Faculdade de Medicina, Universidade Federal da Bahia, Salvador, Bahia, Brazil.
PLoS One. 2011 Jan 5;6(1):e15695. doi: 10.1371/journal.pone.0015695.
Basic courses in most medical schools assess students' performance by conferring scores. The objective of this work is to use a large score databank for the early identification of students with low performance and to identify course trends based on the mean of students' grades.
METHODOLOGY/PRINCIPAL FINDINGS: We studied scores from 2,398 medical students registered in courses over a period of 10 years. Students in the first semester were grouped into those whose ratings remained in the lower quartile in two or more courses (low-performance) and students who had up to one course in the lower quartile (high-performance). ROC curves were built, aimed at the identification of a cut-off average score in the first semesters that would be able to predict low performances in future semesters. Moreover, to follow the long-term pattern of each course, the mean of all scores conferred in a semester was compared to the overall course mean obtained by averaging 10 years of data. Individuals in the low-performance group had a higher risk of being in the lower quartile of at least one course in the second semester (relative risk 3.907; 95% CI: 3.378-4.519) and in the eighth semester (relative risk 2.873; 95% CI: 2.495-3.308). The prediction analysis revealed that an average score of 7.188 in the first semester could identify students that presented scores below the lower quartiles in both the second and eighth semesters (p<0.0001 for both AUC). When scores conferred by single courses were compared over time, three time-trend patterns emerged: low variation, upward trend and erratic pattern.
CONCLUSION/SIGNIFICANCE: An early identification of students with low performance may be useful in promoting pedagogical strategies for these individuals. Evaluation of the time trend of scores conferred by courses may help departments monitoring changes in personnel and methodology that may affect a student's performance.
大多数医学院的基础课程通过评分来评估学生的表现。本研究的目的是利用大量的成绩数据库来早期识别成绩较低的学生，并根据学生成绩的平均值来识别课程趋势。
方法/主要发现：我们研究了 2398 名在 10 年内注册过课程的医学生的成绩。将第一学期的学生分为两个组：成绩在两个或更多课程中始终处于较低四分位数（低绩效）的学生和成绩在一个课程中处于较低四分位数（高绩效）的学生。构建了 ROC 曲线，旨在确定第一学期的平均成绩截断值，以便能够预测未来学期的低成绩。此外，为了跟踪每门课程的长期趋势，将一个学期中所有授予的分数的平均值与通过平均 10 年数据获得的整个课程平均值进行比较。低绩效组的个体在第二学期至少有一门课程处于较低四分位数的风险更高（相对风险 3.907；95%CI：3.378-4.519）和第八学期（相对风险 2.873；95%CI：2.495-3.308）。预测分析表明，第一学期的平均成绩为 7.188 时，可以识别出在第二学期和第八学期的成绩都低于较低四分位数的学生（两个 AUC 的 p 值均<0.0001）。当比较单门课程随时间授予的成绩时，出现了三种时间趋势模式：低变化、上升趋势和不稳定模式。
结论/意义：早期识别成绩较低的学生可能有助于为这些学生制定教学策略。评估课程授予成绩的时间趋势可以帮助系监测可能影响学生成绩的人员和方法的变化。
