On the Concordance of Three Reliable Change Indexes: An Analysis Applying the Dynamic Wisconsin Card Sorting Test

Abstract

Classifying individuals according to learning proficiency is a basic issue in dynamic assessment and requires the identification of a significant change of performance. Previous research on the comparability of three competing methods for the ascertainment of statistically significant intraindividual change (Reliable Change Index [RCI], Gulliksen-Lord-Novick [GLN] method, and the Ultimate RCI/RCINDIV [URCI]) has yielded ambiguous or contradictory results. The present study compared the frequencies of different types of responders in a Wisconsin Card Sorting Test (WCST-64) training. Responder types were determined by each of these RCI methods in a large sample of participants diagnosed with schizophrenia (N = 400). Although agreement of RCI and GLN was high, the URCI differed markedly, and statistical significance of gains was judged more liberally. Although a statistical comparison cannot provide answers to questions concerning the validity of classifications, researchers and practitioners are advised to be aware of these differences when choosing a method for testing the statistical significance of individual change.