Assessing the Human-Likeness of LLM-Driven Digital Twins in Simulating Health Care System Trust