In this paper, we adopted an information-processing perspective to examine the relationship between cognitive engagement and students' performance in a simulation-based training environment. In particular, we examined what forms of cognitive engagement students used while diagnosing virtual patients and whether engagement forms predicted students' diagnostic confidence and efficacy. A total of 88 medical students from a large North American university voluntarily participated in this study. We used latent profile analysis (LPA), a person-centered statistical method, to identify groups of students with similar information processing patterns. Findings from this study revealed that students displayed various forms of cognitive engagement, i.e. recipience, resource management, and task-focused. Moreover, we found that group difference in diagnostic confidence was moderated by task complexity. In terms of diagnostic efficacy, students who were task-focused or resource managers did better than the recipience students. The findings advance our understanding of theories of cognitive engagement as well as inform the design of effective interventions in developing simulation-based learning environments.