Reliability and efficiency of components of clinical competence assessed with five performance-based examinations using standardized patients

Abstract

The present study was conducted to provide in-depth information on the reliabilities of measures of the separate components of clinical competence (e.g. data collection, test interpretation, diagnosis, etc.) assessed by a performance-based examination consisting of standardized-patient cases administered to five classes of senior medical students at Southern Illinois University School of Medicine. In general, the reliabilities of the competencies as they were actually measured on the examination (using the number of cases on which each competency was actually measured) were small, with 54% less than 0.30 and 75% less than 0.40. For generalizability coefficients pooled across the five classes and projected to a common number of k = 10 cases, two of the nine competencies had reliabilities in the 40s, a third was close to 0.40, and the remaining six were in the low 20s. The number of cases needed for the competencies to achieve the recommended reliability of 0.80 ranged from 45 to 170 cases, with six of the nine competencies requiring over 100 cases to reach the 0.80 level. The low reliabilities of these measures of the components of clinical competence raise serious questions about using the scores as indicators of student performance.