The conflict between random assignment and treatment preference: implications for internal validity

Abstract

The gold standard for most clinical and services outcome studies is random assignment to treatment condition because this kind of design diminishes many threats to internal validity. Although we agree with the power of randomized clinical trials, we argue in this paper that random assignment raises other, unanticipated threats to internal validity as a result of failing to consider treatment preference in research participant behavior. Treatment preference arises from an individual's knowledge and appraisal of treatment options. Treatment preferences impact: (1) the recruitment phase because people consider whether they want to participate in a study that involves the possibility of receiving an undesirable treatment or waiting for treatment, (2) degree of engagement in the intervention condition, and (3) attrition from the study. The benefits and limitations of research strategies that augment randomization while respecting treatment preference are reviewed including: approaches that enhance enrollment and engagement; pilot testing assumptions about randomization; and partially randomized clinical trials.