Medical institutions and researchers frequently collect longitudinal data by conducting a series of surveys over time. Such surveys generally collect a consistent and broad set of data elements from large sets of patients at predefined time points. In contrast to the sparse and irregular retrospective observational data found in electronic medical record (EMR) systems, prospectively gathered survey data captures the same variables at the same time steps across the full study population. Most analyses of this type of longitudinal data focus on understanding the how various properties of the patient cohort associate with specific variables or outcomes measures. However, this approach may miss interesting patterns within constellations of correlated variables. In this paper we describe a visual analysis method for survey data that considers interactions across the full, high-dimensional set of collected variables. Our approach first applies cluster analysis algorithms to survey data collected at each time point independently. We then visualize patient cluster dynamics over time, allowing investigators to identify common patient subgroups and evolution patterns, inspect derived statistical summaries, and compare findings between patient subgroups. We demonstrate our method using data from a survey that followed a cohort of approximately 1,000 patients admitted to the emergency department (ED) following a motor vehicle accident. The survey includes data for each patient at four discrete time points, beginning at admission to the ED and continuing for one year.