alternative item without affecting the MoCA’s domain make-up, reliability or thresholds for determining cognitive impairment. A slightly less preferred solution is one where a single substitute is identified for each hearing/vision sensitive item, maintaining the domain structure and reliability, but thresholds for identifying cognitive impairment are different. A third and least preferred solution involves a mix of adapted items that do not replicate the existing domain structure (eg, some domains are measured with more or less items than previously) with thresholds for determining cognitive impairment that are different to the standard MoCA. The planned analysis (see below) is designed to assess, compare and select between these possible solutions.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Patient pathway through study. GPCOG, General Practitioner Assessment of Cognition; ICD-10, International Classification of Disease version 10; MoCA, Montreal Cognitive Assessment.

Participants

The study will be run in six European sites: Athens, Bordeaux, Nice, Nicosia, Bradford and Manchester. Seven hundred and ninety-two older adults and their significant others/study partner will be recruited into the study in total. The sampling frame (table 1) is designed to provide a balanced sample of participants both with and without dementia and also with and without hearing or vision impairment. One hundred and thirty-two individuals will be recruited into each of six groups (table 1). With the exception of the UK, each site will recruit 22 dyads into each of the six groups and 132 dyads in total. Due to local service limitations, Bradford in the UK will recruit only 11 from each group (n=66 in total), and Manchester will correspondingly increase recruitment of people across all six groups (n=198).

Sample size calculation

The sample size is based on achieving acceptably precise estimates of the sensitivity and specificity of the adapted tools for detecting dementia, separately for people with hearing impairment and with vision impairment in relation to the MoCA-H and MoCA-V, respectively. The sample of 264 individuals (132 with dementia and 132 without) within the MoCA-H and MoCA-V groups will enable estimation of the sensitivity to detect dementia and specificity to exclude normal cognition to within 9% of the true value (based on the exact 95% CI for a binary variable, calculated using Stata V.15). After combining across all impairment groups, sensitivity and specificity for each of the three language versions will also be estimated to within 9%.

Recruitment

Participants will be recruited from ophthalmology and audiology services, memory clinics, volunteer databases and the general community. In the UK, participants will also be recruited through the ‘Join Dementia Research’ volunteer database.20 Sites in France, Greece and Cyprus will develop their own recruitment strategies in accordance with local service provision. The member of the clinical care team at each recruitment site will provide information about the study to potential participants. Potential participants would then contact the research team to arrange participation. Participants will be given a minimum of 24 hours to decide whether they wish to participate in the study.

Consent and testing procedures

All study visits will take place at participants’ homes. At the start of the initial study visit, capacity to consent will be evaluated, and written informed consent will be obtained from both the individual participating and their study partner. All individuals taking consent will have received training in checking capacity in accordance with the legal requirements for conducting research in each country (ie, the Mental Capacity Act [2005] in the UK, the Code de la santé publique in France, Article 47 of the Hospital Law of 1992 (2071) in Greece and article 14 of Law No. 1 (I) 2005 in Cyprus21). Consent will be considered on an ongoing basis. If more than one study visit is required, willingness to continue will be discussed at the start of each visit with both the older adult and their significant other.

Following informed written consent, participants would complete the GPCOG, and study partners would complete the GPCOG informant version. Participants would then complete hearing and vision assessments before completing the MoCA. Participants in the ‘vision impairment’ groups would complete the MoCA-V, and participants in the ‘hearing impairment’ group would complete the MoCA-H. Participants with normal sensory function would complete the standard MoCA as well as the novel items from both the MoCA-V and the MoCA-H. The MoCA-H and the MoCA-V will follow standard MoCA testing procedure as closely as possible.

All data collectors in the study will be trained in Good Clinical Practice22 and will have received relevant training on the administration of the screening measures and cognitive tests. Individuals will also have received training on assessing capacity in older adults consistent with relevant local legislation at partner sites. Any individual deemed to be lacking in capacity will not be included in the study. All data transferred between sites will be encrypted, and no individual will be identifiable from the stored data. Identifiable patient information will be stored in a locked cabinet, which will only be accessible to research members at the site of the data collection. Data will be monitored as it comes in for consistency. Data integrity checks will be performed whereby 5% of all data will be checked against source documents for accuracy.

Patient and public involvement

Four research user groups (RUGs) of people with dementia with age-related hearing and/or vision impairment and their support people were established in the UK, France, Cyprus and Greece to provide advice on the research.23 Research awareness training was provided to support involvement (based on the EQUIP training package24). RUGs were consulted with respect to: (i) recruitment materials and study documentation and (2) a dissemination plan for the research. Recruitment materials were revised according to RUG feedback to improve readability. The dissemination plan included face-to-face public engagement events and YouTube video summaries following suggestions from the RUGs.

Data statement

Data will be held in the University of Manchester institutional repository. Published outputs will include a digital object identifier number, and fully anonymised data would be publicly available.

Test–retest

Five participants from each of the study groups (per language site; n=30) will be invited to perform a retest of the study measures 2–4 weeks after the initial testing. At each site, following a run-in period of 10 participants, consecutive participants will be invited to undertake a retest until the target of five has been achieved.

Assessments

Adaptation of MoCA (version 8.1) for people with acquired hearing impairment (MoCA-H)

Adaptation involved presentation of instructions and stimuli from the MoCA items in written rather than spoken format (table 2). Test-takers will be asked to read the written instructions aloud to the examiner. Research using written versions of cognitive tests has previously demonstrated similar performance to verbal versions.1 15 29 Two items—‘language’ and ‘attention to letters’—required substitution with alternative items. The ‘language’ item in the MoCA involves repetition of spoken sentences. Alternative MoCA-H ‘language’ items involve constructing sentences from a list of visually presented words. The ‘attention to letters’ item in the MoCA requires test-takers to tap their finger in response to hearing an ‘A’ in a string of letters that are read aloud. The MoCA-H substitute ‘attention to letters’ items require participants to read the numbers that are in circles as opposed to squares in a string of numbers bordered by different shapes.

Adaptation of the MoCA (version 8.1) for people with acquired vision impairments (MoCA-V)

Adaptation of the MoCA for people with vision impairment involved substitution of the first two sections of the MoCA, which rely on good vision (trail making test [TMT], copy cube, clock draw and naming task; table 2). These visually dependent items were substituted for analogous tasks in the auditory domain: visual TMT was substituted with the oral TMT.30 The clock draw task was substituted with the Verbal Clock Test.31 Both the oral TMT and the Verbal Clock Test are measures of executive function that were designed to remove confounding effects of impaired vision and motor skills on performance and have established validity and reliability. The ‘copy cube’ task was substituted with questions about the shape of a cube. The ‘naming’ task was substituted with object identification based on touch. The latter two substitutions were novel items developed by the authors.

Statistical analysis

The following describes the statistical analysis plan for development of the MoCA-H and the MoCA-V, based on replacing the hearing/vision-sensitive items with adapted items (see table 2). As a first step, each MoCA-H and MoCA-V item will be assessed for the following:

1. Discrimination: no more than 80% of participants achieving the same score.

2. Feasibility: no more than 5% missing responses.

3. Redundancy: correlations with other items >0.75.

4. Independence from hearing/vision ability: degree of association with level of hearing impairment, based on comparison between item performance between non-sensory impaired groups and hearing-impaired/vision-impaired groups.

5. Comparability between versions: where relevant, we will compare performance (% achievement) on the original MoCA item and the adapted version(s) of the item. For the novel items, this will be a within-person comparison based on the data from the non-impaired subgroups collected specifically for this purpose. For other adapted items (eg, where the adaptation involved the provision of written instructions), it will be a comparison between the appropriate non-impaired and sensory-impaired subgroups.

Substitution-based model development

The substitution of items with written rather than spoken instructions has the potential to change scores. Therefore, the substitution-based analysis will focus on the reliability and score characteristics of the overall instrument, rather than of the individual question items. We will begin by including all adapted items in the instrument scoring and examine the distribution of overall scores, reliability and optimum cognitive impairment threshold scores together with area under the curve (AUC), sensitivity and specificity (via receiver operating characteristic curve analysis). Focusing on the domains where we have multiple alternative adapted items, we will then use a stepwise ‘backwards elimination’ method to remove items from these domains one by one, in a way that maximises the AUC (as an index of overall predictive performance) without unduly affecting the tool’s reliability coefficient. Where there is no clear choice of item for removal, we will also take into account each item’s performance indices from the item analysis.

This stepwise procedure will be continued until the adapted instrument has precisely one-to-one substitution of adapted items for original items in every domain throughout, or until it is not possible to remove further items without seriously undermining the level of reliability. The performance measures for the resulting instrument will then be computed.

Exploratory-based model development

We will also conduct a purely exploratory analysis to identify a version of the MoCA-H and the MoCA-V with the highest degree of discriminative ability between people with and without dementia, regardless of domain make-up. This analysis will follow more standard ‘classical’ procedures for scale development. From the results of the item analysis, items showing good discrimination, feasibility, low redundancy, comparability and independence from hearing or vision ability will be retained. Items poor on any of these criteria will be considered for removal prior to further analysis. In the case of the MoCA-H, for example, we anticipate that the four existing MoCA hearing-sensitive items will demonstrate association with hearing impairment but will also check and, if necessary, remove additional items.

Following the removal of poorly performing items, we will apply logistic regression to identify the subset of remaining items that best predicts each participant’s cognitive status (ie, dementia/no dementia). The analysis will be based on the 264 participants with hearing/vision impairment and use a stepwise backwards elimination method for removal of items from the regression model. At the first step, all items that passed the item analysis stage will be entered as a group. At each subsequent step, the item that contributes least to the explanatory power of the model (the item with the largest p value) will be removed. This will continue until all items remaining in the model have a p value of 0.1 or lower. We use a high p value (10%) at this stage for inclusivity, prior to further assessment.

For verification we will then repeat this analysis but using stepwise entry of items in place of stepwise removal. A final selection of items will be decided through comparison of the two models: where there are differences, a final decision will be made taking account of any relevant theoretical and statistical considerations.

Comparison of models

As a final step, we will compare the resulting models from the substitution-based and exploratory approaches to constructing the MoCA-H and the MoCA-V. We will compute participant scores on each model by totalling across the correctly answered items, as per the procedure for the standard MoCA. The models will then be compared on a range of key performance indices including AUC, internal consistency (Cronbach’s alpha), test–retest reliability (intracluster reliability coefficient), sensitivity and specificity, and optimum cut-point for dementia diagnosis. A choice of the final recommended version of the MoCA-H will then be made on the basis of this comparison along with relevant clinical considerations. Assessment at participants’ homes may facilitate performance on the ‘orientation to place’ questions, reduce stress and impact on the total score. Therefore, comparability of scores would be tested with reference to existing MoCA normative data with respect to test site, age and educational level.

The result of the above analytical procedures will be finalised versions of the MoCA-H and MoCA-V instruments, in each of three languages (English, French and Greek) together with recommended threshold values for detecting dementia and measures of internal consistency and test–retest reliability.

Study start and duration

It is anticipated that data collection will start in June 2018 and run for 18 months.

Ethics and dissemination

The results of the study will be disseminated through peer-reviewed publication, conference presentations, the study website (https://www.sense-cog.eu/), the SENSE-Cog Twitter account (@sense_cog) and the MoCA test website (https://www.mocatest.org/).