Remotely administered cognitive screens increase access to crucial health care services, facilitate continuity of care, and assist in tracking cognition over time. Remote screening formats that require high-speed internet and access to technology, however, remain inaccessible for a significant proportion of the population, especially individuals belonging to demographic minorities or socioeconomically disadvantaged groups, or those who reside in rural settings.1,2 Telephone-based cognitive screens mitigate many of these accessibility issues. The majority of telephone-based screens are modified versions of traditional in-person measures.3 When compared with the large body of research that supports the use of in-person screens to detect cognitive impairment,4 research on telephone-delivered screens has been limited, although increasing in the wake of the coronavirus disease 2019 (COVID-19) pandemic, which spurred a rapid shift to remote health care and research.3 For clinicians and researchers seeking to employ both in-person and remote screens interchangeably, it is essential to ensure that remotely administered measures are valid and reliable over follow-up to accurately assess cognitive status over time.
The Telephone-Montreal Cognitive Assessment (T-MoCA) is a cognitive screen that is the telephone-delivered analog of the Montreal Cognitive Assessment (MoCA-30), a widely used, in-clinic cognitive screen.5 The MoCA evaluates a broad range of neurocognitive domains, has been translated into numerous languages, and is the favored screen in many settings to discriminate subtle stages of cognitive decline, such as in mild cognitive impairment (MCI) versus normal cognition.4,6 Research by our team7 and others8–10 demonstrates that the T-MoCA contains adequate psychometric properties, shows strong consistency with in-person MoCA, and discriminates between MCI and normal cognition in clinical populations (including stroke,8,11 atrial fibrillation,12 and Parkinson's disease9) as well as in community-dwelling older adults.7
Studies investigating test-retest reliability of the T-MoCA and its capacity to track cognition over time are lacking. It is well-established that race/ethnicity is associated with MoCA performance13–16; however, little is known of whether there are racial/ethnic differences in changes in T-MoCA performance over time. Given that remote screeners are increasingly utilized to track cognition, there is a critical need to investigate the serial use of the T-MoCA and its utility for clinical monitoring in ethnically/racially diverse groups, controlling for factors that have been shown to be associated with MoCA including age, education, and depression.13,17–19
Systematic Review: There is a paucity of research on telephone-administered instruments used to track cognition over time. The authors carried out a PubMed search for literature on telephone-based cognitive assessment in older adults.
Interpretation: The Telephone Montreal Cognitive Assessment (T-MoCA) is not significantly different from in-person MoCA in its capacity to track cognition in demographically diverse older adults. Across five annual waves, the T-MoCA exhibited an inverted U-shaped trajectory, such that its performance improved from wave 1–3 and then subsequently decreased. Although the T-MoCA demonstrates the ability to capture cognitive decline, we contextualize our findings through the lens of test bias, practice effects, and regression to the mean, and caution the interpretation of T-MoCA scores in isolation.
Future Directions: Longer follow-up periods and larger samples, particularly for minority racial/ethnic groups, will be essential to further examine the racial/ethnic differences on repeated assessments and to evaluate the validity of the telephone versus in-person modalities.
We showed previously that baseline T-MoCA is equivalent to in-person MoCA and sensitive to differentiation of cognitively normal from MCI individuals.7 In this study, we used longitudinal data from the Einstein Aging Study (EAS), a community-residing, racially/ethnically diverse cohort of older adults to (1) examine longitudinal performance on the T-MoCA in comparison with the in-person MoCA; (2) report the test-retest reliability among repeated T-MoCA administration as measured by intraclass correlation coefficient (ICC) and compare with that of the in-person MoCA; and (3) evaluate the association of race/ethnicity with changes and test-retest reliability in T-MoCA and the correlation between T-MoCA and MoCA. These results will inform the utility of the T-MoCA to track cognitive functioning in demographically diverse community settings.
METHODS Overview of participants and proceduresAnalyses were based on data from the EAS, a longitudinal study of community-residing older adults in Bronx, New York (for an overview of recruitment methodology and study procedures, see7). Participants were ≥70 years of age, non-institutionalized, ambulatory, English speaking, and non-demented at study entry. Exclusion criteria included severe audiovisual and physical impairments, or active psychiatric symptomatology that interfered with the ability to complete assessments. The study was approved by the local institutional review board and all participants provided written informed consent at baseline assessment. Telephone interviews and in-person clinic assessments were conducted annually and included neurological and neuropsychological examinations, and characterization of demographic, medical, and other clinical factors.
Analyses were based on the longitudinal administration of the T-MoCA, MoCA-22, and MoCA-30 in 426 EAS participants who had both telephone and in-person assessments at baseline and whose race/ethnicity were non-Hispanic White (NH-W), non-Hispanic Black (NH-B), or Hispanic. Due to the COVID-19 pandemic-related stay-at-home mandates, up to three annual study waves of in-person MoCA (May 2017 to March 2020) and up to five waves of telephone assessed T-MoCA (May 2017 to March 2022) were included in the analysis.
Montreal Cognitive Assessment (MoCA)The MoCA-30 is a freely available cognitive screen (
The T-MoCA is a modified version of the MoCA-30 administered by phone.8 Just as the MoCA-22, this phone version excludes items that require visual stimuli and drawing, with the same maximum score (Table S1).
At baseline (wave 1) and annual follow ups (waves 2–5), the T-MoCA was administered before (average 23.4 days [range 1 to 106]) the MoCA-30. Alternate versions of the MoCA were utilized between formats at the same wave and across waves for the same format to reduce practice effects. Specifically, version 7.1 was administered at baseline during the phone interview, version 7.2 was administered in-person during the in-clinic visit at baseline; at follow-up for wave 2, version 7.2 was administered on the phone and version 7.1 was administered in person. This alternation continued throughout the follow-up waves.
Race/EthnicitySelf-reports of race/ethnicity were obtained and categorized based on U.S. Census Bureau 1994 designations of NH-W, NH-B, Hispanic (including White and Black), Asian, and multiracial. Due to the small number of participants in the Asian and multiracial groups (n = 8), analyses were restricted to NH-W, NH-B, and Hispanic groups (N = 426).
CovariatesDemographic information included self-reported number of years of education, sex, and age. The Geriatric Depression Scale (GDS, short form) was used to screen for depressive symptoms.20 The GDS ranges from 0 to 15, with scores of 6 or above suggestive of clinically significant depression.
Data analysesBaseline demographic and clinical characteristics were summarized by racial/ethnic groups and compared using the Kruskal-Wallis test for continuous variables and chi-square test for categorical variables.
To examine the change across annual visits in performance on the T-MoCA, MoCA-22, and MoCA-30 and association with racial/ethnic identity, we applied linear mixed-effects (LME) models, which account for the correlation between repeated measures using the participant-specific random effect. For each MoCA score, we obtained the test-retest reliability measured by ICC—that is, the ratio of the variance of the participant-specific random effect, and the total variance (sum of the variances of the participant-specific random effect and the residual), across repeated measures. Observed records of participants who were lost to follow-up or who had missing values in the telephone or in-person MoCA during follow-up (mostly the in-person assessments were missing but T-MoCA were available due to the COVID-19 pandemic) were included to avoid loss of power and potential bias from excluding incomplete records. The models can handle missing data during follow-up under the assumption that data are missing at random, which allows the missing data process to depend on the predictors and the observed outcome. A linear trend was observed and thus used for the in-person MoCA-22 and MoCA-30, which had a maximum of three time points: waves 1 (baseline), 2, and 3. Because of the overall increasing trend up to wave 3 and decreasing trend thereafter observed for T-MoCA and the convenience for comparison with the change in MoCA-22, a spline model with knot at wave 3 and piecewise linear trends was used for the longitudinal T-MoCA. To evaluate overall changes and obtain overall ICCs, the model with time trends only (Model 0) was first considered. In Model 1, race/ethnicity: NH-W (reference), NH-B, and Hispanic groups and their interaction with time at assessment were added. Covariates including age and GDS score at baseline, sex, and education were also included. To compare the changes in the longitudinal properties of the T-MoCA and MoCA-22, we applied joint models of the two longitudinal outcomes using LME models with correlated random effects. Due to the observed interaction between race/ethnicity and changes in T-MoCA, we focused on the stratified analysis by racial/ethnic groups. We fit the joint models with time trend only within each racial/ethnic group first to evaluate overall changes and obtain overall ICCs, and then further adjusted for covariates. The changes for waves 1–3 were compared between T-MoCA and MoCA-22. ICCs across repeated measures for T-MoCA and MoCA-22 were obtained and compared. The correlation between concurrent T-MoCA and MoCA-22 was also obtained. All analyses were performed using SAS statistical software version 9.4 (SAS Institute, Inc., Cary, NC).
RESULTS OverviewParticipants’ age at wave 1 ranged from 70 to 94 (mean = 78.0, SD = 5.2) years, the sample was 65.3% female, and educational achievement averaged 15.0 ± 3.4 years. The sample was 47% NH-W, 39% NH-B, and 14% Hispanic (Table 1). Among the 426 subjects at baseline, the average numbers of repeated measures were 1.8 (median = 2, Q1 = 1, Q3 = 2, range 1–3) for in-person MoCA-30, and 3.0 (median = 3, Q1 = 1, Q3 = 4, range 1–5) for the T-MoCA. As shown in Table 1, baseline distributions of age (78.6 ± 5.3 in NH-W, 78.0 ± 5.0 in NH-B, 76.1 ± 4.6 in Hispanic, p = 0.003), percent female (55% in NH-W, 76% in NH-B and 71% in Hispanic, p < .001), years of education (15.9 ± 3.5 in NH-W, 14.3 ± 3.0 in NH-B, 14.0 ± 3.6 in Hispanic, p < 0.001), T-MoCA (18.1 ± 2.8 in NH-W, 16.6 ± 2.6 in NH-B, 15.9 ± 2.9 in Hispanic, p < 0.001), MoCA-22 (18.3 ± 2.7 in NH-W, 16.6 ± 2.9 in NH-B, 16.4 ± 2.6 in Hispanic, p < 0.001), and MoCA-30 (24.6 ± 3.5 in NH-W, 22.2 ± 3.6 in NH-B, 22.1 ± 3.2 in Hispanic, p < 0.001) were significantly different among racial/ethnic groups, whereas there were no significant differences in GDS score and number of waves. The estimates of overall least-squares means across waves for T-MoCA among all (N = 426, 292,270,191,101 at waves 1–5) and each racial/ethnic group (N at waves 1–5: NH-W: 200, 143, 133, 94, 51; NH-B: 165, 114, 107, 75, 38; Hispanic: 61, 35, 30, 22, 12) are illustrated in Figure 1. The estimates of overall least-squares means across waves 1–3 for T-MoCA and MoCA-22 among all and each racial/ethnic group (N of MoCA-22 at waves 1–3: All: 426, 232,113; NH-W: 200, 115, 59; NH-B: 165, 89, 41; Hispanic: 61, 28, 13) are illustrated in Figure 2. An overall linear trend across waves 1–3 was observed for MoCA-22, and an overall increasing and decreasing trend before and after wave 3, respectively, was observed for T-MoCA.
TABLE 1 Baseline descriptive characteristics of sample by race/ethnicity.
All | NH-W | NH-B | Hispanic | ||
Mean (SD) or percentage | N = 426 | N = 200 | N = 165 | N = 61 | p-value* |
Age (in years) at time of T-MoCA | 78.0 (5.2) | 78.6 (5.3) | 78.0 (5.0) | 76.1 (4.6) | 0.003 |
Education, years | 15.0 (3.4) | 15.9 (3.5) | 14.3 (3.0) | 14.0 (3.6) | <0.001 |
Sex, % Female | 65.3% | 54.5% | 76.4% | 70.5% | <0.001 |
GDS Score | 2.3 (2.1) | 2.3 (2.0) | 2.3 (1.9) | 2.4 (2.6) | 0.857 |
MoCA-30–Standard | 23.3 (3.7) | 24.6 (3.5) | 22.2 (3.6) | 22.1 (3.2) | <0.001 |
MoCA-22 | 17.4 (2.9) | 18.3 (2.7) | 16.6 (2.9) | 16.4 (2.6) | <0.001 |
T-MoCA | 17.2 (2.9) | 18.1 (2.8) | 16.6 (2.6) | 15.9 (2.9) | <0.001 |
Abbreviations: GDS, Geriatric Depression Scale; MoCA, Montreal Cognitive Assessment; NH-B, non-Hispanic Black; NH-W, non-Hispanic White; T-MoCA, Telephone-Montreal Cognitive Assessment; SD, standard deviation.
From chi-square test for categorical variables and Kruskal-Wallis test for continuous variables.
FIGURE 1. Least-squares means of T-MoCA across annual waves 1–5. Estimates of overall least-squares means across annual waves for T-MoCA among all (N = 426, 292, 270, 191, 101 at waves 1–5) and each racial/ethnic group (N at waves 1–5: NH-W: 200, 143, 133, 94, 51; NH-B: 165, 114, 107, 75, 38; Hispanic: 61, 35, 30, 22, 12). NH-B, non-Hispanic Black; NH-W, non-Hispanic White; T-MoCA, Telephone-Montreal Cognitive Assessment.
FIGURE 2. Least-squares means across annual waves 1–3 for T-MoCA and MoCA-22. Estimates of overall least-squares means across annual waves 1–3 for T-MoCA (solid lines) and MoCA-22 (dotted lines) among all and each racial/ethnic group (N of MoCA-22 at waves 1–3: All: 426, 232, 113; NH-W: 200, 115, 59; NH-B: 165, 89, 41; Hispanic: 61, 28, 13). NH-B, non-Hispanic Black; NH-W, non-Hispanic White; T-MoCA, Telephone-Montreal Cognitive Assessment.
Table 2 shows results from separate LME models for the T-MoCA, MoCA-22, and MoCA-30. For the T-MoCA, the model examining the piecewise linear trend with knot at wave 3 (Model 0) showed an increasing trend in T-MoCA before the knot (slope = 0.33 points per annual visit, SE = 0.07, p < 0.001), followed by the slope being reduced by 0.61 (SE = 0.15, p < 0.001) after the knot, resulting in a decreasing trend (slope = −0.27, SE = 0.11, p = 0.010).
TABLE 2 Results from linear mixed-effects models for the longitudinal T-MoCA, MoCA-22, and MoCA-30 in the overall sample.a
T-MoCA | MoCA-22 | MoCA-30 | ||||||||
Modelb | Estimates | SE | p-value | Estimates | SE | p-value | Estimates | SE | p-value | |
0 | Intercept | 17.26 | 0.14 | <0.001 | 17.39 | 0.14 | <0.001 | 23.32 | 0.17 | <0.001 |
Slope up to wave 3 | 0.33 | 0.07 | <0.001 | 0.10 | 0.09 | 0.254 | 0.19 | 0.10 | 0.061 | |
Change of slope after wave 3 | −0.61 | 0.15 | <0.001 | |||||||
Slope after wave 3 | −0.27 | 0.11 | 0.010 | |||||||
ICC | 0.59 | 0.72 | 0.75 | |||||||
1 | Intercept | 18.06 | 0.26 | <0.001 | 21.61 | 1.92 | <0.001 | 28.61 | 2.40 | <0.001 |
Slope up to wave 3 among NH-W | 0.11 | 0.11 | 0.308 | 0.00 | 0.12 | 0.999 | 0.07 | 0.14 | 0.629 | |
Change of slope after wave 3 among NH-W | −0.35 | 0.22 | 0.107 | |||||||
NH-B vs NH-W at baseline | −1.46 | 0.29 | <0.001 | −1.49 | 0.28 | <0.001 | −2.02 | 0.34 | <0.001 | |
Hispanic vs NH-W at baseline | −2.27 | 0.40 | <0.001 | −1.79 | 0.38 | <0.001 | −2.28 | 0.47 | <0.001 | |
Difference in slope before wave 3 in NH-B vs NH-W | 0.26 | 0.16 | 0.093 | 0.21 | 0.18 | 0.258 | 0.32 | 0.22 | 0.143 | |
Difference in slope before wave 3 in Hispanic vs NH-W | 0.79 | 0.24 | 0.001 | 0.17 | 0.27 | 0.530 | 0.09 | 0.33 | 0.782 | |
Difference in change of slope in NH-B vs NH-W | −0.10 | 0.33 | 0.762 | |||||||
Difference in change of slope in Hispanic vs NH-W | −1.42 | 0.50 | 0.005 | |||||||
Baseline age (centered at 78) in years | −0.07 | 0.02 | 0.001 | −0.08 | 0.02 | <0.001 | −0.11 | 0.03 | <0.001 | |
Women (compared to men) | 0.89 | 0.24 | <0.001 | 0.71 | 0.25 | 0.005 | 0.60 | 0.31 | 0.057 | |
Years of education (centered at 14) | 0.15 | 0.03 | <0.001 | 0.20 | 0.04 | <0.001 | 0.32 | 0.04 | <0.001 | |
Baseline GDS | −0.27 | 0.06 | <0.001 | −0.21 | 0.06 | <0.001 | −0.22 | 0.07 | 0.003 | |
Slope after wave 3 among NH-W | −0.25 | 0.15 | 0.097 | |||||||
ICC | 0.52 | 0.65 | 0.67 |
Abbreviations: GDS, Geriatric Depression Scale; ICC, intraclass correlation coefficient; MoCA, Montreal Cognitive Assessment; NH-B, non-Hispanic Black; NH-W, non-Hispanic White; T-MoCA, Telephone-Montreal Cognitive Assessment.
From linear mixed-effects models with linear trend for longitudinal in-person (up to 3 waves of data) MoCA-22 and MoCA-30, and spline models with knot at wave 3 with piecewise linear trends for the longitudinal T-MoCA (with up to 5 waves of data).
“Slope up to wave 3” indicates the rate of change in points of T-MoCA, MoCA-22, and MoCA-30 per annual visit across waves 1, 2, and 3.
“Change of slope after wave 3” and “Slope after wave 3” indicate the difference from slope before wave 3 and the value, respectively, of rate of change after wave 3 in points of T-MoCA per annual visit.
ICC: intraclass correlation coefficient, is a measure of test-retest reliability, calculated as the ratio of the variance of the participant-specific random effect and the total variance (sum of variance of the participant-specific random effect and residual).
Model 0 is unadjusted (time trend only); model 1 further includes race/ethnicity and their interaction with time, and covariates including sex, education, baseline age, and GDS score.
In the model incorporating race/ethnicity (using NH-W as reference) and their interactions with time trends added adjusting for covariates (Model 1), significant lower performance at baseline was observed among NH-Bs (difference = −1.46, SE = 0.29, p < 0.001) and Hispanics (difference = −2.27, SE = 0.40, p < 0.001) compared to NH-Ws. Among NH-Ws, a slight and nonsignificant improvement before the knot (slope = 0.11, SE = 0.11, p = 0.308) and reduction of the slope after the knot (change = −0.35, SE = 0.22, p = 0.107) were observed, resulting in a borderline decreasing trend (slope = −0.25, SE = 0.15, p = 0.097) after the knot. Faster improvement before the knot was observed among Hispanics (difference in slope = 0.79, SE = 0.24, p = 0.001) compared to NH-Ws, and borderline improvement among NH-Bs (difference in slope = 0.26, SE = 0.16, p = 0.093); the reduction in slope after the knot was larger among Hispanics (difference = −1.42 SE = 0.50, p = 0.005) compared to NH-Ws, whereas not significantly different among NH-Bs (difference = −0.10, SE = 0.33, p = 0.762). Older age and higher GDS at baseline were associated with worse performance in T-MoCA, whereas being female and having more years of education were associated with better performance on the T-MoCA. ICCs among repeated T-MoCA were 0.59, and 0.52 for Models 0 and 1, respectively.
In-person MoCA performance trends across three wavesFor MoCA-22 and MoCA-30, which had up to three waves, slightly improving trends were observed (MoCA-22: slope = 0.10, SE = 0.09, p = 0.254; MoCA-30: slope = 0.19, SE = 0.10, p = 0.061) when examining time trends only (Model 0). In the model examining race/ethnicity and changes adjusting for covariates (Model 1), significantly lower performance at baseline was observed among NH-Bs and Hispanics compared to NH-Ws, but there was no significant difference in the rate of change among NH-Bs and Hispanics compared to NH-Ws. Similar to T-MoCA, older age and higher GDS at baseline were associated with worse performance on MoCA-22 and MoCA-30; being female and having more years of education were associated with better performance in MoCA-22 and MoCA-30. ICCs were 0.72 and 0.65 among repeated MoCA-22, and 0.75 and 0.67 among repeated MoCA-30, for Models 0 and 1, respectively.
Validating longitudinal T-MoCA performance with in-person MoCAA key subset of analyses focused solely on T-MoCA and MoCA-22, to make direct comparisons, as these measures include the same items. Results from the joint models of the T-MoCA and MoCA-22 that simultaneously examined the changes in performance of the two modalities among each racial/ethnic group are shown in Table 3. We focused on the stratified analysis because of the difference between Hispanics and NH-Ws observed in the change within the T-MoCA. Up to wave 3, mean performance on the T-MoCA increased across visits among NH-Bs (slope = 0.34, SE = 0.12, p = 0.005) and Hispanics (slope = 0.90, SE = 0.22, p < .001), whereas increasing only slightly among NH-Ws (slope = 0.11, SE = 0.10, p = 0.249). After wave 3, the slope was reduced among NH-Ws (change = −0.37, SE = 0.21, p = 0.076), NH-Bs (change = −0.47, SE = 0.26, p = 0.066), and Hispanics (change = −1.81, SE = 0.46, p < 0.001), resulting in decreasing trends after wave 3 among NH-Ws (slope = −0.25, SE = 0.14, p = 0.071), NH-Bs (slope = −0.13, SE = 0.17, p = 0.461), and Hispanics (slope = −0.90, SE = 0.32, p = 0.006). No significant change in MoCA-22, with a slight increasing trend, was observed in any racial/ethnic group over the three waves available for investigation.
TABLE 3 Results from joint model of longitudinal T-MoCA and MoCA-22 in each racial/ethnic group, time trend only.a
NH-W | NH-B | Hispanic | |||||||
Estimates | SE | p-value | Estimates | SE | p-value | Estimates | SE | p-value | |
Intercept: T-MoCA | 18.17 | 0.20 | <0.001 | 16.65 | 0.21 | <0.001 | 15.88 | 0.34 | <0.001 |
Slope up to wave 3: T-MoCA | 0.11 | 0.10 | 0.249 | 0.34 | 0.12 | 0.005 | 0.90 | 0.22 | <0.001 |
Change of slope after wave 3: T-MoCA | −0.37 | 0.21 | 0.076 | −0.47 | 0.26 | 0.066 | −1.81 | 0.46 | <0.001 |
Intercept: MoCA-22 | 18.32 | 0.19 | <0.001 | 16.62 | 0.22 | <0.001 | 16.43 | 0.32 | <0.001 |
Slope: MoCA-22 | −0.02 | 0.11 | 0.871 | 0.24 | 0.15 | 0.106 | 0.23 | 0.26 | 0.376 |
SD of random effect: T-MoCA | 2.28 | 0.14 | <0.001 | 2.07 | 0.16 | <0.001 | 1.94 | 0.24 | <0.001 |
SD of random effect: MoCA-22 | 2.30 | 0.15 | <0.001 | 2.28 | 0.17 | <0.001 | 1.98 | 0.27 | <0.001 |
Correlation between random effects for T-MoCA and MoCA-22 | 0.81 | 0.04 | <0.001 | 0.84 | 0.05 | <0.001 | 0.71 | 0.13 | <0.001 |
SD of residual: T-MoCA | 1.77 | 0.06 | <0.001 | 1.94 | 0.07 | <0.001 | 1.92 | 0.13 | <0.001 |
SD of residual: MoCA-22 | 1.39 | 0.08 | <0.001 | 1.62 | 0.10 | <0.001 | 1.61 | 0.18 | <0.001 |
slope after wave 3: T-MoCA | −0.25 | 0.14 | 0.071 | −0.13 | 0.17 | 0.461 | −0.90 | 0.32 | 0.006 |
difference at baseline between T-MoCA and MoCA-22 | −0.15 | 0.18 | 0.410 | 0.03 | 0.21 | 0.881 | −0.55 | 0.36 | 0.131 |
Difference in slope up to wave 3 between T-MoCA and MoCA-22 | 0.13 | 0.14 | 0.363 | 0.10 | 0.19 | 0.576 | 0.67 | 0.33 | 0.045 |
ICC among repeated T-MoCA | 0.63 | 0.04 | <0.001 | 0.53 | 0.04 | <0.001 | 0.50 | 0.08 | <0.001 |
ICC among repeated MoCA-22 | 0.73 | 0.04 | <0.001 | 0.66 | 0.05 | <0.001 | 0.60 | 0.10 | <0.001 |
Difference in ICC between T-MoCA and MoCA-22 | −0.11 | 0.04 | 0.015 | −0.13 | 0.06 | 0.026 | −0.10 | 0.12 | 0.404 |
Difference in SD of random effect between T-MoCA and MoCA-22 | −0.02 | 0.16 | 0.892 | −0.21 | 0.18 | 0.241 | −0.04 | 0.32 | 0.897 |
Difference in SD of residual between T-MoCA and MoCA-22 | 0.38 | 0.10 | <0.001 | 0.32 | 0.13 | 0.014 | 0.31 | 0.22 | 0.166 |
Correlation between concurrent T-MoCA and MoCA-22 | 0.55 | 0.04 | <0.001 | 0.50 | 0.04 | <0.001 | 0.39 | 0.08 | <0.001 |
Abbreviations: MoCA, Montreal Cognitive Assessment; NH-B, non-Hispanic Black; NH-W, non-Hispanic White; T-MoCA, Telephone-Montreal Cognitive Assessment; SD, standard deviation.
Linear trend for longitudinal MoCA-22 (up to three waves) and spline models with knot at wave 3 with piecewise linear trends for the longitudinal T-MoCA (up to five waves) were used.
A participant-specific random effect was included in each of the two sub-models for T-MoCA and MoCA-22, and the random effects for the two measures are correlated.
“Slope up to wave 3” indicates the rate of change in points of T-MoCA and MoCA-22 per annual visit across waves 1, 2, and 3.
“Change of slope after wave 3” and “Slope after wave 3” indicate the difference from slope before wave 3 and the value, respectively, of rate of change after wave 3 in points of T-MoCA per annual visit.
ICC: intraclass correlation coefficient, is a measure of test-retest reliability, calculated as the ratio of the variance of the participant-specific random effect and the total variance (sum of variance of the participant-specific random effect and residual).
SD of random effect: standard deviation of the participant-specific random effect.
There were no significant differences between the T-MoCA and MoCA-22 at baseline in any racial/ethnic group. The rate of changes up to wave 3 were not significantly different among NH-Ws and NH-Bs, whereas among Hispanics, the slope for T-MoCA was greater than that for MoCA-22 (difference = 0.47, SE = 0.33, p = 0.045). ICC for T-MoCA was lower than that for MoCA-22 among NH-Bs (0.53 vs 0.66, p = 0.026), NH-Ws (0.63 vs 0.73, p = 0.015), and Hispanics (0.50 vs 0.60, p = 0.404), mainly due to larger variation in the error term for the T-MoCA among NH-Ws (SD 1.77 vs 1.39, p < 0.001), NH-Bs (SD 1.94 vs 1.62, p = 0.014), and Hispanics (SD 1.92 vs 1.61, p = 0.166). Correlations between concurrent T-MoCA and MoCA-22 were 0.55, 0.50, and 0.39 among NH-Ws, NH-Bs, and Hispanics, respectively, with the difference between Hispanics and NH-Ws being borderline significant (p = 0.08). Results from the joint model further adjusting for covariates were reported in Table 4, with largely similar conclusions.
TABLE 4 Results from joint model of longitudinal T-MoCA and MoCA-22 in each racial/ethnic group, adjusting for covariates.a
NH-W | NH-B | Hispanic | |||||||
Estimates | SE | p-value | Estimates | SE | p-value | Estimates | SE | p-value | |
Intercept: T-MoCA | 18.03 | 0.33 | <0.001 | 16.49 | 0.43 | <0.001 | 15.82 | 0.56 | <0.001 |
Slope up to wave 3: T-MoCA | 0.10 | 0.10 | 0.301 | 0.34 | 0.12 | 0.006 | 0.89 | 0.22 | <0.001 |
Change of slope after wave 3: T-MoCA | −0.35 | 0.21 | 0.092 | −0.45 | 0.26 | 0.079 | −1.76 | 0.47 | <0.001 |
Baseline age (centered at 78): T-MoCA | −0.07 | 0.03 | 0.022 | −0.06 | 0.04 | 0.088 | −0.11 | 0.06 | 0.058 |
Women: T-MoCA | 1.01 | 0.33 | 0.002 | 0.70 | 0.43 | 0.105 | 0.90 | 0.57 | 0.122 |
Years of education (centered at 14): T-MoCA | 0.20 | 0.05 | <0.001 | 0.11 | 0.06 | 0.060 | 0.17 | 0.07 | 0.018 |
Baseline GDS: T-MoCA | −0.33 | 0.08 | <0.001 | −0.16 | 0.10 | 0.110 | −0.32 | 0.11 | 0.006 |
Intercept: MoCA-22 | 18.08 | 0.32 | <0.001 | 16.51 | 0.45 | <0.001 | 16.56 | 0.57 | <0.001 |
Slope: MoCA-22 | −0.02 | 0.11 | 0.875 | 0.23 | 0.15 | 0.111 | 0.18 | 0.25 | 0.469 |
Baseline age (centered at 78): MoCA-22 | −0.08 | 0.03 | 0.008 | −0.07 | 0.04 | 0.060 | −0.15 | 0.06 | 0.015 |
Women: MoCA-22 | 1.06 | 0.32 | 0.001 | 0.47 | 0.46 | 0.316 | −0.24 | 0.60 | 0.697 |
Years of education (centered at 14): MoCA-22 | 0.24 | 0.05 | <0.001 | 0.12 | 0.07 | 0.083 | 0.23 | 0.08 | 0.003 |
Baseline GDS: MoCA-22 | −0.33 | 0.08 | <0.001 | −0.08 | 0.10 | 0.428 | −0.10 | 0.11 | 0.388 |
SD of random effect: T-MoCA | 1.96 | 0.13 | <0.001 | 1.93 | 0.15 | <0.001 | 1.55 | 0.23 | <0.001 |
SD of random effect: MoCA-22 | 1.90 | 0.14 | <0.001 | 2.11 | 0.17 | <0.001 | 1.68 | 0.25 | <0.001 |
Correlation between random effects for T-MoCA and MoCA-22 | 0.75 | 0.06 | <0.001 | 0.82 | 0.06 | <0.001 | 0.67 | 0.16 | <0.001 |
SD of residual: T-MoCA | 1.76 | 0.06 | <0.001 | 1.94 | 0.07 | <0.001 | 1.93 | 0.13 | <0.001 |
SD of residual: MoCA-22 | 1.39 | 0.08 | <0.001 | 1.62 | 0.10 | <0.001 | 1.58 | 0.17 | <0.001 |
Slope after wave 3: T-MoCA | −0.25 | 0.14 | 0.078 | −0.12 | 0.17 | 0.510 | −0.87 | 0.32 | 0.008 |
Difference at baseline between T-MoCA and MoCA-22 | −0.05 | 0.32 | 0.864 | −0.02 | 0.40 | 0.955 | −0.75 | 0.62 | 0.236 |
Difference in slope up to the wave 3 between T-MoCA and MoCA-22 | 0.12 | 0.14 | 0.409 | 0.10 | 0.19 | 0.582 | 0.70 | 0.33 | 0.036 |
ICC among repeated T-MoCA | 0.55 | 0.04 | <0.001 | 0.50 | 0.05 | <0.001 | 0.39 | 0.09 | <0.001 |
ICC among repeated MoCA-22 | 0.65 | 0.05 | <0.001 | 0.63 | 0.05 | <0.001 | 0.53 | 0.11 | <0.001 |
Difference in ICC between T-MoCA and MoCA-22 | −0.10 | 0.06 | 0.075 | −0.13 | 0.06 | 0.037 | −0.14 | 0.13 | 0.281 |
Difference in SD of random effect between T-MoCA and MoCA-22 | 0.06 | 0.16 | 0.700 | −0.18 | 0.18 | 0.318 | −0.14 | 0.31 | 0.667 |
Difference in SD of residual between T-MoCA and MoCA-22 | 0.37 | 0.10 | <0.001 | 0.32 | 0.13 | 0.012 | 0.35 | 0.21 | 0.105 |
Correlation between concurrent T-MoCA and MoCA-22 | 0.45 | 0.04 | <0.001 | 0.46 | 0.05 | <0.001 | 0.30 | 0.08 | 0.001 |
Abbreviations: ICC, intraclass correlation coefficient; MoCA, Montreal Cognitive Assessment; NH-B, non-Hispanic Black; NH-W, non-Hispanic White; T-MoCA, Telephone-Montreal Cognitive Assessment; SD, standard deviation.
Linear trend for longitudinal MoCA-22 (up to three waves) and spline models with knot at wave 3 with piecewise linear trends for the longitudinal T-MoCA (up to five waves) were used. Covariates adjusted include baseline age (centered at 78), sex (men as reference), years of education (centered at 14), and baseline GDS score.
A participant-specific random effect was included in each of the two sub-models for T-MoCA and MoCA-22, and the random effects for the two measures are correlated.
“Slope up to wave 3” indicates the rate of change in points of T-MoCA and MoCA-22 per annual visit across waves 1, 2, and 3.
“Change of slope after wave 3” and “Slope after wave 3” indicate the difference from slope before wave 3 and the value, respectively, of rate of change after wave 3 in points of T-MoCA per annual visit.
ICC: intraclass correlation coefficient, is a measure of test-retest reliability, calculated as the ratio of the variance of the participant-specific random effect and the total variance (sum of variance of the participant-specific random effect and residual).
SD of random effect: standard deviation of the participant-specific random effect.
DISCUSSIONA seamless transition between in-person and remotely administered cognitive screens optimizes both clinical and research outcomes by facilitating follow-up, disease monitoring, and continuity of care. The MoCA is an extensively used cognitive screen to detect MCI and dementia in both clinical and research settings.5 A growing body of literature establishes that the telephone version of the MoCA, the T-MoCA, is a valid screening tool to remotely detect cognitive impairment without requiring the traditional face-to-face, paper-and-pencil administration.7–12,21,22 This provides clinicians, researchers, and patients/participants alike with the flexibility and convenience to implement the MoCA either in person or via telephone.
The current study builds on prior research by our team7 by examining the utility of the T-MoCA to assess cognition over longitudinal follow-up, and by exploring the racial/ethnic differences on T-MoCA performance in diverse urban community dwelling older adults (age ≥70 years) who were English speaking and non-demented at study entry. Overall, findings revealed longitudinal trends of increasing T-MoCA scores from wave 1 to wave 3, followed by a change in the slope of the trajectory, such that scores decreased from waves 3 to 5. At baseline, worse T-MoCA performance was associated with older age, elevated depression, and NH-B and Hispanic identity; better performance on the T-MoCA was associated with higher levels of education and female sex. Racial/ethnic identity was associated with distinct longitudinal T-MoCA performance patterns, such that: (1) the T-MoCA scores of both Hispanic and NH-B participants improved more over waves 1–3 than did the NH-W participants; and (2) the scores of the Hispanic participants fell more precipitously over waves 3–5 than both NH-B and NH-W participants, who declined at similar slopes.
The longitudinal trajectory of T-MoCA is determined by several factors, including practice effects, regression to the mean, and true cognitive decline. The ascending pattern up to wave 3 suggests a dominant presence of practice effect across the initial three repeated assessments during the 2-year follow-up after baseline. The descending pattern after wave 3 suggests the dominance of cognitive decline due to normative aging, neurodegenerative diseases, or other factors. The literature on practice effects on MoCA performance has been mixed, although there is a paucity of studies that present >2–3 repeated measures. For example, practice effects have been observed on MoCA performance in healthy older adults,23 whereas other studies have identified meaningful decline in MoCA scores that reflect the progression to MCI24 or mild dementia.25 Still others have suggested that stable scores should be interpreted as the absence of cognitive decline.26 A small study (N = 53) of cognitively normal older adults over three annual waves showed that improvement with repeat MoCA administration was most pronounced in individuals with the lowest initial scores, whereas higher scoring individuals remained relatively stable over follow-up.23 This regression to the mean was reflected in our findings, and may explain the results, particularly when stratified by race/ethnicity. Indeed, Hispanic and NH-B participants performed worse at baseline compared with NH-Ws, but they showed significantly improved performances across waves 1–3. In contrast, NH-W participants, who performed the highest at baseline, remained largely stable from waves 1–3. Consistent with existing literature,13–16 this suggests that the lower scores observed at baseline in NH-B and Hispanic older adults likely reflects a culture-related test bias of the MoCA, which may yield artificially reduced scores in minoritized groups that are not necessarily predictive of cognitive decline (see27 for a discussion of cross-cultural test bias). Of note, our study suggests that with repeated exposure to the MoCA, scores more closely approximate those of NH-Ws.
After wave 3, our findings showed that any effects of practice or regression to the mean reached an equilibrium, and subsequent trajectories of declining performance were observed. Decreasing T-MoCA scores occurred regardless of racial/ethnic group identity but were most pronounced in Hispanic participants. We contend that these worsening T-MoCA scores from waves 3–5 likely capture the effects of true cognitive decline and demonstrate the utility of the T-MoCA to track cognitive change. Overall, this non-linear trajectory of T-MoCA performance overtime, and distinct slopes related to racial/ethnic identity, highlights the need for a nuanced interpretation of any given T-MoCA score in isolation.
Our findings of racial/ethnic differences in performance on the T-MoCA are consistent with the established literature. MoCA scores associate with numerous demographic variables, including race/ethnicity,13–16 age,18,28 educational background and literacy,13,17,18,28,29 and psychiatric factors.19 Although literature on the T-MoCA is still emerging, research suggests that age,7,10 race/ethnicity,7,10 education,7,12,22 and gender7 can impact T-MoCA performance. In this context, caution should be used when interpreting T-MoCA scores for minority or other health disparity groups. It is essential to underscore that although NH-Bs and Hispanics achieved lower scores at baseline on the T-MoCA when compared with NH-W participants, their pattern of follow-up scores increased meaningfully with practice. That is, lower scores at baseline among these demographic groups do not confer immediate risk for cognitive decline and likely reflect test bias of the MoCA that should be carefully considered when interpreting performance in minoritized groups.
When transitioning between in-person and remotely administered screeners over longitudinal follow-up, it is essential to ensure concordance and reliability of the screen to accurately measure cognition over time. Reassuringly, we found no significant differences between the T-MoCA and the corresponding in-person MoCA, overall or among any racial/ethnic group at wave 1. Across waves 1–3, the trajectories of change in performance were not significantly different among NH-W and NH-B participants. The slope for T-MoCA from waves 1–3 was greater in the Hispanic group relative to MoCA-22, which may be driven by the relatively low sample size in the Hispanic group (n = 61). We also found that the T-MoCA is relatively less reliable than the in-person MoCA version. This reduction in reliability is unsurprising and possibly unavoidable, as the remote version is subject to more plausibly confounding variables. For example, the remote-assessment examiner is unable to ensure a distraction-free testing environment for the examinee, the examinee may be more prone to utilize unauthorized cognitive supports (i.e., taking notes on a memory test), and there is the risk for telephone administration difficulties, such as hearing difficulty. The overall observed ICC for the T-MoCA (0.59) is fair or moderate by most common criteria (e.g.,30–32). The overall observed ICCs for the in-person MoCA-22 (0.72) and MoCA-30 (0.75) are good and consistent with MoCA-30 results from prior studies. A wide range of test-retest reliability has been reported for the MoCA-30, for example, low to moderate (0.33–0.48) among the annual measurements during up to a 2-year follow-up,23 and high (0.92) between two repeated measurements ≈1 month apart.5 Although in person testing, when possible, remains the gold standard, it is likely that the many benefits of remote cognitive screening outweigh the cons to using T-MoCA when necessary.
Although telephone screeners may increase study engagement and/or clinical diagnostic screening in disadvantaged groups, our study found some suggestion of lower test-retest reliability for the T-MoCA among ethnic minority groups when compared with NH-W older adults. For example, the variation in the residual error for T-MoCA and MoCA-22 among NH-B participants was relatively higher than that among NH-W participants. We also observed weaker associations between the T-MoCA and MoCA-22 among Hispanic groups compared to NH-W groups as shown by borderline lower correlations. Of note, these racial/ethnic differences in ICCs for the T-MoCA and MoCA-22 and the correlation between T-MoCA and MoCA-22 were reduced after adjusting for covariates. Overall, these results caution that the widespread accessibility of the T-MoCA may not necessarily equate to equality of cognitive health screening for diverse, health disparity populations. Indeed, repeated administration of the T-MoCA may be less reliable in NH-B and Hispanic older adults when compared with NH-W older adults, thus increasing the risk for misdiagnosis and mishandling of clinical tracking in these racial/ethnic groups.
Our study has several noteworthy strengths. To our knowledge, it is the first to specifically investigate the utility of the T-MoCA for tracking cognition. We present up to five waves of annual longitudinal follow-up occurring over 5 years. Our study contributes meaningfully to the small but growing literature on the T-MoCA by extending the field's understanding of the change and test-retest reliability of this measure over follow-up. We carried out this study in a diverse sample of older adults, adding to the understanding of how factors such as race/ethnicity may correlate with performance on the T-MoCA.
In terms of study limitations, we employed different versions of the MoCA, which were not counterbalanced with regard to sequence of the format of administration during each data collection wave, but were counterbalanced with regard to alternate forms of each instrument. It is plausible that elimination of the visuo-constructive items in order to administer the MoCA by telephone may diminish its ability to capture cognitive change, given that visuospatial dysfunction in older adults has been found to be highly prognostic for dementia.33 We did not perform formal hearing assessments, rather we utilized informal documentation of hearing difficulties as detected by research assistants. Although our sample was quite diverse, our sample of Hispanic older adults was relatively smaller than our NH-B and NH-W groups. We also did not have a sufficient number of Asian participants to include in our analyses. Future research should recruit a larger, more racial/ethnically diverse sample and employ a longer follow-up period, possibly capturing more meaningful cognitive decline. Future work should also investigate more clinical (e.g., vascular risk factors), genetic, behavioral, social-economic, and cultural (e.g., English as first language) factors as potential confounders or modifiers, and include biomarkers of neurodegeneration to better understand the underlying etiology of any cognitive decline captured via T-MoCA. It will also be crucial for future studies to more closely explore culture-related test bias on the MoCA in an item-by-item or cognitive domain by domain fashion to better understand the utility of this cognitive screen in diverse community samples.
In sum, this study contributes to the remote cognitive assessment literature by demonstrating that the T-MoCA is a reliable telephone-based cognitive screen to track cognitive functioning over longitudinal follow-up. Consistent with prior research, individuals from racial/ethnic minorities scored lower on the T-MoCA at baseline when compared with NH-W participants. However, changes in T-MoCA were only slightly different between NH-B and NH-W participants, and there were no significant differences in changes in in-person MoCA in NH-Bs and Hispanics compared to NH-Ws. Although we found some evidence of reduced test-retest reliability of the T-MoCA in NH-B and Hispanic participants, this disparity was eliminated when analyses adjusted for covariates. Taken together, these results provide valuable information for researchers and clinicians employing the T-MoCA to measure longitudinal cognitive change in racially/ethnically diverse populations.
ACKNOWLEDGMENTSThis work was supported by the National Institutes of Health (grants NIA 2 P01 AG003949, R21 AG056920, and K23 AG063993), the Alzheimer's Association (2019-AACSF-641329), the Leonard and Sylvia Marx Foundation, and the Czap Foundation and Cure Alzheimer Fund. The authors would like to thank the dedicated Einstein Aging Study (EAS) participants for their time and effort in support of this research. This research was made possible through the hard work of EAS research assistants: Diane Sparracio and April Russo, for assistance with participant recruitment; Betty Forro, Maria Luisa Giraldi, and Sylvia Alcala, for assistance with clinical and neuropsychological assessments; and Michael Potenza for assistance with data management. All authors contributed to and approved the final manuscript.
CONFLICT OF INTEREST STATEMENTDr. Ali Ezzati serves as consultant, advisory board member, or has received honoraria from Eisai, PCORI Health Care Horizon Scanning System, GlaxoSmithKline, Mist Research, and BioDelivery Sciences International Inc, and is the founder of Neurodiction, LLC. Dr. Richard B. Lipton receives research support from the National Institutes of Health (NIH): 2PO1 AG003949 (Program Director), S&L Marx Foundation, and Czap Foundation, the US Food and Drug Administration (FDA), the Migraine Research Foundation, and the National Headache Foundation. He serves as consultant, advisory board member, or has received honoraria from Abbvie (Allergan), American Academy of Neurology, American Headache Society, Amgen, Biohaven, Biovision, Boston, Dr. Reddy's (Promius), Electrocore, Eli Lilly, eNeura, Equinox, GlaxoSmithKline, Grifols, Lundbeck (Alder), Merck, Pernix, Pfizer, Teva, Vector, and Vedanta. He holds stock options in Biohaven and Manistee. All other authors: none. Author disclosures are available in the supporting information
You have requested "on-the-fly" machine translation of selected content from our databases. This functionality is provided solely for your convenience and is in no way intended to replace human translation. Show full disclaimer
Neither ProQuest nor its licensors make any representations or warranties with respect to the translations. The translations are automatically generated "AS IS" and "AS AVAILABLE" and are not retained in our systems. PROQUEST AND ITS LICENSORS SPECIFICALLY DISCLAIM ANY AND ALL EXPRESS OR IMPLIED WARRANTIES, INCLUDING WITHOUT LIMITATION, ANY WARRANTIES FOR AVAILABILITY, ACCURACY, TIMELINESS, COMPLETENESS, NON-INFRINGMENT, MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE. Your use of the translations is subject to all use restrictions contained in your Electronic Products License Agreement and by using the translation functionality you agree to forgo any and all claims against ProQuest or its licensors for your use of the translation functionality and any output derived there from. Hide full disclaimer
© 2023. This work is published under http://creativecommons.org/licenses/by-nc/4.0/ (the “License”). Notwithstanding the ProQuest Terms and Conditions, you may use this content in accordance with the terms of the License.
Abstract
Introduction
We investigated the utility of the Telephone-Montreal Cognitive Assessment (T-MoCA) to track cognition in a diverse sample from the Einstein Aging Study.
Methods
Telephone and in-person MoCA data, collected annually, were used to evaluate longitudinal cognitive performance. Joint models of T-MoCA and in-person MoCA compared changes, variance, and test-retest reliability measured by intraclass correlation coefficient by racial/ethnic group.
Results
There were no significant differences in baseline performance or longitudinal changes across three study waves for both MoCA formats. T-MoCA performance improved over waves 1–3 but declined afterward. Test-retest reliability was lower for the T-MoCA than for the in-person MoCA. In comparison with non-Hispanic Whites, non-Hispanic Blacks and Hispanics performed worse at baseline on both MoCA formats and showed lower correlations between T-MoCA and in-person versions.
Conclusions
The T-MoCA provides valuable information on cognitive change, despite racial/ethnic disparities and practice effects. We discuss implications for health disparity populations.
Highlights
You have requested "on-the-fly" machine translation of selected content from our databases. This functionality is provided solely for your convenience and is in no way intended to replace human translation. Show full disclaimer
Neither ProQuest nor its licensors make any representations or warranties with respect to the translations. The translations are automatically generated "AS IS" and "AS AVAILABLE" and are not retained in our systems. PROQUEST AND ITS LICENSORS SPECIFICALLY DISCLAIM ANY AND ALL EXPRESS OR IMPLIED WARRANTIES, INCLUDING WITHOUT LIMITATION, ANY WARRANTIES FOR AVAILABILITY, ACCURACY, TIMELINESS, COMPLETENESS, NON-INFRINGMENT, MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE. Your use of the translations is subject to all use restrictions contained in your Electronic Products License Agreement and by using the translation functionality you agree to forgo any and all claims against ProQuest or its licensors for your use of the translation functionality and any output derived there from. Hide full disclaimer
Details
1 Saul R. Korey Department of Neurology, Albert Einstein College of Medicine, Bronx, New York, USA; Department of Epidemiology and Population Health, Albert Einstein College of Medicine, Bronx, New York, USA
2 Department of Psychology, Brooklyn College, City University of New York (CUNY), Brooklyn, New York, USA; Department of Psychology, The Graduate Center, City University of New York (CUNY), New York, New York, USA
3 Saul R. Korey Department of Neurology, Albert Einstein College of Medicine, Bronx, New York, USA; Department of Psychology, Brooklyn College, City University of New York (CUNY), Brooklyn, New York, USA; Department of Psychology, The Graduate Center, City University of New York (CUNY), New York, New York, USA
4 Saul R. Korey Department of Neurology, Albert Einstein College of Medicine, Bronx, New York, USA
5 Saul R. Korey Department of Neurology, Albert Einstein College of Medicine, Bronx, New York, USA; Department of Epidemiology and Population Health, Albert Einstein College of Medicine, Bronx, New York, USA; Department of Psychiatry and Behavioral Sciences, Albert Einstein College of Medicine, Bronx, New York, USA