In 2016, Uruguay started gathering longitudinal student data to improve educational trajectories by putting in place an ‘early alert’ system. Underlying the system is the understanding that prior schooling predicts likelihood of grade repetition and grade repetition predicts later school dropout, while close follow‐up can help prevent both repetition and dropout. We used a database of administrative registries from a national public primary school graduating cohort on their last year in primary and first year in secondary education (2015–2016, n = 36,754). We conducted two‐level cross‐classified logistic regression analyses to assess the suitability of using features of Uruguayan students’ primary school trajectories, individual, family and primary school characteristics to predict their success or failure in their first year of secondary school. All considered prior schooling factors (previous repetition experiences, achievement, behaviour and absenteeism), the student’s family socio‐economic status (SES) and primary school’s SES composition, as well as the location of the school in an urban or rural setting, help explain differences in chances of first‐year success or failure (grade repetition) in secondary school. While these results support the ‘early alert’ system’s approach, predictive performance analyses are needed when using explanatory models for planning interventions with scarce resources and making decisions affecting individual students’ trajectories. The importance of testing resulting models’ sensitivity, as well as their false positive rates, is highlighted.

中文翻译：

---

可以预防失败吗？使用纵向数据识别进入中学的高危学生

2016年，乌拉圭开始建立纵向预警系统，开始收集纵向的学生数据，以改善教育轨迹。该系统的基础是这样的理解，即以前的教育可以预测年级复读的可能性，而年级的重复则可以预测以后的辍学，而密切的跟进可以帮助防止重复和辍学。我们使用了来自国家公立小学毕业队列的行政管理数据库，这些数据库分别是初等教育和初等教育的第一年（2015-2016年，n = 36,754）。我们进行了两级交叉分类的逻辑回归分析，以评估使用乌拉圭学生小学轨迹的特征，个人，家庭和小学特征来预测他们在中学一年级中的成功或失败的适当性。所有考虑过的就学背景（先前的重复经历，成就，行为和旷课），学生的家庭社会经济地位（SES）和小学的SES构成以及学校在城市或农村地区的位置都有助于解释中学一年级成功或失败（复读）机会的差异。这些结果支持“预警”系统的方法，当使用解释性模型来规划资源匮乏的干预措施并做出影响个别学生轨迹的决策时，需要进行预测性绩效分析。强调了测试结果模型的敏感性及其假阳性率的重要性。