Combining Raw Score and Normative Score Cut-Offs for Remote Digital Cognitive Screening: Impact on Sensitivity and Specificity in Cognitively Unimpaired Participants and Participants With Mild Cognitive Impairment and Dementia

Abstract

Objective: The Mayo Test Drive (MTD) screening battery is a digital, self-administered, web-based cognitive assessment (Stricker et al., 2025 TCN). We aimed to compare diagnostic accuracy of raw versus normative score approaches to distinguishing cognitively unimpaired (CU) from mild cognitive impairment/mild dementia (MCI/DEM) participants. We examined sensitivity and specificity across age (< 75 versus 75+ years), sex and amyloid PET positive (A+; ≥25 Centiloid) and negative (A-) subgroups. Method/Data Selection: 2,249 participants (2,121 CU/96 MCI/22 DEM; age=70±12 years; 48% male; education=16±2 years; 95% White) completed MTD remotely and 54% had amyloid PET imaging. Area under the receiver operating curve (AUC) was derived using logistic regression; AUCs were compared. Low raw (< 85 MTD composite; corresponding to < 7 unadjusted scaled score) and low normative (< 40 MTD composite age-, sex- and education-adjusted T-score) cut-offs were defined. Results/Data Synthesis: MTD raw composite showed excellent ability to differentiate CU from MCI/DEM participants (AUC=0.89; 95% CI=0.86-0.92). MTD T-score AUC remained high (AUC = 0.86; 95% CI = 0.84–0.88) but was significantly lower than the raw score (p< 0.001). Higher sensitivity was observed for normative scores in participants < 75 and for raw scores in participants 75+ and men (often converse for specificity; Table). Defining low performance by either low raw or low normative scores helped balance sensitivity and specificity for most subgroups. Lower specificity was observed in those 75+, but specificity was acceptable (81%) when amyloid-refined clinical groups were examined. Conclusions: Results suggest that a combination of both raw and normative score cut-offs has benefits for balancing sensitivity and specificity.