Psychometric Innovations and Advances in Medical Educational Assessments

Описание: The proposed coordinated session provides an overview of new and innovative psychometric approaches based on data science methods and item response theory (IRT) model extensions for improving the quality, validity, and accuracy of test items, test designs and test scores in the context of medical licensing exams and medical education. In the assessment of cognitive constructs, methods such as natural language processing (NLP), machine learning, feature generation, and cluster analyses of process data provide additional information about examinee ability and item characteristics. This information can be used to improve the test development process and the efficiency of test designs, allow for the automated coding and scoring of open-ended or constructed responses for an increased measurement precision, and a better interpretation of generated process data features in relation to test scores. Moreover, new IRT model extensions allow the modeling of different latent variables related to response biases and measurement error for a more valid interpretation of noncognitive constructs. These new approaches and model extensions are not only helpful for the enhancement of high stakes tests, but can also be used to advance the education and professional development of medical students, young physicians and practitioners prior to and beyond licensure testing.