Colliver J A, Vu N V, Markwell S J, Verhulst S J
Division of Statistics and Research Consulting, Southern Illinois University School of Medicine, Springfield 62708.
Med Educ. 1991 Jul;25(4):303-10. doi: 10.1111/j.1365-2923.1991.tb00071.x.
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.
本研究旨在深入了解临床能力各独立组成部分（如数据收集、检查结果解读、诊断等）测量的可靠性，这些测量通过基于表现的考试进行，该考试由标准化病人案例组成，施测于南伊利诺伊大学医学院的五个班级的高年级医学生。总体而言，在考试中实际测量的能力（使用实际测量每个能力的案例数量）的可靠性较低，54%低于0.30，75%低于0.40。对于五个班级汇总并推算到共同的k = 10个案例数量的概化系数，九个能力中有两个的可靠性在40多，第三个接近0.40，其余六个在20多。各能力要达到推荐的0.80可靠性所需的案例数量从45到170个不等，九个能力中有六个需要超过100个案例才能达到0.80水平。临床能力这些组成部分测量的低可靠性引发了关于将分数用作学生表现指标的严重问题。
