OBJECTIVES Rasch modelling theory and its extensions have become popular tools in assessing psychometric properties of patient-reportedoutcome (PRO) instruments. Since the Rasch model assumes a unidimensional structure, it is important to assert that this assumption has not been violated. However, there has been much debate about using factor analysis as the first step to assess dimensionality or using the Rasch model directly to identify items not fitting the unidimensional model. This study uses simulated data to compare the two techniques to examine the unidimensional assumptions. METHODS Simulated data that represent a typical PRO instrument are generated based on the following variables: sample size (200 or 400), number of factors (1, 2, or 3), and correlation among factors (0.4 or 0.7). Each simulated PRO assessment contains 15 items with 5 response categories. Exploratory factor analysis is conducted, and the number of factors proposed by the results are noted. A Rasch model theory analysis is also conducted, and the number of mis-fit items is noted. The unidimensional test associated with Rasch model is also conducted. RESULTS Preliminary results suggest that when there is one dominant factor, the Rasch results identify the simulated factor and it becomes the underlying trait. Items not belonging to the dominant factor are flagged as mis-fit items or as forming secondary factors based on the unidimensional test. In situations where there is no dominant factor, factor analysis is able to identify separate factors, whereas the results are inconsistent using Rasch model. CONCLUSIONS Rasch first or factor first? The preliminary findings suggest that, when there is prior knowledge of a dominant factor, Rasch modelling can be conducted first and will result in a unidimensional measure. When there is no prior knowledge of a dominant factor, then factor analysis should be conducted first to examine the dimensionality.