Validity of interpretive algorithms for the neuropsycholopgical batteries in the Tracking and Comprehensive databases: Foundation for research on cognition in the CLSA

Normative data were created for neuropsychological measures that accounted for language of administration, sex, advanced age and education. These normative data were novel in the use of weights to adjust for CLSA sampling procedures, but the validity of this approach is unclear. We will compare validity of the clinical algorithm based on normative data with and without adjustment for sampling weights. A second algorithmic approach will detail the lower order, intermediate and higher order relations between normatively corrected neuropsychological tests with measurement of cognitive domains, such as memory and executive function and overall cognition to create continuous cognitive summary scores. Finally we will explore validity of algorithms for the telephone-administered neuropsychological battery to support its use in remote testing.

We described the properties of the cognitive tests used in the CLSA. We found that advanced age was associated with increased likelihood of misclassification, but medical conditions were not. Consequently, we did not plan to adjust for medical conditions when creating normative comparison standards. We found that use of sampling weights did not impact the normative comparison standards, so we planned to provide norm-referenced scores that were both unweighted and weighted to allow for flexibility in use by CLSA researchers. Finally, we explored different models for creating normative comparison standards related to how well they reduced measurement bias due to covariates and how different methods impacted precision of point estimates. This work led to the creation of normative comparison standards that use a hybrid regression approach where they were stratified by sex and education level (four levels) based on age predicting cognitive test scores.