Item Response Theory (IRT) analysis of the SKT Short Cognitive Performance Test

In this article, the SKT Short Cognitive Performance Test (SKT = Syndrom Kurztest; Erzigkeit, 2001) was tested with models of IRT. Here, the SKT test functions as an example on how IRT- analyses with any cognitive test, which results in a summary score, can be applied. The SKT was newly scaled in 2015 (Stemmler, Lehfeld & Horn, 2015) using a regression-based continuous norming approach (Crawford & Garthwaite, 2006) while leaving the testing material unchanged. The data were from a sample of 811 older adults ( x  73.3 years; SD=7.2) assessed at multiple testing sites in Germany. The sample consisted of cognitively healthy older adults (n =250), of older adults suffering from mild cognitive impairment (MCI; n=290) as well as adults suffering from dementia (n=271). Analyses to check the psychometric validity of the SKT encompass the Partial Credit Model (PCM), the Graded Response Model (GRM) and Multidimensional Response Models (MRM).

The strict assumptions of the one parametric Rasch model with regard to specific objectivity were not met. The two parametric GRM provided a better goodness-of-fit. A two-dimensional GRM revealed that the subtests assessing Memory and those assessing Attention can be distinguished as two separate dimensions. However, the use of a total score seems still valid, but, next to a careful inspection of the total score the scores for Memory and Attention should be taken into account and considered separately.