Applying Modern Test Theory in PRO Measure Development: Rasch modelling

Techniques applying classical test theory (CTT), such as item-total correlation analysis, stepwise regression and factor analysis, have been traditionally used in the development and testing of measurement models for PROs (Coste et al. 1997). Item response theory, especially the Rasch model, is an alternative approach and is increasingly being used. This is regarded by some as setting new rules in measurement and thus creating a new ‘gold standard’ in PRO instrument development (Reise and Henson 2003; Nijsten 2012). The Rasch model generates a linear metric scaled in logit units, representing the construct being measured, on which both the items and persons are located hierarchically reflecting their levels on the construct (Prieto et al. 2003). Further, the probability of a given response on an item by an individual is then given by a logistic function of the difference between the item location and person location and nothing else (Twiss et al. 2011); and the item and person parameters of the RM demonstrate invariance (DeMars 2010). Then, following a prescriptive approach, items and persons are assessed for conformity to the model by applying fit statistics (Nijsten et al. 2006) (Table 5.1).