The Role of an Evidence-Based Risk Score in Predicting Baseline and Follow-up Cognition in a Multiethnic Sample

Andrea Ochoa Lopez, University of Houston, Houston, United States
Stephanie Torres, Massachusetts General Hospital, Boston, United States
Kasey Escamilla, University of Houston, Houston, United States
Luis Medina, University of Houston, Houston, United States


Given the expected increase in adults living with dementia in the next few decades, there is a critical need to identify populations at higher risk for cognitive decline. Risk scores accounting for multiple health factors predict dementia incidence in U.S. American populations, and may also predict changes in cognitive performance. However, many risk scores have been primarily developed in largely non-Hispanic White samples, and little is known about their applicability in diverse populations. We evaluated the association between an evidence-based risk score, the Mexican American Dementia Nomogram (MADeN), and cognition across different ethnoracial subgroups in the United States and Mexico.

Participants and Methods:

Using data from the Mexican Health and Aging Study (MHAS; n = 14,115, Mage = 62.9 ± 10.5, Medu = 4.55 ± 4.38, 57.6% female), and the United States’ Health and Retirement Study (HRS; n = 15,676, Mage = 68.4 ± 10.5, Medu = 12.6 ± 3.22, 58.9% female) we calculated MADeN scores for each participant using ten risk factors: female, older age, lower educational attainment, limited social support, limited community involvement, diabetes, depression, pain, instrumental activity of daily living impairment, and limited mobility. Using principal component analysis, we calculated cognitive composite scores for baseline and follow-up performance (2003/2012 for MHAS, 2002/2012 for HRS), and a cognitive change score, ΔCog. Using linear regression, we examined the MADeN score’s utility in predicting cognition across four ethnoracial groups: Mexican, U.S. Hispanic/Latin American (H/L), White, and African American.


MADeN scores explained 14.4% of the variance in baseline cognitive performance for Mexicans, 3.2% for U.S. H/Ls, 2.3% for African Americans, and 3.3% for Whites (all ps < .001). Regarding cognition at follow-up (9 to 10 years later), MADeN scores explained 8.4% of variance for Mexicans, 5.3% for U.S. H/Ls, 1.9% for African Americans, and 1.2% for Whites (all ps < .001). Prediction of ΔCog by MADeN scores was significant but minimal, explaining ≤1% of variance for all participant groups (all ps < .001).


An evidence-based risk score developed to predict dementia incidence in Mexican American populations explained a modest but significant degree of variance in cognitive performance across two time points for additional, U.S.-based ethnoracial groups. Consistent with previous findings in a Mexican sample, the risk score was less useful in explaining change in cognitive performance. Additional research is needed to elucidate the cross-cultural differences in the predictive utility of risk scores.

Category: Cross Cultural Neuropsychology/ Clinical Cultural Neuroscience

Keyword 1: cross-cultural issues
Keyword 2: cognitive functioning
Keyword 3: ethnicity