This paper explores three dimensions of digital twin development in healthcare: knowledge extraction, clinical pathway modeling, and conceptual visualization. Using the Medical Information Mart for Intensive Care IV (MIMIC-IV) database, the study investigates how unsupervised learning, event-log analysis, and visualization design can be combined to support the instantiation of clinical digital twins. The proposed framework integrates patient similarity modeling, temporal pathway reconstruction, and user-oriented visualization into a coherent design science research artifact. Although the resulting implementation does not yet constitute a fully realized clinical digital twin, it illustrates how digital twins can be initialized from real-world clinical data, support early predictive modeling, and evolve through iterative refinement. The findings highlight the importance of data granularity, temporal structure, and interpretability, and demonstrate how such models may bridge analytical and clinical perspectives. Future work will extend the framework through richer data inclusion, clinical validation, and real-time deployment.