The need to improve geriatric care quality presents a challenge that requires insights from stakeholders. While simulation-based training can enhance caregiving competencies, extracting meaningful insights from these experiences to inform simulation design remains a challenge. In this study, we employed Epistemic Network Analysis (ENA), Ordered Network Analysis (ONA), and Chronologically-Ordered Representations of Discourse and Tool-Related Activity (CORDTRA) within an Augmented Reality simulation to analyze caregiver competencies and engagement. Twenty participants interacted with a virtual geriatric patient across two conditions: an unaware condition, where the virtual patient lacked contextual awareness, and an aware condition, where the patient offered personalized responses and reacted to environmental and conversational cues. Results showed that participants provided more supportive care and demonstrated stronger person-centered caregiving behaviors when interacting with an aware virtual patient. In addition, ONA during a reasoning task revealed a significant difference between the two conditions, suggesting more adaptive strategies in the aware condition. CORDTRA analysis further indicated higher participant engagement when the virtual patient expressed awareness. These findings have implications for Human-Computer Interaction and nursing education by demonstrating how Quantitative Ethnography can be applied to dynamic, multimodal simulations to evaluate and inform the design of effective training systems.