In today's rapidly evolving, technology-driven and data-rich environment, we are increasingly being offered new information with which to make decisions. This dissertation examines how these changes are reshaping operations in healthcare settings, and seeks to better understand their impact on the system as a whole. We first consider the use of machine-learned predictions of patient risk for triage and prioritization. We model this as a learning-augmented online scheduling problem where we are given good, but imperfect, information about each arriving patient’s urgency level in advance. In this formulation, we face the challenges of decision making under imperfect information, and of responding dynamically to prediction error as we observe better data in real time. We propose a simple online policy for minimizing the total urgency-weighted costs of delay across all patients, and show that this policy is in fact the best possible in certain stylized settings. Our work in the area of scheduling has also prompted some theoretical questions. We present a new proof of correctness of the celebrated Shortest Processing Time rule using geometric insights and linear programming techniques. Finally, we examine the impact of online appointment scheduling platforms that offer patients the ability to observe and choose physicians that best meet their needs. In particular, some patients value the flexibility of seeing readily available physicians while others prefer dedicated service by a primary care provider. We study the effects of added flexibility in a multi-server queueing framework and show that even a small number of flexible patients can greatly benefit overall system performance.