Threats to validity in assessment: implications in medical education
Abstract
Validity threats in educational assessment are elements that interfere with the proposed interpretation of a test score. They can occur in written tests as well as in performance and clinical competency assessments. They are usually grouped in two major categories: construct underrepresentation and construct-irrelevant variance. The former refers to tests with insufficient items, cases, or observations to make a proper generalization towards the full to-be-assessed domain. The latter is related to the presence of biases that can interfere systematically with the interpretation of a test score, such as item quality and raters’ systematic errors, among other factors that may have an effect on the obtained score. In this paper we describe the characteristics of some of these threats, their importance, and some recommendations to avoid them during the development of assessment instruments in health sciences education. The insights offered can be useful to devise tests and assessment instruments that allow us to draw more valid inferences about students’ knowledge and abilities.