Diagnostic accuracy of cognitive assessment in diverse older adults: effects of normative sample and demographic adjustments

Abstract

AbstractBackgroundRacial and ethnic disparities in ADRD diagnosis are well documented but remain poorly understood. Clinical ADRD diagnosis relies on neuropsychological assessment, which is interpreted by comparing individual performance to demographically adjusted reference group norms. Differential use of demographic adjustment procedures and sources of normative data may compromise diagnostic accuracy of neuropsychological findings in diverse individuals. Using two different demographic adjustment approaches derived from two distinct normative samples, we generated cognitive impairment categories and evaluated their associations with longitudinal trajectories of independent executive functioning and memory scores.MethodParticipants were 717 older adults (age: 77.5±6.9, 56% women) enrolled in the Kaiser Healthy Aging and Diverse Life Experiences (KHANDLE) study who completed up to 3 annual visits. The main predictor was cognitive impairment category (no impairment, single‐domain, multiple‐domain) based on Uniform Data Set Neuropsychological Battery (UDSNB) scores at baseline. UDSNB scores were either demographically unadjusted or adjusted (age, sex, education, and race/ethnicity) using data from the National Alzheimer’s Coordinating Center (NACC; N = 8313, 16% Black, 84% White) or KHANDLE (N = 397, 26% Asian, 18% Black, 30% LatinX, 26% White) samples. Linear mixed effect models were used to examine the associations between differentially derived cognitive impairment categories with longitudinal changes on episodic memory and executive functioning composites from an independent cognitive battery (Spanish and English Neuropsychological Assessment Scales, SENAS).ResultCognitive impairment categories derived from the KHANDLE sample were more strongly associated in all four racial and ethnic groups with longitudinal decline on episodic memory compared to the NACC sample‐derived categories (Figure 1). Similar, albeit attenuated, effects were observed in relation to longitudinal decline on the executive function composite. There were no clear differences in the associations between cognitive impairment categories with executive functioning or memory decline using demographically adjusted vs. unadjusted scores.ConclusionOur results highlight the need for greater representation of diverse individuals in normative samples underlining the limitations of the NACC‐derived normative algorithms for cognitive assessment of individuals from diverse populations. Further investigations into the clinical value of the practice of demographic adjustments are also warranted.