This application proposes to probe the meaning of differences in the measurement of mental status according to racial/ethnic group or level of educational attainment. We will do this by conducting methodological research, namely, comparative analyses using simulated and real data of different approaches to detecting item bias or differential item functioning, a particular form of measurement bias or measurement non-invariance. Our research is motivated by the often reported but as yet incompletely understood relationship of race/ethnicity and years of completed education and performance on mental status measures among older adults. The overarching goal of this research project is to evaluate the extent that item and test bias by race/ethnicity and level of educational attainment influence performance on cognitive assessment measures in late life. The research team includes investigators from different institutions who have studied measurement invariance using the methods of item response theory. Members of our research team have been at the vanguard of efforts to apply modern psychometric techniques to the detection of bias in the measurement of cognitive function, yet, we recognize that this area of research is in its infancy. To achieve our goal of moving the field forward in this regard, our specific aims are to (1) Conduct 'basic psychometric science' research on differing methodological approaches to detecting biased test items, comparing IRT techniques that employ different estimation strategies; (2) Evaluate the extent that item bias attributable to race/ethnicity and educational attainment impact assessments of group differences in assessed cognitive status in three large general population surveys; and (3) Evaluate, among cognitively healthy older individuals, the extent that educational attainment influences individual items and overall ability estimates in (a) tasks of everyday cognitive functioning, and (b) neuropsychological performance (memory, reasoning ability, speed of processing).
These aims will be accomplished by analyzing existing and publicly available data sets, or data sets available to the investigators (pending public release). This is an important area of research that will both examine the extent that mechanisms other than assessment bias explain group differences in assessed cognition among older adults, and inform efforts to develop education appropriate and culture-fair assessment tools.
Showing the most recent 10 out of 14 publications
