Education correction for neurocognitive testing

Abstract

Comprehensive neuropsychological assessments, composed of specific tests targeting different aspects of cognition, are standard diagnostic tools to characterise the singleor multidomain impairments due to neurodegeneration, depression, or other brain disorders. Individual test results are frequently corrected for age-related changes in performance, and for the years of formal education, a proxy measure for cognitive reserve, that also comprises occupational attainment and intelligence. Education correction for neurocognitive tests could influence a physician’s decision how to diagnose and treat an individual, and has been used to reclassify demented and nondemented individuals in diagnostic assessments. Others point to education differences in European and American populations to explain heterogeneous performances in a common screening test for cognitive impairment. Both scenarios imply that we expect a well-educated person to perform better, reflected by a correlation between the years of formal education and cognitive abilities. We hypothesised that generational differences in education opportunities would challenge this assumption. In the 20th century, many societies experienced substantial changes in education opportunities, specifically regarding access to university education. This is reflected by an increasing percentage of university students in these populations and a strong shift in the ratio of men and women attending the institutions. In Germany for example, ‘universities of applied sciences’ have complemented the higher education sector since the 1970s, increasing tertiary education possibilities in addition to other quantitative changes in the education system. Furthermore, over the course of the last century, high tertiary education inequality due to social class effects has been substantially declining, specifically for women. We investigated 276 people (age range 44–94 years, median: 73 years, 124 male, 152 female), presenting for the first time to a German university memory clinic with cognitive impairment in the spectrum from subjective complaints to dementia. Written informed consent was obtained in accordance with the university’s ethics committee requirements. In addition to an interview (with caregiver, if available) and examination by an experienced physician, all patients underwent routine laboratory testing, magnetic resonance brain imaging and detailed neuropsychological assessments, in line with standard guidelines for how to evaluate cognitive impairment. We performed the statistical analyses using SPSS 24.0 (IBM, Armonk, NY, USA), with a significance level of P < 0.05, two-sided, for all analyses. Participants reporting university education (n = 101) compared with all other individuals (n = 175) were more likely to be older, male, and intrinsically motivated for diagnostic assessments, possibly reflecting the protective effect of cognitive reserve, gender differences in education opportunities, and greater dementia worries. We also found weak to moderate correlations between the years of formal education and a multiple choice vocabulary test, a common measure of premorbid intelligence. The correlation strength shows a notable trend for linear decline with increasing age (from Pearson’s r = 0.60, P = 0.01 among our youngest subjects to Pearson’s r = 0.30, P = 0.02 in our oldest participants, Fisher’s z-transformation: z = 1.38, P = 0.08). However, this can only be detected among subjects without university education (patients younger than 65 years: Pearson’s r = 0.39, P < 0.01; patients older than 65 years: Pearson’s r = 0.12, P = 0.28, z = 1.50, P = 0.07), but not among those individuals with university education (patients younger than 65 years Pearson’s r = 0.20, P = 0.46; patients older than 65 years: Pearson’s r = 0.30, P = 0.02, z = 0.35, P = 0.37). When investigating patients with dementia only, we did not observe the effect. The multiple choice vocabulary test is a valid measure of global premorbid intelligence, appropriately insensitive for the influence of neuropsychiatric disease. Considering the test’s limitations, our data would be in line with reduced validity in the case of severe brain disease. However, neither aging nor the presence of cognitive impairment per se sufficiently explained the observed results. We may speculate that older