Hearing loss and dementia are among the most common chronic diseases in older age and frequently co-occur (World Health Organization [WHO], 2020, 2021; McDonough et al., 2021). By 2050 up to 900 million people will be living with hearing loss (World Health Organization [WHO], 2018) and 152.8 million with dementia (Nichols, 2022). In a German sample, dementia cases increased by 0.23% with an increasing prevalence of hearing impairment by 1 SD (Teipel et al., 2015), and an increased risk of cognitive decline was described in case of a bilateral hearing loss with a hazard ratio of 1.43 (Fritze et al., 2016). Hearing loss often remains undetected (Ramage-Morin et al., 2019). Likewise, subjective cognitive complaints do not reliably predict cognitive impairment (Edmonds et al., 2018).

But, people with hearing loss have a higher risk of dementia (Livingston et al., 2020) and cognitive impairment is associated with a higher risk of anxiety, of depression (Mirza et al., 2017 ), and of functional impairment (Brown et al., 2011), loss of independence (McLaughlin et al., 2010), institutionalization (Eska et al., 2013), delirium (Tsui et al., 2022) and mortality (Diwell et al., 2018). Therefore it is an imperative to reliably detect cases of cognitive impairment. Due to the high comorbidity of hearing loss and cognitive impairment (Lin et al., 2013; Teipel et al., 2015; Fritze et al., 2016; Huang et al., 2023; Tamblay et al., 2023), cognitive screening has gained increasing awareness beyond geriatric or psychiatric settings (Mordenfeld et al., 2020; Susano et al., 2020; Canavan and O’Donnell, 2022; Jammula et al., 2022; Tsui et al., 2022) including ENT departments and audiology (Mosnier et al., 2018; Humes, 2020; Sanders et al., 2021; Völter et al., 2022).

Therefore, the development and evaluation of tests for cognitive impairment for people with hearing loss are required to enable early detection as well as timely care and clinical intervention.

However, most cognitive screening tests are auditory-based and require good hearing. Hearing loss or simulated hearing loss leads to a false positive diagnosis of dementia or an overestimation of the actual cognitive impairment (Dupuis et al., 2015; Pye et al., 2017; Füllgrabe, 2020; Völter et al., 2020). For example, simulation of a moderate hearing loss on performance of the Mini Mental Status Examination [MMSE, a similar test to the MoCA (Folstein et al., 1975)] resulted in performance within the ‘dementia’ range among college students with normal cognition (Jorgensen et al., 2016). To mitigate the effects of hearing loss on performance, several attempts were made to adjust cognitive screening tests for people with hearing loss (Pye et al., 2017). Tests were typically adjusted by either i) deleting hearing-depended spoken items or ii) presenting the test in visual form. The problem with the former is that deleting items may have an adverse effect on the reliability of the test (Al-Yawer et al., 2019). The problem with the latter is that changing the modality of the test alters the cognitive demands of the test. One must therefore re-validate a test that has been transposed to visual format with respect to identification of cognitive impairment.

The MoCA is a commonly used, freely available cognitive screening test that has been translated into more than 100 languages (Nasreddine et al., 2005) showing a better sensitivity especially in detecting MCI than other test batteries such as the MMSE (Jia et al., 2021). Around one third of the MoCA items are spoken (Völter et al., 2020). Dawes et al. developed and validated a visually presented version of the MoCA for people with hearing loss (MoCA-H) which was translated into German by forward and backward translation as proposed by Cha et al. (2007), Völter and Götze (2021). In addition to providing written instructions and items, this adaptation also included the replacement of two auditory tasks from the original MoCA (sentence repetition and attention) by alternative items. The MoCA-H was conceived to assess the same cognitive domains with a similar level of difficulty as the original MoCA (Dawes et al., 2019).

In a validation sample of people with hearing loss, including 76 with normal cognition and 83 with dementia, a cut-off of ≤24 yielded high sensitivity and specificity (92.8 and 90.8%, respectively) (Dawes et al., 2023). One strength of the MoCA is the ability to discriminate mild cognitive impairment (MCI) from normal cognition (Nasreddine et al., 2005). Dawes et al. (2023) only established performance of the MoCA-H with respect to identification of dementia.

The performance cut-offs derived by Dawes et al. (2023) for the English MoCA-H may not be optimal for translations of the MoCA-H in other languages. Cultural or linguistic factors have been found to impact upon performance of the MoCA (Ng et al., 2018; Theocharous et al., 2023). Validation studies with the original MoCA have resulted in different estimates of accuracy and optimal cut-offs varied between different languages (O’Driscoll and Shaikh, 2017; Carson et al., 2018). In a previous study we developed normative data for the German-language MoCA-H with people with normal cognition using z-scores and taking age, education and sex into account (Völter et al., 2022). We showed that people with hearing loss performed worse than those with normal hearing on spoken items from the standard MoCA, but there was no difference in performance on the novel visually presented items from the MoCA-H.

The aim of the present study was to determine suitable cut-offs for the German MoCA-H with respect to identification of both MCI and dementia versus normal cognition. We hypothesize that the MoCA-H is a sensitive screening test for MCI and dementia, but that cut-offs and diagnostic accuracy may differ from the English version.

The present study is the first to establish performance criteria for identifying MCI or dementia among people with and without hearing loss using the German-language MoCA-H.

Recently Dawes et al. validated the English MoCA-H for dementia participants and showed a high sensitivity of 92.8% and a specificity of 90.8% at a cut-off of ≤24 points (Dawes et al., 2023). Using this cut-off point, a high sensitivity of 100% could also be obtained in the German MoCA-H in the present dementia sample. However, specificity was low (64.4%). Studies have shown that optimal cut-offs of the original MoCA vary between different languages (Carson et al., 2018), and cultural or linguistic factors may affect the performance of the MoCA (Ng et al., 2018). Therefore, the performance cut-offs derived for the English MoCA-H may not be optimal for the German translation. Further MCI patients were not included in the English study (Dawes et al., 2023). We consequently developed new cut-off values for the German sample defining cut-offs of the MoCA-H for the detection of dementia as well as for MCI.

Diagnostic accuracy of the German MoCA-H differed depending on the severity of the cognitive impairment. While a cut-off of ≤21 points was considered as optimal in the total cognitive impairment group (IC), showing a sensitivity of 87.6% and a specificity of 90%, sensitivity was significantly lower at this cut-off for the MCI sample (75.9%) and higher for the dementia group (96.3%).

Finding an appropriate cut-off score that can distinguish cognitively healthy individuals from those with mild cognitive impairment is challenging (Carson et al., 2018). While the optimal cut-off for patients with dementia was at ≤21 points in the present study with a sensitivity comparable to the results obtained by the English version, setting a cut-off for patients with MCI was much more difficult. This was already reported by others (Summers and Bondi, 2017; Thomann et al., 2020; Zhuang et al., 2021; Saunders et al., 2022).

Using the optimal cut-off of ≤22 points showed only low sensitivity (84.8%) and specificity (82.5%) in MCI, as there was a large overlap in the MoCA-H score in cognitively healthy and MCI patients. This is in line with a meta-analysis covering 9 studies worldwide on the original MoCA published by Carson et al. (Carson et al., 2018), who found an optimal cut-off of <23 points which strongly deviates from the proposed cut-off of <26 points of the original validation study (Nasreddine et al., 2005) for MCI with a sensitivity of 83% and a specificity of 88%.

To address the issue of diagnostic accuracy, Thomann et al. studying 496 outpatients in the Memory Clinic and 283 normal controls proposed two cut-offs in a range of ≤23 and ≤26 points in the auditory-based original MoCA in MCI (Thomann et al., 2020). This is in line with Yang et al., who questioned whether a single cut-off for the Chinese Beijing version of the MoCA is feasible in order to differentiate the wide spectrum of cognitive disorders and applied a range between >18 and <24 points to achieve a high discrimination rate for the diagnosis of MCI in a large study in 697 Chinese participants aged ≥60 years with a suspicion of cognitive impairment (Yang et al., 2021).

Therefore, we have also developed two cut-offs (>24 points and <22 points) with a range in between. Subjects scoring >24 points are assumed to be cognitively healthy, since at this cut-off point sensitivity is high (97.5%), i.e., only 2.5% of the MCI-subjects score better. Scoring 21 points or less means that there is a high probability that the subject is cognitively impaired, as only 10% of cognitively healthy persons score worse, with a score of ≤18 points indicating dementia and a score between 18 and 21 indicating MCI. In case subjects score in between, a more detailed examination by an experienced psychologist as well as retesting after 6 to 12 months is recommended (Thomann et al., 2020). The approach of using two separate cut-offs with an indecisive area in between might reflect the clinical reality (Saunders et al., 2022) more accurately than a single cut-off (Wong et al., 2015; Thomann et al., 2020; Yang et al., 2021).

Consultation of demographically corrected norms is recommended in the original MoCA to further increase diagnostic accuracy of the test (Nasreddine et al., 2005). In the present study, regression analysis showed a significant impact of age and education on the MoCA-H total score, but adjusting the score by these demographic factors did not significantly change the AUC as already shown by others (Thomann et al., 2020). Whether differences in the educational systems across the countries might account for this discrepancy, remains unclear. However, as sensitivity slightly increased in case education was adjusted for, we decided to adjust as proposed by Dawes et al. (2023). Our findings fit to Thomann et al. (2020), who did not observe a significant improvement in diagnostic accuracy using demographically corrected z-scores in the original MoCA, but reported on differences in the balance of sensitivity and specificity when applying demographic adjustments (Thomann et al., 2020).

Studies on the impact of gender on the original MoCA and the MoCA-H total score are contradictory. While some, including the validation study of the English MoCA-H (Dawes et al., 2023), report about an influence of gender (Konstantopoulos et al., 2016; Borland et al., 2017), the present study found an impact of age and education, but not of gender. This is in line with other normative studies on the original MoCA (Conti et al., 2015; Kopecek et al., 2017).

The presented study is the first on the German MoCA-H in dementia and MCI samples, but it also has some limitations. Although we applied exclusion criteria to exclude individuals with acute infection, delirium, severe psychiatric or brain disorders, we did not perform extensive neuropsychological assessments or magnetic resonance images in all participants. Thereby, the sample of the cognitively unimpaired might also include some people with undetected cognitive dysfunction. However, it should be mentioned that only recruitment of individuals without any kind of abnormal history might result in samples which are not representative of the general population. We believe that the cognitively healthy subjects in the present study are representative for individuals with normal cognitive findings in the clinical routine.

Furthermore, there may be a selection bias in the cognitively normal and the cognitively impaired groups. Whereas the first one was mainly recruited from out-patient settings the MCI and the dementia group consisted of patients attending a geriatric setting. Besides, regional differences in study populations should be taken into account, when applying our findings to other settings. Further, one has to keep in mind that sensitivity, specificity, and the AUC indicate the quality of the test with respect to the reference, but these parameters do not inform about the probability whether a tested person has a specific disease, as predictive values are influenced by prevalence rates. The prevalence rate of dementia and mild cognitive impairment (MCI) in clinical settings beyond geriatric departments might be lower as in the present study where the cognitively impaired patients were recruited in the geriatric clinic.

Ideally, the diagnostic accuracy of a test should be evaluated in the same setting where it is clinically applied (Thomann et al., 2020). Moreover, it is important to note that abnormal scores on a brief neuropsychological screening test like the MoCA-H are insufficient to properly diagnose dementia and require further neuropsychologic evaluation.

A non-auditory based neurocognitive screening test addresses the issue of over- or misdiagnosis of cognitive impairment in the growing number of hearing-impaired. The German MoCA-H represents a reliable screening tool for MCI and dementia among older people with hearing impairment. In the present study the diagnostic accuracy of the German-language MoCA-H was evaluated in a sample of cognitively healthy individuals as well as in those with mild cognitive impairment (MCI) and dementia. The use of a single cut-off score may be too simple, and a cut-off range should be utilized in MCI diagnostics. However, availability and accuracy regarding biomarkers especially in MCI are still limited (Zhuang et al., 2021). Mixture modeling based on a biopsychosocial perspective and including blood tests, biomarkers, or neuroimaging might improve classification accuracy in the diagnosis of MCI or dementia in the future.

The original contributions presented in the study are included in the article/supplementary material, further inquiries can be directed to the corresponding author.

The study involving human participants was reviewed and approved by Ruhr-University Bochum, Germany. The patients/participants provided their written informed consent to participate in this study.

CV and PD designed the study. HF and SF collected the data. HF did the statistical analysis with critical feedback from PD. CV and HF wrote the manuscript with contributions from all other authors. All authors contributed to the article and approved the submitted version.