Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2023 Jul 17.
Published in final edited form as: Clin Neuropsychol. 2022 Jun 14;37(3):562–576. doi: 10.1080/13854046.2022.2086487

Classification statistics of the Montreal Cognitive Assessment (MoCA): Are we interpreting the MoCA correctly?

Lauren N Ratcliffe a, Taylor McDonald a, Brittany Robinson a, John R Sass b,c, David W Loring d,e, Kelsey C Hewitt d
PMCID: PMC10351673  NIHMSID: NIHMS1915792  PMID: 35699222

Abstract

Objective:

The Montreal Cognitive Assessment (MoCA) is a common cognitive screener for detecting mild cognitive impairment (MCI). However, previously suggested cutoff scores of 26/30 and above is often criticized and lacks racial diversity. The purpose of this study is to investigate the potential influence of race on MoCA classification cutoff score accuracy.

Method:

Data were obtained from the National Alzheimer’s Coordinating Center (NACC) Uniform Data Set and yielded 4,758 total participants. Participants were predominately White (82.8%) and female (61.7%) with a mean age of 69.3 years (SD = 10.3) and education level of 16.3 years (SD = 2.6). Based on NACC’s classification, participants were either cognitively normal (n = 3,650) or MCI (n = 1,108).

Results:

Sensitivity and specificity analyses revealed that when using the cutoff score of ≤26/30, the MoCA correctly classified 73.2% of White cognitively normal participants and 83.1% of White MCI participants. In contrast, this criterion correctly classified 40.5% of Black cognitively normal participants and 90.8% of Black MCI participants. Our sample was highly educated; therefore, we did not observe significant differences in scores when accounting for education across race. Classification statistics are presented.

Conclusions:

Black participants were misclassified at a higher rate than White participants when applying the ≤26/30 cutoff score. We suggest cutoff scores of ≤25/30 be applied to White persons and ≤22/30 for Black persons. These findings highlight the need for racially stratified population-based norms given the high misclassification of Black participants without such adjustment.

Keywords: Cognitive aging, cognitive screening, diagnosis and classification, MoCA, race, disparities

Introduction

The Montreal Cognitive Assessment (MoCA) is a commonly used cognitive screening test to detect mild cognitive impairment (MCI). The initial validation study resulted in a suggested cutoff score of 26/30, where scores less than 26/30 suggest cognitive impairment (McPherson & Koltai, 2018; Nasreddine et al., 2005; Rossetti et al., 2011). The standard scoring adds a one-point correction for individuals with 12 or fewer years of education (McPherson & Koltai, 2018; Rossetti et al., 2011). However, this one-point correction may adversely affect reliability in community and hospital-based samples (Bernstein et al., 2011; Gagnon et al., 2013; Ng et al., 2013). Gagnon et al. (2013) reported a decrease in sensitivity, from 80% to 69%, when applying the education correction but with a small increase in specificity, from 89% to 92%. Therefore, the loss in sensitivity can lead to an increased number of false negatives.

Different cutoff scores based on population, age, race, ethnicity, and level of education have been proposed (Cecato et al., 2016; Jeffers, 2019; Milani et al., 2018). Although many international studies exist to characterize classification accuracy in different languages and cultures, (e.g., Bosco et al., 2017; Freitas et al., 2013; Memória et al., 2013; Ng et al., 2013; Tsai et al., 2012; Zhou et al., 2015), few studies in North America have formally explored cutoff scores with population-based scores, and even fewer have explored possible differences across racial and ethnic minorities. Consequently, a range of cutoffs from ≤23/30 to ≤27/30 across national and international samples to indicate abnormal cognitive functioning have been suggested (Carson et al., 2018; Ng et al., 2013).

Racial and ethnic demographics are important influences on MoCA performance levels, as measurement bias may inflate rates of cognitive impairment among historically underserved and marginalized populations (Gianattasio et al., 2019; Milani et al., 2018). Clinically, Gianattasio et al. (2019) observed that non-Hispanic Black persons’ risk for underdiagnosed dementia were nearly doubled compared to non-Hispanic White persons. To address this, Goldstein et al. (2014) suggested a cutoff score of ≤24/30 to be more appropriate for Black persons with MCI, Milani et al. (2018) reported ≤25/30 as the optimal cutoff score in a sample of non-Hispanic White, ≤23/30 non-Hispanic Black, and ≤24/30 Hispanic persons, and Rossetti et al. (2019) suggested a cutoff score of ≤24/30. However, the small sample sizes may limit the generalizability of these findings for Black persons (Goldstein et al., 2014, n = 81; Milani et al., 2018, N = 3,895 [n = 586 non-Hispanic Black], and Rossetti et al., 2019, n = 135). If the suggested cutoff score of ≤26/30 had been used in any of these samples, it would have resulted in higher rates of misdiagnosis of MCI. While racial differences have been reported, Sachs et al. (2021) developed robust demographically-adjusted normative data for White, Black, and Hispanic participants. Sachs et al. (2021) also explored the discrimination of using a cutoff score suggested by Carson et al.’s (2018) meta-analysis and observed that a cutoff score of 23/30 classified 29.2% of participants with MCI. When using the suggested cutoff score of ≤26/30, 61.4% of their 5,338 total participants were classified as having MCI. Even with the logistical regression-based norms, Black and Hispanic participants generally scored lower on the MoCA compared to White participants, suggesting diversity influence in MoCA performance and the need for appropriately derived cutoff scores rather than a universal cutoff score.

The purpose of this study is to examine the influence of race on MoCA classification and to establish suggested MoCA cutoff scores based upon a larger representative patient group compared to previous studies. We predict that: 1) the suggested cutoff of ≤26/30 will inaccurately classify our sample, and 2) the MoCA will demonstrate poorer classification statistics when applying the suggested cutoff score to Black participants compared to White participants. We anticipate different cutoff points to reflect appropriate discrimination of cognitively normal from MCI participants across race.

Methods

Participant data were obtained from the National Alzheimer’s Coordinating Center (NACC), a longitudinal referral-based or volunteer case series with 39 Alzheimer’s Disease Research Centers (ADRCs) supported by the United States (U.S.) National Institute on Aging/NIH (NACC, 2010). The Uniform Data Set (UDS) includes participants since 2005 with a range of cognitive statuses, including cognitively normal, MCI, and dementia.

A diagnosis in the ADRC’s protocol is based on a constellation of patient data where the MoCA is only one factor in the study protocol. ADRC’s study protocol states diagnoses are made by either a consensus team or physician who conducted the evaluation. Normal cognition was indicated by 1) a global Clinical Dementia Rating (CDR) score of zero or neuropsychological testing within the normal range and 2) absence of pathologic behavior. The MCI core clinical criteria include: 1) subjective or collateral concern about a change in cognition, 2) impairment in one or more cognitive domains, and 3) largely preserved independence in functional abilities.

Data were derived from 30 ADRCs, and Version 3 (v3) of the UDS from study visits between March 2015 and September 2019. The UDS v3 neuropsychological battery consists of the MoCA, Craft Story (Craft et al., 1996), Benson Figure (Possin et al., 2011), Number Span (forward and backward), Multilingual Naming Test (MINT; Gollan et al., 2012), Word Fluency (F and L words, animals, vegetables), and Trail Making Test Part A and B (Army Individual Test Battery, 1944).

The MoCA was selected as a brief cognitive screener consisting of only 13 tasks to measure various cognitive domains. Executive functioning and visuospatial functioning (5 points) are measured via an alternation task adapted from the Trail Making B task, copy a three-dimensional cube, and draw and detail a clock, examining the contour, numbering, and hand labeling. Naming (3 points) includes a three-item confrontation naming task. Attention (6 points) includes digits forward and backward, vigilance, and serial subtraction. Language (3 points) includes sentence repetition and phonemic fluency. Memory (5 points) is assessed with a delayed recall of five nouns with two non-scorable learning trials. Lastly, temporal and spatial orientation is obtained for month, date, year, place, and city. The total MoCA score is derived by adding the points from each successfully completed task, in a range of 0 to 30 points with higher scores denoting better cognitive functioning.

MoCA scores were available for 6,943 participants at their baseline visit (see Figure 1). Exclusion criteria included non-English administration (n = 279), non-amnestic MCI (n = 137), dementia diagnosis (n = 1,393), severe impairment due to other conditions (e.g., traumatic brain injury, substance-related, schizophrenia/other psychosis; n = 19), and incomplete MoCAs due to sensory impairments (n = 21). Other non-White or non-Black races (n = 336) races were also excluded due to limited sample size (e.g., American Indian/Alaska Native [n = 72], Native Hawaiian/Pacific Islander [n = 8], Asian [n = 177], Other [n = 35], and Unknown [n = 44]). The final sample included 4,758 participants who were grouped into either cognitively normal (n = 3,650) or MCI (n = 1,108). The cognitively normal group included individuals with cognitive inefficiencies who did not meet criteria for MCI to represent normal aging. The MCI group excluded participants who met the 2011 National Institute of Aging-Alzheimer’s Association (NIA-AA) dementia criteria.

Figure 1.

Figure 1.

Open in a new tab

Flowchart on the inclusion of participants.Note: MoCA = Montreal Cognitive Assessment. naMCI = nonamnestic mild cognitive impairment. MCI = mild cognitive impairment

Results

Statistical analyses were conducted using SPSS version 27 (SPSS, Inc., Chicago, IL). Descriptive statistical analyses were performed regarding the distribution of age, sex, race, and educational level. Independent sample t-tests were used to assess differences in race and education levels across scores. Receiver operating characteristics (ROC) analysis was used to identify alternative cutoff scores and the optimum balance between sensitivity and specificity.

Study demographic characteristics are presented in Table 1. Participants were predominantly White (n = 3,940; 82.8%) and female (n = 2,935; 61.7%) with a mean age of 69.3 years (SD = 10.3; range = 18-101) and average education level of 16.3 years (SD = 2.6). Diagnosis was largely made by a consensus panel across race and diagnostic categories. White participants averaged 69.4 years of age (SD = 10.7) with an average education level of 16.5 years (SD = 2.5). Black participants averaged 69.2 years of age (SD = 8.1) with an average education level of 15.3 years (SD = 2.7). White participants (M = 16.3, SD = 2.6) had significantly more years of education than Black participants (M = 15.2, SD = 2.8); t = 11.7, p <.001, Cohen’s d = .44). Cognitively normal subjects (M = 68.2, SD = 10.5) were significantly younger than MCI participants (M = 73.2, SD = 8.3); t = −16.6, p < 0.001, Cohen’s d = −0.50).

Table 1.

Descriptive characteristics by group and total (N = 4,758).

Cognitively Normal
 MCI
 Total
White n = 3,006 (82.4%) Black n = 644 (17.6%) Total n = 3,650 White n = 934 (84.3%) Black n = 174 (15.7%) Total n = 1,108 White n = 3,940 (82.8%) Black n = 818 (17.2%) Total N = 4,758
Age
 Mean (SD) 68.2 (11.0) 68.2 (7.9) 68.2 (10.5) 73.2 (8.3) 72.7 (7.8) 73.1 (8.3) 69.4 (10.7) 69.2 (8.1) 69.3 (10.3)
 Range 18-97 44-101 18-101 29-99 54-89 29-99 18-99 44-101 18-101
Education
 Mean (SD) 16.6 (2.4) 15.4 (2.6) 16.4 (2.5) 16.5 (2.8) 15.0 (2.7) 16.2 (2.8) 16.5 (2.5) 15.3 (2.7) 16.3 (2.6)
 Range 1-26 2-25 1-26 0-26 6-20 0-26 0-26 2-25 0-26
Sex, n (%)
 Female 1874 (62.3) 497 (77.2) 2371 (65.0) 435 (46.6) 129 (74.1) 564 (50.9) 2309 (58.6) 626 (76.5) 2935 (61.7)
 Male 1132 (37.7) 147 (22.8) 1279 (35.0) 499 (53.4) 45 (25.9) 544 (49.1) 1631 (41.4) 192 (23.5) 818 (17.2)
MoCA Scores, Mean (SD)
Education-Corrected* 26.7 (2.4) 24.5 (3.1) 26.3 (2.7) 22.4 (3.3) 21.1 (3.7) 22.2 (3.4) 25.7 (3.2) 23.8 (3.5) 25.3 (3.4)
 Median 27.0 25.0 27.0 23.0 21.0 22.0 26.0 24.0 26.0
Uncorrected** 26.6 (2.5) 24.3 (3.2) 26.2 (2.8) 22.3 (3.4) 20.8 (3.8) 22.0 (3.5) 25.6 (3.3) 23.6 (3.6) 25.2 (3.4)
 Median 27.0 25.0 27.0 23.0 21.0 22.0 26.0 24.0 26.0
Diagnosis Method, n (%)
 Single Clinician 371 (12.3) 82 (12.7) 453 (12.4) 75 (8.0) 14 (8.0) 89 (8.0) 446 (11.3) 96 (11.7) 542 (11.4)
 Consensus Panel 2222 (73.9) 502 (78.0) 2724 (74.6) 715 (76.6) 146 (83.9) 861 (77.7) 2937 (74.5) 768 (79.2) 3585 (75.3)
 Other*** 413 (13.7) 60 (9.3) 473 (13.0) 144 (15.4) 14 (8.0) 158 (14.3) 557 (14.1) 74 (9.0) 631 (13.3)
Open in a new tab

Note: Total: N = 4,758. Cognitively Normal: n = 3,650. MCI: n = 1,108. Mild Cognitive impairment (MCI). Montreal Cognitive Assessment (MOCA). Standard deviation (SD).

*

MoCA total score with the 1-point education correction.

**

MoCA total raw score without the applied education point.

***

Two or more clinicians or other informal group.

Both the education-corrected and uncorrected scores were analyzed since the one-point educational correction may adversely affect the MoCA reliability. Across both White and Black cognitively normal participants, both the education-corrected (M = 26.3, SD = 2.7) and uncorrected (M = 26.2, SD = 2.8) MoCA total scores were above the suggested cutoff score. In the cognitively normal group, there was a statistically significant difference between education-corrected and uncorrected MoCA scores, t = −20.8, p < 0.001, Cohen’s d = 0.3. Although statistically significant, the difference in the mean scores (Mdifference = −0.1 points) is small as reflected by its small effect size and are not considered clinically meaningful. We used education-corrected scores in subsequent analyses.

There was a significant difference in MoCA scores for White (M = 26.7, SD = 2.4) and Black participants (M = 24.5, SD = 3.2; t = 16.9, p < 0.001, Cohen’s d = 0.9) in the cognitively normal group, and these groups also differed in years of education [White (M = 16.6, SD = 2.4), Black participants (M = 15.4, SD = 2.6); t = 10.0, p < 0.001, Cohen’s d = .5)]. On a MANOVA, there was a significant difference in White and Black participants’ MoCA scores when accounting for years of education, F = 122.88, p < 0.001, partial eta squared = 0.03.

For MCI patients, the average MoCA was below the cutoff score (M = 22.2, SD = 3.4). There was a significant difference in MoCA scores for White (M = 22.4, SD = 3.29) and Black MCI subjects (M = 21.1, SD = 3.7; t = 4.8, p < 0.001, Cohen’s d = 0.4). White (M = 16.5, SD = 2.8) and Black participants (M = 15.0, SD = 2.7) also differed in years of education; t = 6.5, p < 0.001, Cohen’s d = .54).

Though the average MoCA score for cognitively normal participants was above the suggested cutoff of 26/30 (M = 26.3, SD = 2.7), this cutoff misclassified 32.3% of cognitively normal participants and 15.8% of MCI participants when applied at the individual subject level (Table 2). This MoCA threshold had adequate sensitivity but poor specificity, respectively, 84.2% and 67.7%. Using the 23.7% base rate for MCI from this sample, the positive predictive power (PPP) was .49 and negative predictive power (NPP) was .93, resulting in a positive likelihood ratio of 2.6 times more likely to correctly classify an individual with a score below 26 as impaired. The negative likelihood ratio suggests that an individual with MCI is 0.23 times as likely to have a normal MoCA score. Overall, the MoCA accurately classified 71.6% of cognitively normal and MCI participants.

Table 2.

Classification accuracy using the suggested Montreal Cognitive Assessment (MoCA) cutoff score.

Cognitively Normal % MCI %
Total ≥26 (Normal) 2,472 67.7 175 15.8
<26 (Impaired) 1,178 32.3 933 84.2
White Participants ≥26 (Normal) 2,201 73.2 158 16.9
<26 (Impaired) 805 26.8 776 83.1
Black participants ≥26 (Normal) 271 42.1 17 9.8
<26 (Impaired) 373 57.9 157 90.2
Open in a new tab

Note: Total: N = 4,758. White participants (n = 3,940). Black participants (n = 818). Montreal Cognitive Assessment (MoCA). Mild Cognitive Impairment (MCI).

The cutoff score of ≤26/30 on the MoCA correctly classified 73.2% of White cognitively normal and 83.1% of White MCI participants (Table 2). The MoCA had adequate sensitivity and specificity, 83.1% and 73.2% respectively. With the MCI base rate of 21.3% in this sample, the PPP was .30 and NPP was .94. This yielded a positive likelihood ratio of 3.10 and a negative likelihood ratio of 0.23. Overall, the MoCA accurately classified 75.5% of White participants.

The MoCA correctly categorized only 42.1% of Black cognitively normal and 90.2% of Black MCI participants (Table 2). The MoCA had good sensitivity but poor specificity, 90.2% and 42.1% respectively. This yielded a positive likelihood ratio of 1.56 and a negative likelihood ratio of 0.23. When applying the cutoff, the MoCA accurately classified 52.3% of Black participants.

ROC analyses were performed to empirically establish MoCA cutoff scores that maximized classification rates for each racial group separately. For White participants, the Kolmogorov-Smirnov (K-S) goodness of fit derived cutoff score was 25.5 points, associated with an area under the curve (AUC) = 0.86 (CI = 0.85-0.87). This contrasts with the K-S cutoff score for Black participants, which was 22.5 points (AUC = 0.76, CI = 0.72-0.81). This not only suggests different MoCA thresholds across racial groups but also reflects different classification accuracy when optimized thresholds are applied based upon the 0.10 AUC racial difference. This is reflected by the 3-point group K-S difference, 1-point greater than the average group difference (White = 25.6, SD = 3.3; Black = 23.6, SD = 3.6).

Sensitivity, specificity, and accuracy were calculated to assess the diagnostic accuracy of the MoCA at various cutoff scores. Table 3 shows these values for White and Black participants. The cutoff for White participants is a 3-point difference from the cutoff for Black participants. In White participants, a cutoff score of ≤25/30 on the MoCA balanced modest sensitivity (72%) and specificity (84%). A more liberal cutoff score of ≤24/30 produced poor sensitivity (60%) and excellent specificity (90%). In Black participants, a cutoff score of ≤22/30 had poor sensitivity (55%) but good specificity (84%). A liberal cutoff score of ≤21/30 had poor sensitivity (40%) but good specificity (89%). However, an appreciable change in accuracy was not observed with the more liberal scores for both White and Black participants.

Table 3.

Operational characteristics of Montreal Cognitive Assessment (MoCA) scores by race.

Participant Race Cutoff Sensitivity Specificity Correctly Classified Cases, % PPP NPP Positive Likelihood Ratio Negative Likelihood Ratio
White  ≤26  .83  .73  75.5% .49 .93  3.1  .23
Black  .90  .42  52.3% .30 .94  1.56  .23
White  ≤25  .72  .84  80.8% .58 .91  4.36  .34
Black  .78  .54  59.3% .32 .90  1.71  .40
White  ≤24  .60  .90  83.1% .66 .89  6.21  .83
Black  .73  .69  70.8% .39 .90  2.37  .39
White  ≤23  .48  .94  83.3% .72 .85  8.41  .55
Black  .66  .78  75.6% .45 .90  3.02  .43
White  ≤22  .34  .97  81.9% .76 .83  10.06  .68
Black  .55  .84  77.6% .48 .87  3.38  .78
White  ≤21  .25  .98  80.9% .81 .81  13.68  .76
Black  .40  .89  78.9% .50 .85  3.75  .67
Open in a new tab

Note: White participants (n = 3,940). Black participants (n = 818). Montreal Cognitive Assessment (MoCA). Positive Predictive Power (PPP). Negative Predictive Power (NPP).

Base Rate for MCI in White participants is 23.7% and 21.3% in Black participants.

Discussion

This study demonstrates differential race classification accuracy when using the ≤26/30 MoCA threshold for normal aging. Cognitively normal participants’ mean scores were at the ≤26/30 cutoff. However, when scores were stratified by race, the MoCA only correctly classified 67.7% of cognitively normal participants. The MoCA only accurately categorized 75.5% of the White participants and 52.3% of the Black participants. In contrast to Gagnon and colleagues’ findings (2013), there were no appreciable differences in total MoCA scores when comparing education-corrected to uncorrected MoCA scores; the discrepancy in findings may be reflective of sample size, as well as the limited distribution of education in this highly educated sample (Gagnon et al., N = 101; this study, N = 4,758). Additionally, there was statistical significance when evaluating MoCA scores based on race and education level; however, when we accounted for these differences, they were relatively small. Again, the differences in education level by race and across groups was relatively small (Mdifference = 1.2 years) and may not have been clinically meaningful, particularly in the context of the large variability in characterizing educational level (e.g., Battistin et al., 2014). Nevertheless, this is an important area to continue to investigate in future prospective studies.

These findings highlight the benefit of race-derived norms when interpreting the MoCA. Although our optimal cutoff score (White= ≤25/30; Black= ≤22/30) had modest to low sensitivity (72% and 55%), they had good specificity (84% and 84%), similar to other reports (Goldstein et al., 2014; Rossetti et al., 2017, 2019; Sink et al., 2015). Applying our race-derived cutoff scores to Sachs et al.’s (2021) median score would result in group classification for both White (≤25/30) and Black (≤23/30) participants as being cognitively normal. Since our findings are consistent with Sachs et al. (2021), we believe that our suggested cutoff scores should be generalizable. It is at the clinician’s discretion as to which cutoff score is most appropriate for their population of interest, particularly when considering demographically-matched characteristics; however, given our larger sample size of Black participants (n = 818), we believe that our suggested cutoff score of ≤22/30 is more robust compared to Goldstein et al. (2014) and Rossetti et al. (2019).

Our rate of overdiagnosis was relatively low in Black participants, which is consistent with findings reported by Gianattasio et al. (2019). However, misdiagnosis may adversely affect quality of life and needlessly misdirect health care resources. Racial differences in cognitive test performance may result from differences in socioeconomic status, health, quality of education, lifetime stress, and educational attainment among racial groups (Brewster et al., 2019; Driscoll & Shaikh, 2017; Sink et al., 2015; Zuelsdorff et al., 2020).

Racial differences may also suggest specific MoCA content that may be culturally biased or have poor classification statistics across studies, settings, and targeted populations (Nasreddine et al., 2012). McDonald et al. (2022) conducted an item level analysis in 3,560 cognitively normal White and Black participants and observed significant racial differences on most MoCA subtests except for the clock contour, delayed recall category cue, delayed recall recognition, and orientation domain. Modes revealed worse performance on the cube copy and delayed recall (no cue) for Black participants compared to White participants. Although the results were statistically significant, the difference in scores was small (Mdifference < 0.04 points). They concluded that the items were not culturally biased, but the differences were reflective of poor psychometric properties, a general limitation of many cognitive screeners (Molnar et al., 2020). Similarly, Milani et al. (2019) found that some subtests (e.g., serial 7 s, abstraction, delayed recall, and clock contour and hands) had higher discrimination and more diagnostic utility compared to other subtests for non-Hispanic Black and Hispanic participants. Therefore, future studies are warranted to replicate these findings and expand to include other races/ethnicities and cultural factors.

The Mini Mental State Examination (MMSE), another commonly used cognitive screener, has diminished diagnostic utility in ethnic minorities (Rossetti et al., 2017). Tappen et al. (2012) suggested that the combination of screening measures, such as the MMSE and Functional Activities Questionnaire, may reduce cultural bias when compared to the MMSE alone. These findings have been supported by Gonzalez et al. (2021) with improvement in diagnostic accuracy and equity when combining a cognitive screener with a functional status assessment. Overall, the evaluation of racial and cultural influences on cognitive screeners is largely understudied (Driscoll & Shaikh, 2017; Rossetti et al., 2017; Sink et al., 2015).

Accuracy and sensitivity have generally been found to be low in cognitive screeners – a finding that is supported by the results of this study. Inadequate classification accuracy can cause false positives leading to over-diagnosis and psychological distress (Roebuck-Spencer et al., 2017). Similarly, false negatives may result in missed opportunities for early intervention. The poor sensitivity observed in this study highlights that the use of the MoCA alone would miss 15.8% of participants with true cognitive impairment. Some factors that alter classification accuracy include disease severity, setting, education, and adopted cutoffs (Molnar et al., 2020). Sensitivity, specificity, predictive values, and likelihood ratios are not fixed and vary across studies as a result of the aforementioned factors. In clinical trials, high specificity is preferred over high sensitivity in order to decrease the chance of false positive errors (Dodge et al., 2020); therefore, it will be important for these cutoff scores to be reassessed in other samples.

The racial composition for the initial validation study (Nasreddine et al., 2005) is unclear; however, the significant difference in scores between racial groups found in the present study suggests that researchers should develop normative data derived by race. These findings further suggest that a single cutoff is inappropriate for Black individuals (Milani et al., 2018), and that using an inappropriate cutoff score for Black individuals, based on a general cutoff score, could inflate false positive error rates. Therefore, the development of race-derived norms may better facilitate accurate classification for individuals who may need supplemental services.

While there are limitations in the existing normative data across neuropsychological measures, the concept of race-derived norms should not be dismissed, as suggested by National Football League (NFL) 2021 plaintiffs. Rather, the field of neuropsychology, in conjunction with the American Academy of Clinical Neuropsychology (AACN) Relevance 2050 Initiative, advocates for updated and appropriate multicultural and multilinguistic norms (Postal, 2018). While the position paper by AACN (2021) suggests the removal of race as a variable with demographically-based normative test interpretations, when not accounting for demographic variables, this study demonstrated significant misclassification among Black participants. Thus, emphasizing the importance for relevant cutoff scores and normative data to consider demographic influence on performance.

Limitations

The present study is not without limitations. Our subjects had significantly higher educational attainment compared to the national average. According to the U.S. Census, only 32.1% of the U.S. population obtained ≥16 years of education (McElrath & Martin, 2021), whereas in the present study, 71.3% of participants completed ≥16 years of education (White participants: 75.1% and Black participants: 52.3%). This discrepancy can likely be attributed, at least in part, to the observed trend of individuals with greater educational attainment being more likely to participate in clinical trials than their less educated counterparts overall (Jeffers, 2019).

Age was also found to be a potential confounding variable, which may reflect disease process and differences in cognitive performance between groups. This finding is consistent Sachs et al. (2021) who also observed a decline in MoCA performance with increasing age. Another study limitation is the proportional lack of racial minority participants. Although the inclusion of multiple racial groups in the NACC database is unique, the number of non-White and non-Black participants is relatively small, making it inappropriate to generalize these results to less represented racial groups (Brewster et al., 2019; Milani et al., 2018).

Other study limitations involve potential biases resulting from study design relying on the existing NACC dataset. Since most participants were referral-based or volunteer-based, there is potential for selection bias, especially given that some ADRCs require consent to autopsy prior to their baseline visit (Milani et al., 2018). Prospective participants may opt out of research due to this requirement, which likely also contributes to the low rates of racial minorities within the sample. The timing of MoCA administration is another potential source of bias since the MoCA was administered prior to clinician evaluation, which may have influenced clinicians’ conceptualizations (Milani et al., 2018). While the CDR is supposed to be scored independently from the other neuropsychological tests, clinicians may use the cognitive test scores (e.g., MoCA) to guide CDR interpretation (Dodge et al., 2020). Specifically, this pre-conceptualization of the participant and unknown emphasis placed on the MoCA for informing diagnosis may result in some degree of criterion contamination; therefore, classification statistics may partially reflect the influence of the MoCA in classifying the participant rather than the equal contribution of diagnostic criteria, although this cannot fully explain the group differences between Black and White participants.

While the NACC includes longitudinal data spanning several years, test-retest reliabilities for the MoCA range from 0.6 to 0.8, which may not diminish the influence of racial bias across multiple timepoints (Bruijnen et al., 2020; Cooley et al., 2015). The conjunction of these limitations could have reduced the MoCA sensitivity and specificity demonstrated in the current study. However, it will be important to replicate these findings with other diagnostic methods.

Future directions

Future studies are warranted in developing normative data across diverse demographic populations to facilitate accurate clinical MoCA classification and interpretation in these groups. Furthermore, research could replicate and expand upon the robust demographically-adjusted normative data developed by Sachs et al. (2021). Demographically-adjusted normative data may help decrease potential systematic bias in diagnostic decision making when using the MoCA, particularly among racial and ethnic minorities; therefore, future studies are warranted in expanding the normative data in other minority populations (e.g., Asian/Pacific Islander, Indigenous American, etc.) Demographically corrected T-scores could enhance diagnostic accuracy and interpretation. The qualitative descriptors could mimic those used by either Heaton (Heaton et al., 2004), Q-Simple (Schoenberg & Rum, 2017), or the Uniform Labeling System (Guilmette et al., 2020) proposed by the AACN, as these are readily understood by referring clinicians and may help to reduce communication errors (Schoenberg et al., 2018). While we acknowledge that these aforementioned directions are not solutions, per se, but rather minimize racial disparities, future studies should focus on advancing psychometric techniques to address measurement invariance.

Conclusion

Black participants are misclassified as MCI to a greater degree than White participants when applying the MoCA cutoff of ≤26/30. An optimal classification cutoff is ≤25/30 points for White participants and ≤22/30 points for Black participants based on study results. Although cutoff scores are helpful for identifying an individual who may require a more comprehensive evaluation, race-corrected normative scores may increase clinician confidence, thus contributing to reduced misclassification at time of screening.

Funding

The NACC database is funded by NIA/NIH Grant U24 AG072122. NACC data are contributed by the NIA-funded ADRCs: P30 AG019610 (PI Eric Reiman, MD), P30 AG013846 (PI Neil Kowall, MD), P50 AG008702 (PI Scott Small, MD), P50 AG025688 (PI Allan Levey, MD, PhD), P50 AG047266 (PI Todd Golde, MD, PhD), P30 AG010133 (PI Andrew Saykin, PsyD), P50 AG005146 (PI Marilyn Albert, PhD), P50 AG005134 (PI Bradley Hyman, MD, PhD), P50 AG016574 (PI Ronald Petersen, MD, PhD), P50 AG005138 (PI Mary Sano, PhD), P30 AG008051 (PI Thomas Wisniewski, MD), P30 AG013854 (PI Robert Vassar, PhD), P30 AG008017 (PI Jeffrey Kaye, MD), P30 AG010161 (PI David Bennett, MD), P50 AG047366 (PI Victor Henderson, MD, MS), P30 AG010129 (PI Charles DeCarli, MD), P50 AG016573 (PI Frank LaFerla, PhD), P50 AG005131 (PI James Brewer, MD, PhD), P50 AG023501 (PI Bruce Miller, MD), P30 AG035982 (PI Russell Swerdlow, MD), P30 AG028383 (PI Linda Van Eldik, PhD), P30 AG053760 (PI Henry Paulson, MD, PhD), P30 AG010124 (PI John Trojanowski, MD, PhD), P50 AG005133 (PI Oscar Lopez, MD), P50 AG005142 (PI Helena Chui, MD), P30 AG012300 (PI Roger Rosenberg, MD), P30 AG049638 (PI Suzanne Craft, PhD), P50 AG005136 (PI Thomas Grabowski, MD), P50 AG033514 (PI Sanjay Asthana, MD, FRCP), P50 AG005681 (PI John Morris, MD), P50 AG047270 (PI Stephen Strittmatter, MD, PhD).

Footnotes

Disclosure statement

No potential conflict of interest was reported by the authors.

References

  1. American Academy of Clinical Neuropsychology (AACN). (2021). Relevance 2050 Subcommittee on Use of Race in Neuropsychological test Norming and Performance Prediction. Position statement on use of race as a factor in neuropsychology test norming and performance prediction.
  2. Army Individual Test Battery. (1944). Manual of directions and scoring. War Department, Adjutant Generals’ Office. [Google Scholar]
  3. Battistin E, De Nadai M, & Sianesi B (2014). Misreported schooling, multiple measures and returns to educational qualifications. Journal of Econometrics, 181(2), 136–150. 10.1016/j.jeconom.2014.03.002 [DOI] [Google Scholar]
  4. Bernstein I, Lacritz L, Barlow C, Weiner M, & DeFina L (2011). Psychometric evaluation of the Montreal Cognitive Assessment (MoCA) in three diverse samples. The Clinical Neuropsychologist, 25(1), 119–126. 10.1080/13854046.2010.533196 [DOI] [PubMed] [Google Scholar]
  5. Bosco A, Spano G, Caffò AO, Lopez A, Grattagliano I, Saracino G, Pinto K, Hoogeveen F, & Lancioni GE (2017). Italians do it worse. Montreal Cognitive Assessment (MoCA) optimal cut-off scores for people with probable Alzheimer’s disease and with probable cognitive impairment. Aging Clinical and Experimental Research, 29(6), 1113–1120. 10.1007/s40520-017-0727-6 [DOI] [PubMed] [Google Scholar]
  6. Brewster P, Barnes L, Haan M, Johnson JK, Manly JJ, Nápoles AM, Whitmer RA, Carvajal-Carmona L, Early D, Farias S, Mayeda ER, Melrose R, Meyer OL, Zeki Al Hazzouri A, Hinton L, & Mungas D (2019). Progress and future challenges in aging and diversity research in the United States. Alzheimer’s & Dementia : The Journal of the Alzheimer’s Association, 15(7), 995–1003. [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Bruijnen CJWH, Dijkstra BAG, Walvoort SJW, Budy MJJ, Beurmanjer H, De Jong CAJ, & Kessels RPC (2020). Psychometric properties of the Montreal Cognitive Assessment (MoCA) in healthy participants aged 18-70. International Journal of Psychiatry in Clinical Practice, 24(3), 293–300. 10.1080/13651501.2020.1746348 [DOI] [PubMed] [Google Scholar]
  8. Carson N, Leach L, & Murphy K (2018). A re-examination of Montreal Cognitive Assessment (MoCA) cutoff scores. International Journal of Geriatric Psychiatry, 33(2), 379–388. 10.1002/gps.4756 [DOI] [PubMed] [Google Scholar]
  9. Cecato J, Martinelli J, Izbicki R, Yassuda M, & Aprahamian I (2015). A subtest analysis of the Montreal Cognitive Assessment (MoCA): Which subtests can best discriminate between healthy controls, mild cognitive impairment and Alzheimer’s disease. International Psychogeriatrics, 28(5), 825–832. 10.1017/S1041610215001982 [DOI] [PubMed] [Google Scholar]
  10. Cooley S, Heaps J, Bolzenius J, Salminen L, Baker L, Scott S, & Paul R (2015). Longitudinal change in performance on the Montreal Cognitive Assessment in older adults. The Clinical Neuropsychologist, 29(6), 824–835. 10.1080/13854046.2015.1087596 [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Craft S, Newcomer J, Kanne S, Dagogo-Jack S, Cryer P, Sheline Y, Luby J, Dagogo-Jack A , & Alderson A (1996). Memory improvement following induced hyperinsulinemia in Alzheimer’s disease. Neurobiology of Aging, 17(1), 123–130. 10.1016/0197-4580(95)02002-0 [DOI] [PubMed] [Google Scholar]
  12. Dodge H, Goldstein F, Wakim N, Gefen T, Teylan M, Chan K, Kukull W, Barnes L, Giordani B , Hughes T, Kramer J, Loewenstein D, Marson D, Mungas D, Mattek N, Sachs B, Salmon D, Willis-Parker M, Welsh-Bohmer K, … Weintraub S (2020). Differentiating among stages of cognitive impairment in aging: Version 3 of the Uniform Data Set (UDS) neuropsychological test battery and MoCA index scores. Translational Research & Clinical Interventions, 7(1), 1–21. 10.1002/trc2.12103 [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Driscoll C, & Shaikh M (2017). Cross-cultural applicability of the Montreal Cognitive Assessment (MoCA): A systematic review. Journal of Alzheimer’s Disease : JAD, 58(3), 789–801. 10.3233/JAD-161042 [DOI] [PubMed] [Google Scholar]
  14. Freitas S, Simões M, Alves L, Duro D, & Santana I (2013). Montreal Cognitive Assessment: Validation study for mild cognitive impairment and Alzheimer disease. Alzheimer Disease and Associated Disorders, 27(1), 37–43. 10.1097/WAD.0b013e3182420bfe [DOI] [PubMed] [Google Scholar]
  15. Gagnon G, Hansen K, Woolmore-Goodwin S, Gutmanis I, Wells J, Borrie M, & Fogarty J (2013). Correcting the MoCA for education: Effect on sensitivity. The Canadian Journal of Neurological Sciences. Le Journal Canadien Des Sciences Neurologiques, 40(5), 678–683. [DOI] [PubMed] [Google Scholar]
  16. Gianattasio K, Prather C, Glymour M, Ciarleglio A, & Power M (2019). Racial disparities and temporal trends in dementia misdiagnosis risk in the United States. Alzheimer’s & Dementia (New York, N. Y.), 5(1), 891–898. 10.1016/j.trci.2019.11.008 [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Goldstein F, Ashley A, Miller E, Alexeeva O, Zanders L, & King V (2014). Validity of the Montreal Cognitive Assessment as a screen for mild cognitive impairment and dementia in African Americans. Journal of Geriatric Psychiatry and Neurology, 27(3), 199–203. [DOI] [PubMed] [Google Scholar]
  18. Gollan TH, Weissberger GH, Runnqvist E, Montoya RI, & Cera CM (2012). Self-ratings of spoken language dominance: A Multi-Lingual Naming Test (MINT) and Preliminary Norms for Young and Aging Spanish-English Bilinguals. Bilingualism (Cambridge, England), 15(3), 594–615. 10.1017/S1366728911000332 [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. González D, Gonzales M, Jennette K, Soble J, & Fongang B (2021). Cognitive screening with functional assessment improves diagnostic accuracy and attenuates bias. Alzheimer’s & Dementia, 13, 1–10. 10.1002/dad2.12250 [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Guilmette TJ, Sweet JJ, Hebben N, Koltai D, Mahone EM, Spiegler BJ, Stucky K, Westerveld M, & Conference Participants. (2020). American Academy of Clinical Neuropsychology consensus conference statement on uniform labeling of performance test scores. The Clinical Neuropsychologist, 34(3), 437–453. 10.1080/13854046.2020.1722244 [DOI] [PubMed] [Google Scholar]
  21. Heaton R, Miller S, Taylor M, & Grant I (2004). Revised comprehensive norms for an expanded Halstead-Reitan Battery: Demographically adjusted neuropsychological norms for African American and Caucasian adults. PAR. [Google Scholar]
  22. Jeffers S (2019). Refining brief assessment tools for Mild Cognitive Impairment: Increasing accuracy on the Montreal Cognitive Assessment for United States Populations [Doctoral dissertation, Howard University, Department of Psychology; ]. ProQuest Dissertations & Theses A&I; ProQuest Dissertations & Theses Global. https://search.proquest.com/docview/2307191291?accountid=12381 [Google Scholar]
  23. McDonald T, Ratcliffe L, & Sass J (2022). Is the MoCA culturally biased? [Paper presentation]. Poster Presented at the 50th Annual Meeting of the International Neuropsychological Society, New Orleans, LA. [Google Scholar]
  24. McElrath K, & Martin M (2021, February). Bachelor’s degree attainment in the United States: 2005 to 2019. United States Census Bureau. https://www.census.gov/content/dam/Census/library/publications/2021/acs/acsbr-009.pdf [Google Scholar]
  25. McPherson S, & Koltai D (2018). A practical guide to geriatric neuropsychology. Oxford University Press. [Google Scholar]
  26. Memória CM, Yassuda MS, Nakano EY, & Forlenza OV (2013). Brief screening for mild cognitive impairment: Validation of the Brazilian version of the Montreal Cognitive Assessment. International Journal of Geriatric Psychiatry, 28(1), 34–40. 10.1002/gps.3787 [DOI] [PubMed] [Google Scholar]
  27. Milani S, Marsiske M, & Striley C (2019). Discriminative ability of Montreal Cognitive Assessment subtests and items in racial and ethnic minority groups. Alzheimer Disease and Associated Disorders, 33(3), 226–232. 10.1097/WAD.0000000000000310 [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Milani S, Marsiske M, Cottler L, Chen X, & Striley C (2018). Optimal cutoffs for the Montreal Cognitive Assessment vary by race and ethnicity. Alzheimer’s & Dementia (Amsterdam, Netherlands), 10, 773–781. 10.1016/j.dadm.2018.09.003 [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Molnar FJ, Benjamin S, Hawkins SA, Briscoe M, & Ehsan S (2020). One size does not fit all: Choosing practical cognitive screening tools for your practice. Journal of the American Geriatrics Society, 68(10), 2207–2213. 10.1111/jgs.16713 [DOI] [PubMed] [Google Scholar]
  30. Nasreddine ZS, Phillips NA, Bédirian V, Charbonneau S, Whitehead V, Collin I, Cummings JL, & Chertkow H (2005). The Montreal Cognitive Assessment, MoCA: A brief screening tool for mild cognitive impairment. Journal of the American Geriatrics Society, 53(4), 695–699. 10.1111/j.1532-5415.2005.53221.x [DOI] [PubMed] [Google Scholar]
  31. Nasreddine Z, Rossetti H, Phillips N, Chertkow H, Lacritz L, Cullum M, & Weiner M (2012). Normative data for the Montreal Cognitive Assessment (MoCA) in a population-based sample author response. Neurology, 78(10), 765–766. 10.1212/01.wnl.0000413072.54070.a3 [DOI] [PubMed] [Google Scholar]
  32. National Alzheimer’s Coordinating Center. (2010). The National Alzheimer’s Coordinating Center. The National Alzheimer’s Coordinating Center. https://www.alz.washington.edu/index.html [Google Scholar]
  33. Ng A, Chew I, Narasimhalu K, & Kandiah N (2013). Effectiveness of Montreal Cognitive Assessment for the diagnosis of mild cognitive impairment and mild Alzheimer’s disease in Singapore. Singapore Medical Journal, 54(11), 616–619. 10.11622/smedj.2013220 [DOI] [PubMed] [Google Scholar]
  34. Possin K, Laluz V, Alcantar O, Miller B, & Kramer J (2011). Distinct neuroanatomical substrates and cognitive mechanisms of figure copy performance in Alzheimer’s disease and behavioral variant frontotemporal dementia. Neuropsychologia, 49(1), 43–48. 10.1016/j.neuropsychologia.2010.10.026 [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. Postal K (2018). President’s annual state of the academy report. The Clinical Neuropsychologist, 32(1), 1–9. 10.1080/13854046.2017.1406993 [DOI] [PubMed] [Google Scholar]
  36. Roebuck-Spencer T, Glen T, Puente A, Denney R, Ruff R, Hostetter G, & Bianchini K (2017). Cognitive screening tests versus comprehensive neuropsychological test batteries: A National Academy of Neuropsychology education Paper†. Archives of Clinical Neuropsychology : The Official Journal of the National Academy of Neuropsychologists, 32(4), 491–498. 10.1093/arclin/acx021 [DOI] [PubMed] [Google Scholar]
  37. Rossetti H, Lacritz L, Cullum C, & Weiner M (2011). Normative data for the Montreal Cognitive Assessment (MoCA) in a population-based sample. Neurology, 77(13), 1272–1275. doi:WNL.0b013e318230208 10.1212/WNL.0b013e318230208a [DOI] [PubMed] [Google Scholar]
  38. Rossetti H, Lacritz L, Hynan L, Cullum C, Van Wright A, & Weiner M (2017). Montreal Cognitive Assessment performance among community-dwelling African Americans. Archives of Clinical Neuropsychology, 32, 238–244. [DOI] [PMC free article] [PubMed] [Google Scholar]
  39. Rossetti H, Smith E, Hynan L, Lacritz L, Cullum C, Van Wright A, & Weiner M (2019). Detection of mild cognitive impairment among community-dwelling African Americans using the Montreal Cognitive Assessment. Archives of Clinical Neuropsychology, 34(6), 809–813. 10.1093/arclin/acy091 [DOI] [PMC free article] [PubMed] [Google Scholar]
  40. Sachs B, Chelune G, Rapp S, Couto A, Willard J, Williamson J, Sink K, Coker L, Gaussoin S, Gure T, Lerner A, Nichols L, Still C, Wadley V, & Pajewski N (2021). Robust demographically-adjusted normative data for the Montreal Cognitive Assessment (MoCA): Results from the systolic blood pressure intervention trial. The Clinical Neuropsychologist, 1–16. 10.1080/13854046.2021.1967450 [DOI] [PMC free article] [PubMed] [Google Scholar]
  41. Schoenberg MR, Osborn KE, Mahone EM, Feigon M, Roth RM, & Pliskin NH (2018). Physician preferences to communicate neuropsychological results: Comparison of qualitative descriptors and a proposal to reduce communication errors. Archives of Clinical Neuropsychology : The Official Journal of the National Academy of Neuropsychologists, 33(5), 631–643. 10.1093/arclin/acx106 [DOI] [PubMed] [Google Scholar]
  42. Schoenberg M, & Rum R (2017). Towards reporting standards for neuropsychological study results: A proposal to minimize communication errors with standardized qualitative descriptors for normalized test scores. Clinical Neurology and Neurosurgery, 162, 72–79. 10.1016/j.clineuro.2017.07.010 [DOI] [PubMed] [Google Scholar]
  43. Sink KM, Craft S, Smith SC, Maldjian JA, Bowden DW, Xu J, Freedman BI, & Divers J (2015). Montreal Cognitive Assessment and modified Mini Mental State Examination in African Americans. Journal of Aging Research, 2015, 1–6. 10.1155/2015/872018 [DOI] [PMC free article] [PubMed] [Google Scholar]
  44. Tappen R, Rosselli M, & Engstrom G (2012). Use of the MC-FAQ and MMSE-FAQ in cognitive screening of older African Americans, Hispanic Americans, and European Americans. The American Journal of Geriatric Psychiatry, 20(11), 955–962. 10.1097/JGP.0b013e31825d0935 [DOI] [PubMed] [Google Scholar]
  45. Tsai C-F, Lee W-J, Wang S-J, Shia B-C, Nasreddine Z, & Fuh J-L (2012). Psychometrics of the Montreal Cognitive Assessment (MoCA) and its subscales: Validation of the Taiwanese version of the MoCA and an item response theory analysis. International Psychogeriatrics, 24(4), 651–658. 10.1017/S1041610211002298 [DOI] [PubMed] [Google Scholar]
  46. Zhou Y, Ortiz F, Nuñez C, Elashoff D, Woo E, Apostolova LG, Wolf S, Casado M, Caceres N, Panchal H, & Ringman JM (2015). Use of the MoCA in detecting early Alzheimer’s disease in a Spanish-speaking population with varied levels of education. Dementia and Geriatric Cognitive Disorders Extra, 5(1), 85–95. 10.1159/000365506 [DOI] [PMC free article] [PubMed] [Google Scholar]
  47. Zuelsdorff M, Okonkwo OC, Norton D, Barnes LL, Graham KL, Clark LR, Wyman MF, Benton SF, Gee A, Lambrou N, Johnson SC, & Gleason CE (2020). Stressful life events and racial disparities in cognition among middle-aged and older adults . Journal of Alzheimer’s Disease : JAD, 73(2), 671–682. 10.3233/JAD-190439 [DOI] [PMC free article] [PubMed] [Google Scholar]

RESOURCES