Reliability enhancement and estimation in multisite clinical trials

Abstract

Clinical trials, particularly those in addictions research, often rely on self-report data for primary dependent variables, and it is imperative to evaluate and to minimize both random and systematic error. This article describes methods for enhancing and assessing reliability of measurement in multisite clinical research. It begins with an overview of the two major approaches to evaluating the reliability of psychiatric diagnosis and verbal self-reports. A model of the interview process is then described, and the major sources of inconsistency that arise in the data collection process are identified. Based on the model, staff selection criteria, training techniques and quality assurance procedures are recommended for enhancing the reliability of interview assessments, and a research design appropriate for evaluating reliability in multisite clinical investigations is proposed.