In vocational education and training, it can be assumed that the digital transformation increases the relevance of individuals' cross-occupational collaboration competency (COCC) across vocational and professional sectors, thus improving and reorganizing processes across time and domains on the basis of available information. This poses a challenge to instruction and diagnostics in VET, which can be supported by thorough, empirically grounded theoretical modeling. In Strikovic and Wittmann (2022), we suggested a role-theoretical approach to modeling COCC involving five role-theoretical dimensions: knowing one's own occupational role, knowing the occupational role of one's collaboration partners, latent processes of role distance, role-taking, and object-oriented role coordination. In the present study, we evaluate this role-theoretical model empirically using the example of nursing, specifying and operationalizing the model with a computer-based situational judgment test and using a data set of 328 nursing students to carry out an in-depth analysis of (1) dimensionality and (2) role-related task features thought to influence item difficulty. While model comparisons showed the superiority of a unidimensional vector compared to five- and two-dimensional structures, linear additive regression analysis revealed that the role-related task features on the basis of the five dimensions showed significant effects, as did the number of collaboration partners and the test subjects' subject-matter competence, which can also be linked to the nurses' situational role. These findings are in line with our theoretical reasoning in the COCC model. We discuss their limitations as well as their theoretical and educational significance, outlining needs for further research--also in other professional and vocational domains.