• Curt Hagquist, Malin Bruce, and J Petter Gustavsson.
• Karlstad University, SE-651 88 Karlstad, Sweden. curt.hagquist@kau.se
• Int J Nurs Stud. 2009 Mar 1; 46 (3): 380-93.

ObjectiveThe purpose was to introduce the Rasch model by showing an application in nursing research.MethodsThe Rasch model was used to examine the psychometric properties of the nursing self-efficacy (NSE) scale. Data were collected among nursing students in Sweden. Two sets of items were analysed more thoroughly: an original set of nine items with eleven response categories and a revised set of seven items with seven response categories. Invariance of the item functioning and the categorisation of the items were analysed. Targeting was examined by comparisons of the items and persons locations. Differential Item Functioning across sample groups such as gender was examined using analysis of variance. The final set of seven items was also analysed more closely with respect to possible multidimensionality and response dependence.ResultsThe Rasch analysis of the original set of nine items showed high reliability measured by a person separation index, but it also indicated severe problems with the targeting, the categorisation of the items as well as lack of invariance. Although the revised set comprising seven items with seven categories performed better than the original item set some items showed misfit according to formal test statistics. Graphical examination showed, however, that the items operated in the right direction. The formal test of local independence of the items indicated minor signs of multidimensionality, alternatively response dependence.ConclusionsThe Rasch model is useful for rigorous examination and development of measurement instruments in nursing research. The Rasch model facilitates disclosure of lack of invariance and other measurement problems that may not be easily detected by traditional analyses. Hence, the NSE-scale would probably have performed much better if the developmental work had been guided by Rasch analyses. In future work on the scale, priority should be given to improving the targeting and the categorisation of the items.

27 keywords...

hide…