This article presents a new model for constructing annual schedules for medical residents based on the regulations of a German teaching hospital as well as the program restrictions of the German Medical Association. Since resident programs of physicians do not only vary between disciplines but also between countries, it is essential to evaluate the main characteristics of the program. The main difference between the already well-studied resident programs in the US and the one of this article is the task-related structure. Residents need to perform different interventions several times to become specialists. This study will focus on Germany since there was a judgement in 2015 that hospital management needs training schedules guaranteeing the success of the resident program in time. Therefore, a new formulation of a tactical resident scheduling problem is presented. The problem is formulated in two stages considering the total number of interventions, equal progress in training as well as continuity of care. As the second stage of our formulation is a quadratic program and even by linearization standard solvers are not able to generate high-quality solutions within 24 h, a genetic algorithm using standard crossovers is developed for the second stage constructing annual schedules for an existing stock of residents. We evaluate our algorithm by comparing the solutions of the genetic algorithm and standard software with a real-world situation of a German training hospital from 2016.

中文翻译：

---

任务相关计划中麻醉医学居民的年度计划，重点是护理的连续性

本文介绍了一种根据德国教学医院的规定以及德国医学会的计划限制构建医疗居民年度时间表的新模型。由于医师的常驻计划不仅在学科之间而且在国家之间也不同，因此评估该计划的主要特征至关重要。在美国，经过精心研究的居民计划与本文之一之间的主要区别是与任务相关的结构。居民需要多次执行不同的干预措施才能成为专家。这项研究将集中在德国，因为在2015年有一项判断，即医院管理层需要培训时间表以确保住院医师计划及时获得成功。因此，提出了战术居民调度问题的新表述。考虑到干预的总数，培训的均等进展以及护理的连续性，问题分两个阶段提出。由于我们制定的第二阶段是二次程序，即使通过线性化，标准求解器也无法在24小时内生成高质量的解决方案，因此第二阶段开发了使用标准交叉法的遗传算法，为现有库存的年度计划构建居民。我们通过将遗传算法和标准软件的解决方案与2016年以来德国培训医院的实际情况进行比较来评估我们的算法。在培训和护理连续性上均等。由于我们的公式编制的第二阶段是二次程序，即使通过线性化，标准求解器也无法在24小时内生成高质量的解决方案，因此第二阶段开发了使用标准交叉法的遗传算法，为现有库存的年度计划编制了时间表居民。我们通过将遗传算法和标准软件的解决方案与2016年以来德国培训医院的实际情况进行比较来评估我们的算法。在培训和护理连续性上均等。由于我们的公式编制的第二阶段是二次程序，即使通过线性化，标准求解器也无法在24小时内生成高质量的解决方案，因此第二阶段开发了使用标准交叉法的遗传算法，为现有库存的年度计划编制了时间表居民。我们通过将遗传算法和标准软件的解决方案与2016年以来德国培训医院的实际情况进行比较来评估我们的算法。