This paper describes a quasi-experimental three-phase sequential cohort design used in the Study of Patient–Nurse Effectiveness with Assisted Communication Strategies (SPEACS) to test two interventions to improve nurse–patient communication in the intensive care unit (ICU). The sample consists of 10 nurses and 30 nonspeaking ICU patients in each phase (total n=90 nurse–patient dyads). Observational techniques (video recording, transcription, and rating) measure nurse–patient communication performance. Descriptive and covariate data are collected through clinical assessment tools, questionnaires, and field notes. We discuss the practical and scientific considerations in constructing and implementing this type of clinical trial. Specifically, primary threats to validity, history and the Hawthorne effect, are considered and efforts to minimize and track these potential threats are described.