Robotic patient simulators are the most commonly used robot in clinical education and provide low-risk, high-fidelity learning experiences. They are life-sized humanoid robots that can simulate human physiological responses. Commercially available RPSs lack realistic facial and social cues, which limits their ability to engage human learners and immerse them in the simulation. This may cause poor skill transfer, which can result in adverse patient outcomes. We address this by introducing an expressive RPS capable of conveying expressivity far beyond the state of the art, including pain and neurological impairment, and a new shared control system to support clinical educators. This paper presents our ongoing work, and discusses its implications for the HRI and medical education communities.