Krämer Stefan, Flöge A, Handt S, Juzek-Küpper F, Vogt K, Ullmann J, Rauen T
Medizinische Klinik II für Nieren- und Hochdruckkrankheiten, rheumatische und immunologische Erkrankungen, Uniklinik der RWTH Aachen, Aachen, Deutschland.
Z Rheumatol. 2025 Apr;84(3):169-178. doi: 10.1007/s00393-024-01550-7. Epub 2024 Aug 16.
The timely allocation of appointments for new patients is a daily challenge in rheumatological practice, which can be supported by digital solutions. The question is to find the simplest and most effective possible method for prioritization when allocating appointments.
Using a registration form for new patients, standardized symptoms and laboratory results were collated. After reviewing this information by a medical specialist the allocation of appointments was carried out in three categories: a) < 6 weeks, b) 6 weeks up to 3 months and c) > 3 months. The waiting time between the time of registration and the presentation appointment was calculated and compared between patients with and without a diagnosis of an inflammatory rheumatic disease (IRD). In addition a decision tree (DT), a method taken from the field of supervised learning within artificial intelligence (AI), was established and the resulting classification was compared with respect to the accuracy and calculated saving in waiting time.
In this study 800 appointments between 2020 and 2023 (including 555 women, 69.4%, median age 53 years, interquartile range, IQR 39-63 years) were analyzed. An IRD could be confirmed in 409 (51.1%) cases with a waiting time of 58 vs. 93 days for non-IRD cases (-38%, p < 0.01). An AI-based stratification resulted in an accuracy of 67% for IRD and a predicted saving of 19% waiting time. The accuracy increased up to 78% with a time saving for IRD cases of up to 31%, when all basic laboratory results were known. Simplified algorithms, e.g., stratification by the use of laboratory findings alone, resulted in a lower accuracy and time savings.
Manual allocation of appointments by a medical specialist is effective and significantly reduces the waiting times for patients with IRD. An automated categorization can lead to a reduction in waiting times for appointments when taking complete laboratory results and a lower sensitivity into consideration.
为新患者及时安排预约是风湿病科日常面临的一项挑战，数字解决方案可为其提供支持。问题在于在安排预约时找到最简单、最有效的优先排序方法。
通过新患者登记表整理标准化症状和实验室检查结果。医学专家审核这些信息后，将预约安排分为三类：a）<6周，b）6周至3个月，c）>3个月。计算登记时间与就诊预约时间之间的等待时间，并比较已诊断和未诊断炎性风湿性疾病（IRD）的患者之间的等待时间。此外，建立了一种决策树（DT），这是一种从人工智能（AI）监督学习领域借鉴的方法，并将所得分类在准确性和计算出的等待时间节省方面进行比较。
本研究分析了2020年至2023年期间的800个预约（包括555名女性，占69.4%，中位年龄53岁，四分位间距IQR为39 - 63岁）。409例（51.1%）确诊为IRD，IRD患者的等待时间为58天，非IRD患者为93天（-38%，p<0.01）。基于AI的分层对IRD的准确率为67%，预计等待时间节省19%。当所有基本实验室检查结果已知时，准确率提高至78%，IRD患者的时间节省高达31%。简化算法，例如仅使用实验室检查结果进行分层，准确率和时间节省较低。
医学专家手动安排预约是有效的，可显著减少IRD患者的等待时间。考虑完整实验室检查结果和较低敏感性时，自动分类可减少预约等待时间。
