This dissertation explores the decision to access health care, in which individuals must weigh between perceived costs and benefits before choosing to seek care. Each chapter tackles a different aspect of this decision-making process. The first chapter develops a framework to explore how insurance generosity or other selection pressures impact emergency health care utilization. We consider two mechanisms which both can affect health: selection on the extensive margin, where individuals seek health care during health shocks they would not have otherwise; and selection on the intensive margin, where individuals seek health care after reduced delays within a given health shock, thus initiating treatment earlier in acute disease processes. While the former is easily quantified, less is known about the latter due to lack of data and methods. Our intervention is to develop a method to measure severity of illness on presentation to the emergency room using machine learning and electronic health record data. We apply these methods to investigate how changes in insurance coverage around age 65 impact selection into emergency room presentations. The second chapter aims to shed light on a primary motivator which drives people to seek out health care: symptoms. Symptoms, or subjective experiences of patients which can indicate underlying pathology, are difficult to study retrospectively because information is primarily stored in clinical free text, not in structured data fields. We leverage natural language processing to unlock symptom information from clinical narratives. The third chapter examines outpatient utilization patterns following emergency care, comparing trajectories between patients who potentially did and did not receive a surprise out-of-network bill associated with the emergency event. Together, these essays demonstrate how methodological innovation and diverse datasets--electronic health record data, clinical notes, and commercial claims data--can illuminate distinct facets of the decision to seek health care.