|Home | About | Journals | Submit | Contact Us | Français|
Normative neuropsychological data for U.S. racial/ethnic minorities is limited. Extant norms are based on small, regional groups that may not be nationally representative. The objectives of this study were to 1) provide norms for a modified Symbol Digit Modalities Test (M-SDMT) based on a nationally representative sample of African Americans, Caribbean Blacks and non-Latino Whites (NLW) living in areas with large populations of Blacks, and 2) determine significant correlates of M-SDMT performance. The M-SDMT was administered to a subset of respondents from the National Survey of American Life in standard, face-to-face interviews. M-SDMT performance was influenced by race/ethnicity, age, education, and gender. African Americans and NLW groups had similar M-SDMT performances, which differed from Caribbean Blacks. The Black ethnic differences in M-SDMT were not explained by the sociodemographic factors considered in this study. Unlike previous work, this study supports the consideration of Black ethnicity when evaluating Black neuropsychological test performance.
A principle objective of the National Institute of Health’s Strategic Plan and Budget to Reduce Health Disparities is “developing new and improved approaches for detecting, diagnosing, preventing, treating or delaying onset or progression of diseases and disabilities that contribute to health disparities (U.S. DHHS, 2002).” For clinicians and cognitive scientists, it is essential that appropriate normative test data be available to screen for cognitive impairment and to better understand disease process versus bias artifact in conducting research with diverse and representative U.S. populations (Lampley-Dallas, 2001).
The Symbol Digit Modalities Test (SDMT) is a widely used, brief screening test for neurologic disorders (Smith, 1968, 1982). The cognitive demands of the SDMT include attention, visual scanning, and motor and psychomotor speed. The SDMT is purportedly sensitive to a wide-range of neurologic and neuropsychiatric disorders, but may lack disorder specificity (Lezak, Howieson, & Loring, 2004). Impaired SDMT performance has been reported in otherwise healthy adults with white matter hyperintensities, and a variety of pathologic conditions, such as traumatic brain injury, and subtypes of multiple sclerosis (Felmingham, Baguley, & Green, 2004; Huijbregts et al., 2004; Sachdev, Wen, Christensen, & Jorm, 2005). Thus, the SDMT is a useful screening test provided that it is appropriately calibrated to the population of intended use.
A recent review of published SDMT normative data indicates that it is widely used; however, normative data is lacking for adult racial/ethnic minority groups (Sheridan et al., 2005). In recent years, more normative data for African Americans have become available for neuropsychological tests (e.g., (Byrd, Touradji, Tang, & Manly, 2004; Moering, Schinka, Mortimer, & Graves, 2004). Most studies have focused on adults over age 60, which may be of limited utility to clinicians and investigators working with a wider range of adults (Fillenbaum, Heyman, Huber, Ganguli, & Unverzagt, 2001; Friedman, Schinka, Mortimer, & Graves, 2002; Lucas et al., 2005; Manly, Byrd, Touradji, & Stern, 2004; Mast, Fitzgerald, Steinberg, MacNeill, & Lichtenberg, 2001). Additionally, ethnicity among Blacks and test performance has rarely been examined (Byrd, Sanchez, & Manly, 2005). Currently, normative data is unavailable for younger Caribbean Blacks who, some argue, may be culturally and socioeconomically distinct from other Blacks (Williams & Jackson, 2000). The purposes of this study were to determine significant correlates of a modified version of the SDMT performance and provide normative data for clinical and research use from a large, nationally representative sample of African Americans, Blacks of Caribbean ancestry, and non-Latino Whites (NLW) living in areas with large populations of Blacks.
Participants in this study were from the National Survey of American Life (NSAL), which is a population-based, multi-stage probability sample of 6,082 African American, Caribbean Blacks, and non-Latino White (NLW), ages 18–94) residing in the U.S. The NSAL, along with the National Comorbidity Survey Replication (NCS-R) and the National Latino and Asian American Study (NLAAS), is a part of the Collaborative Psychiatric Epidemiology Surveys. Details of the sampling methodology and sampling weight calculations for the NSAL have been previously published (Heeringa et al., 2004).
The NSAL adult sample was an integrated national household probability sample of 3,570 African Americans, 891 NLW, and 1,621 Blacks of Caribbean descent (Caribbean Blacks), for a total sample of 6,082 individuals aged 18 and over. The core African American sample is a nationally representative sample of households located in the 48 coterminous states with at least one Black adult 18 years or over who did not identify ancestral ties in the Caribbean. The Caribbean Black sample was selected from two area probability sample frames (i.e., 265 came from the households in the core sample frame, while 1,356 came from an area probability sample of housing units from geographic areas with a relatively high density of persons of Caribbean descent). The NSAL analysis weights for the African American and Caribbean Black samples were designed to provide population representation for these populations in the 48 coterminous states. Caribbean Blacks included persons who self-identified as Black, and answered affirmatively when asked if they 1) were of West Indian or Caribbean descent, or 2) indicated that they were from a country included on a list of Caribbean area countries presented by the interviewers, or 3) indicated that their parents or grandparents were born in a Caribbean area country. The NLW sample was a stratified, disproportionate sample of NLW adults residing in households located in Census 2000 tracts and blocks that have 10% or greater African American populations. The sample design and sample analysis weights were determined so that they would be optimal for comparative analyses in which residential, environmental, and socioeconomic characteristics were controlled for in Black-NLW statistical contrasts.
The interviews used in this study were conducted face-to-face using interviews that averaged two hours and twenty minutes in length. Data collection was conducted between February 2001 and June 2003. The overall response rate was 72.3% and varied by race and ethnicity: 70.7% for African Americans; 77.7% for Caribbean Blacks; and 69.7% for the NLW. Additional information on the NSAL data collection has been previously published (Pennell et al., 2004).
Sample respondents were included in this study if they had completed at least one section of the cognitive test and did not meet criteria for any of the following Diagnostic and Statistical Manual (DSM) IV psychiatric conditions in the past twelve months: 1) panic disorder; 2) agoraphobia without panic disorder; 3) social phobia; 4) generalized anxiety disorder; 5) major depression; 6) dysthymia; and 7) bipolar I or II (APA, 1994). The DSM-IV, World Mental Health Composite Interview (WHO CIDI), which is a fully structured diagnostic interview, was used to assess the mental disorders itemized above. Of the 6,082 respondents in the NSAL sample, 4,545 (74.73%) were included in the study per the criteria above, 712 (11.71%) had at least one condition, and the remaining 825 (13.56%) did not have information collected on the disorder or SDMT measures due to incomplete data, physical impairment and refusals. Commands appropriate for subpopulation analyses of survey data in SAS Version 9.1.3 (SAS Institute, Inc., Cary, NC) and Stata Release 9 (StataCorp LP, College Station, TX) software were utilized when analyzing the data collected for this subpopulation.
The SMDT is a timed (90s), paper and pencil test. It consists of 1) a key with two rows, with nine stimulus symbols in the upper row, and matched numbers (1–9) in the lower row, and 2) a two-row grid with the same nine stimulus symbols in the upper row and blank cells for numeric responses in the lower row (Smith, 1982). This modified version of the SDMT, called the Modified Symbol Digit Modalities Test (M-SDMT) has 50 response cells and two parts (A and B). Parts A and B use the same stimulus and numeric coding system. The M-SDMT has been previously used in a study of genetics among African Americans twins (Whitfield, Brandon, Wiggins, Vogler, & McClearn, 2003); however, detailed characteristics of the M-SDMT have not been previously published. The examinee is given 45s to complete each part (A and B). All tests were administered to NSAL respondents in the paper and pencil format.
There are three separate scores for the M-SDMT. They are 1) the total number of correct responses for Part A (out of 50 possible); 2) the total correct responses for Part B (out of 50); and 3) the total correct responses for Parts A and B combined (total possible was 100).
All statistical analyses performed in the study were design-based, taking the NSAL sampling weights into account to ensure that estimates were representative of the respective ethnic populations in the U.S., and using the NSAL sample design variables to calculate standard errors for statistical estimates appropriately reflecting the stratified, clustered design of the NSAL sample. Analyses were performed using procedures for design-based analysis of survey data in SAS Version 9.1.3 (SAS Institute, Inc., Cary, NC) and Stata Release 9 (StataCorp LP, College Station, TX).
Descriptive analyses of the survey data were performed to generate tables of normative data for the scores on the M-SDMT, indicating weighted estimates of means, weighted estimates of standard deviations, and sample sizes for cross-classifications of the NSAL population defined by gender, ethnicity, age, and education. The relationships between SDMT performance and demographic variables were also evaluated in the analyses. Linear regression models were used to model the three dependent variables representing the three M-SDMT scores. Independent variables included ethnicity (African Americans, Caribbean Blacks and NLW) and gender as categorical factors, in addition to age and education as continuous independent variables. Terms to model the interaction of ethnicity with other independent variables were also included in the models. Thus, the following independent variables were included in the models: ethnicity, education, gender, age, ethnicity × age, ethnicity × education, and ethnicity × gender. Values for the age and education variables were mean-centered by subtracting from each variable the overall mean for the subpopulation of interest meeting the inclusion criteria, to avoid problems associated with collinearity of interaction terms and main effects (Rawlings, Pantula, & Dickey, 1998). Standard diagnostic tests were used to assess the appropriateness of all assumptions behind the linear regression models (e.g., normality and constant variance for the random errors).
The demographic characteristics of the NSAL subsample meeting the inclusion criteria for this study are presented in Table 1. Consistent with 2000 U.S. Census data, the African American and Caribbean Black samples were younger, more likely to be female, and had fewer years of education than the NLW sample (Census, 2000). Overall, ages ranged from 18 to 94 years across the three groups.
Table 2 presents regression models indicating the estimated relationships of age, education and gender to the dependent variables measuring scores on Parts A, B and the Total (Part A + Part B) of the M-SDMT. In each of the three models (not shown), none of the two-way interactions between the demographic variables and ethnicity were found to be significant (p<.05), based on design-adjusted Wald tests for the interaction effects. The results of these models indicate that there are significant main effects of these demographic variables on the scores, and that the effects do not vary for different subgroups. Age was found to have a significant negative relationship with each of the three scores, education was found to have a positive relationship with each of the three scores, and females were found to have significantly higher means on each score than males. In addition, Caribbean Blacks were found to have a significantly lower mean than both African Americans and NLW on each of the three scores. These main effects are evident in the weighted means and standard deviations presented in Tables 3 through through55.
Tables stratified by ethnicity, age, education and gender are presented for M-SDMT Part A (Table 3), Part B (Table 4) and the Total Score (Table 5), with weighted estimates of means and standard deviations based on the NSAL subsample. The respective cell sample sizes used to compute the table estimates are provided as well.
The M-SDMT was developed to improve detection and aid in the diagnosis of disorders for prevention, treating and delaying onset or progression of diseases and disabilities in different ethnic/racial groups. The normative data presented here for the M-SDMT are based on the largest sample of U.S. Blacks ever published. In addition, due to the extensive sampling procedures, the data are highly representative of U.S. Blacks and geographically similar NLW. Further, statistical corrections for significant predictors of M-SDMT performance should further assist the user in controlling demographic biases. The relationships of the demographic predictors with M-SDMT performance reported in this study were found in additional analyses (not reported) to be similar and significant in each of the three ethnic subgroups. This suggests that demographic differences between the groups were not attenuating the relationships of important demographic predictors with M-SDMT performance, and that the demographic predictors considered should be controlled for when predicting performance. While the results of this study should be generalizable to African Americans and Caribbean Blacks in the U.S., it is important for users of the M-SDMT to be cognizant of limited generalizability of the M-SDMT to all NLW because the sample in this study is a unique group that may differ from the majority of NLW. The normative data in this report for the M-SDMT should provide clinicians and researchers with appropriate information for using this screening test with African American, Caribbean Blacks and NLW from areas with larger concentrations of Blacks.
In this study, African American and NLW groups did not differ significantly on M-SDMT performance; however, the Caribbean Black group had significantly lower performances compared to the African American and NLW groups. The findings of this study are consistent with previous work with older adults that ethnic differences among Black Americans should be considered when evaluating cognitive performance using tests like the M-SDMT (Byrd, Sanchez, & Manly, 2005). It is not clear from the results of this study if other neuropsychological tests would be similarly affected by Black ethnicity.
The M-SDMT performance differences between Black ethnicity groups were likely to have been associated with factors other than the demographic variables considered in this study. The demographic variables examined in this study explained a small fraction (16.2%) of the possible total variance associated with M-SDMT performance. This suggests that other factors could explain the ethnic differences in M-SDMT. Reading performance has been previously shown to predict cognitive test performance among African Americans and Caribbean Blacks (Byrd, Sanchez, & Manly, 2005; Manly, Byrd, Touradji, & Stern, 2004); however, reading level was not examined in the NSAL. In previous work with Latinos, associations between nativity, acculturation, and cognitive function were found (González et al., unpublished manuscript). The influences of nativity and acculturation have only recently been examined among Blacks. Previous work with the NSAL Caribbean Black sample has shown that psychopathology varies by U.S. acculturation (Williams et al., 2007). It is possible that the level of acculturation within the Caribbean Black sample may have contributed to the ethnic differences observed in this study. This study did not consider the associations between M-SDMT performance and indicators of acculturation, such as nativity and years of U.S. residency. Although previous work has found trends between acculturation and cognitive performance among older Caribbean Blacks (Byrd, Sanchez, & Manly, 2005; Manly, Byrd, Touradji, & Stern, 2004), these associations merit further investigation in future studies with larger and younger adult samples (Whitfield, 1998a, 1998b).
In summary, the M-SDMT is a brief screening test, and the normative data presented in this study should prove useful to clinicians and researchers interested in assessing the cognitive functioning of African American and Caribbean Blacks and NLW living in areas with larger Black populations. Further examinations of the between and within group variability are needed to better explain the similarities and differences between these ethnic/racial groups.
The research was supported by grants from the National Institute of Mental Health (MH 67726, H.M. González; MH 57716, J. Jackson). The authors would like to express our thanks to Myriam Torres and Julie Sweetman for editing this manuscript.