Education-Based Cutoffs for Cognitive Screening of Alzheimer's Disease.

Introduction: The educational background and size of the elderly population are undergoing significant changes in Finland during the 2020s. A similar process is likely to occur also in several European countries. For cognitive screening of early Alzheimer's disease (AD), using outdated norms and cutoff scores may negatively affect clinical accuracy. The aim of the present study was to examine the effects of education, age, and gender on the Consortium to Establish a Registry for Alzheimer's Disease neuropsychological battery (CERAD-nb) in a large register-based, clinical sample of patients with mild AD and nondemented at-risk persons from the general population (controls) and to examine whether corrected cutoff scores would increase the accuracy of differentiation between the 2 groups.

Methods: CERAD-nb scores were obtained from AD patients (n = 389, 58% women, mean age 74.0 years) and from controls (n = 1,980, 52% women, mean age 68.5 years). The differences in CERAD-nb performance were evaluated by univariate GLM. Differentiation between the 2 groups was evaluated using a receiver operating characteristic (ROC) curve, where a larger area under the ROC curve represents better discrimination. Youden's J was calculated for the overall performance and accuracy of each of the measures.

Results: Of the demographic factors, education was the strongest predictor of CERAD-nb performance, explaining more variation than age or gender in both the AD patients and the controls. Education corrected cutoff scores had better diagnostic accuracy in discriminating between the AD patients and controls than existing uncorrected scores. The highest level of discrimination between the 2 groups overall was found for two CERAD-nb total scores.

Conclusions: Education-corrected cutoff scores were superior to uncorrected scores in differentiating between controls and AD patients especially for the highest level of education and should therefore be used in clinical cognitive screening, also as the proportion of the educated elderly is increasing substantially during the 2020s. Our results also indicate that total scores of the CERAD-nb are better at discriminating AD patients from controls than any single subtest score. A digital tool for calculating the total scores and comparing education-based cutoffs would increase the efficiency and usability of the test.