Gonnella Joseph S, Erdmann James B, Hojat Mohammadreza
Center for Research in Medical Education and Health Care, Jefferson Medical College, Thomas Jefferson University, 1025 Walnut Street, Philadelphia, PA 19107, USA.
Med Educ. 2004 Apr;38(4):425-34. doi: 10.1111/j.1365-2923.2004.01774.x.
Context It is important to establish the predictive validity of medical school grades. The strength of predictive validity and the ability to identify at-risk students in medical schools depends upon assessment systems such as number grades, pass/fail (P/F) or honours/pass/fail (H/P/F) systems. Objective This study was designed to examine the predictive validity of number grades in medical school, and to determine whether any important information is lost in a shift from number to P/F and H/P/F grading systems. Subjects The participants in this prospective, longitudinal study were 6656 medical students who studied at Jefferson Medical College over 3 decades. They were grouped into 10 deciles based on their number grades in Year 1 of medical school. Methods Participants were compared on academic accomplishments in Years 2 and 3 of medical school, medical school class rank, delayed graduation and attrition, performance on medical licensing examinations and clinical competence ratings in the first postgraduate year. Results Results supported the short- and longterm predictive validity of the number grades. Ratings of clinical competence beyond medical school were predicted by number grades in medical school. We demonstrated that small differences in number grades are statistically meaningful, and that important information for identifying students in need of remedial education is lost when students who narrowly meet faculty's expectations are included with the rest of the class in a broad 'pass' category. Conclusions The findings refute the argument that knowledge of sciences basic to medicine is not critical to subsequent performance in medical school and beyond if an appropriate evaluation system is used. Furthermore, the results of this study raise questions about abandoning number grades in favour of a pass/fail system. Consideration of these findings in policy decisions regarding assessment systems of medical students is recommended.
背景 确立医学院成绩的预测效度很重要。医学院预测效度的强度以及识别高危学生的能力取决于诸如数字评分、通过/不及格（P/F）或优等/通过/不及格（H/P/F）等评估系统。
目的 本研究旨在检验医学院数字评分的预测效度，并确定从数字评分系统转变为P/F和H/P/F评分系统是否会丢失任何重要信息。
受试者 这项前瞻性纵向研究的参与者是6656名医科学生，他们在杰斐逊医学院学习了30多年。根据他们医学院第一年的数字成绩，他们被分为10个十分位数组。
方法 比较参与者在医学院第二年和第三年的学业成绩、医学院班级排名、延迟毕业和退学情况、医学执照考试成绩以及研究生一年级的临床能力评分。
结果 结果支持了数字成绩的短期和长期预测效度。医学院的数字成绩可预测医学院毕业后的临床能力评分。我们证明，数字成绩上的微小差异具有统计学意义，并且当勉强达到教师期望的学生与班上其他学生被归为宽泛的“通过”类别时，识别需要补习教育学生的重要信息就会丢失。
结论 这些发现驳斥了一种观点，即如果使用适当的评估系统，医学基础科学知识对于医学院及以后的后续表现并不关键。此外，本研究结果对放弃数字评分而采用通过/不及格系统提出了质疑。建议在有关医学生评估系统的政策决策中考虑这些发现。
