A – 90 Capturing Distributional Variations of Demography in Tele-neuropsychological Cognitive Performance Using Quantile Regression

Objective The growing interest in telephone-based neuropsychological assessments is undercut by the absence of viable, demographically adjusted norms for clinical interpretation. This study addresses the gap by using quantile regression (QR) to model the distribution of neuropsychological test scores. Method 536 adults (Age=41.4 (18.6), Range: 18-85; Education=15.8 (2.0), Range:12-18; Sex=59.1% Female; Asian=10.6%, Black=16.2%, Hispanic=8.58%, Other =3.17%, White=61.4%; English as a second language (ESL)=10.8%) completed a battery of tele-neuropsychological assessments (i.e., Delis-Kaplan Executive Function System [DKEFS] Switching, Hopkins Verbal Learning Test—Revised [HVLT-R], Prospective Memory, Action Fluency [AF], Oral Trail Making Test [OTMT] parts A and B, WAIS-IV’s Digit Span Forward and Backward, Information subtests). QR models were fitted on Tau values of .02, .08, .25, .50, .75, .91, and .98 that corresponded to the median cleavage points of the percentile range in the AACN Consensus Conference on uniform test score labeling (Guilmette et al., 2020). A bootstrap method with 500 replications and demographic predictors was included (i.e., age, education, biological sex, ESL, and race/ethnicity). Results QR yielded heterogeneous quantile-specific coefficients of predictors across the cognitive measures. Specifically, the demographic factors (e.g., education, sex, ESL) had a non-uniform predictive strength and direction across the quantile distribution for specific tests (e.g., DKEFS Switching, HVLT-R, WAIS-IV subtests, AF, OTMT-A). Conclusion The findings emphasize the importance of considering demographic factors when interpreting performance on telephone-based cognitive test scores. An automated normative score generator is provided for clinicians and researchers for a nuanced understanding of identifying individuals with cognitive impairment in practice.