Methods/Statistics
Cognitive data harmonization across two racially diverse cohorts in the United States Michelle Flesaker* Michelle Flesaker A. Zarina Kraal Justina F. Avila-Rieger M. Maria Glymour Emily M. Briceño Jennifer J. Manly Lindsay C. Kobayashi Marcia Pescador Jimenez
Introduction: Few cohorts have sufficient racial diversity to identify drivers of racial disparities in Alzheimer’s Disease and related dementias (ADRD). Pooling individual participant data from different samples can effectively increase sample size and diversity. We harmonized cognitive function data from two US samples, the Reasons for Geographic and Racial Differences in Stroke (REGARDS) and the Health and Retirement Study (HRS) for these purposes.
Methods: Data were from dementia-free participants in the 2010 HRS (n = 19,887) and 2009-2013 REGARDS waves (n = 19,690). We evaluated comparability of cognitive test items between cohorts. We used confirmatory factor analysis (CFA) models to derive harmonized factor scores for general and domain-specific (memory, orientation, language) cognitive function, leveraging common cognitive test items across cohorts to link scales, while retaining those unique to each study. We assessed construct validity of the resultant factor scores by regressing them on variables known to be associated with cognitive function (age, sex/gender, and education).
Results: The combined sample (n = 39,577) had a mean age of 67.4 (SD = 10.3) years. 28% were Black/African American and 68% were White. We identified 4 cognitive items shared by the cohorts and 12 unique items. CFA models demonstrated adequate fit based on predetermined criteria. Younger age (β = -0.36, 95% CI [-0.37, -0.35]), female sex/gender (β = 0.29, 95% CI [0.27, 0.31]), and higher education (β = 0.63, 95% CI [0.61, 0.66] for some college, reference = less than high school) were associated with higher general (results shown) and domain-specific cognitive scores, supporting construct validity.
Discussion: We harmonized cognitive tests across two population-based studies of aging in the US to increase sample size for the examination of racial differences. This work provides a foundation for researchers to improve inferences on drivers of ADRD racial disparities.