In operational testing programs using item response theory (IRT), item parameter invariance is one of essential requirements in common item equating designs. However, in practice, the stability of item parameters can be affected by many factors. Particularly, this study utilised data from the large-scale Graduate Skills Assessment (GSA) in 2010 to investigate the change of Rasch item difficulty by item positions. The test included 78 multiple-choice items and was presented in eight test forms by arranging the items in different orders. Items addressed three components of generic skills: Critical Thinking, Problem Solving and Interpersonal Understandings. Each test form was randomly administered to about 8000 Colombian university students in November 2010. In the cohort there were roughly equal numbers of males and females. For each component, a three-faceted Rasch model was applied to evaluate the variation of test form difficulty and item difficulty estimates by these forms. Then for each item, the difference of item difficulty estimates from each pair of the forms was examined in relation to the position difference of the item in the forms. Findings showed that there was a small variation in the test form difficulty. However, in general, items themselves became more difficult when they were located towards the end of the test. The change of the item difficulty correlated significantly with the corresponding change of its position in the test. Moreover, the correlation was highest for Problem Solving items and lowest for Interpersonal Understandings items. The correlations were higher for males than for females.