Kouz Hicham, Bouaud Jacques, Guézennec Gilles, Séroussi Brigitte
Sorbonne Université, Université Sorbonne Paris Nord, INSERM, LIMICS, UMR_S 1142, Paris, France.
AP-HP, DRCI, Paris, France.
Stud Health Technol Inform. 2020 Nov 23;275:107-111. doi: 10.3233/SHTI200704.
How textual clinical practice guidelines are written may have an impact on how they are formalized and on the kind of recommendations issued by the clinical decision support systems (CDSSs) that implement them. Breast cancer guidelines are mostly centered on the description of the different recommended therapeutic modalities, represented as atomic recommendations, but seldom provide comprehensive plans that drive care delivery. The objective of this work is to implement a knowledge-based approach to develop a care plan builder (CPB) that works on atomic recommendations to build patient-centered care plans as sequences of chronologically ordered therapeutic steps. The CPB uses the atomic recommendations issued by the guideline-based decision support system (GL-DSS) of the DESIREE project. The domain knowledge is represented as the list of all care plans that apply to breast cancer patients. Scenarios are introduced to locate the patient on these theoretical care plans. The CPB has been evaluated on a sample of 99 solved clinical cases leading to an overall performance of 89,8%.
文本临床实践指南的编写方式可能会对其形式化方式以及实施这些指南的临床决策支持系统（CDSS）所发布的推荐类型产生影响。乳腺癌指南大多集中于对不同推荐治疗方式的描述，这些方式以原子推荐的形式呈现，但很少提供推动护理实施的全面计划。这项工作的目标是实施一种基于知识的方法来开发护理计划生成器（CPB），该生成器基于原子推荐来构建以患者为中心的护理计划，这些计划是按时间顺序排列的治疗步骤序列。CPB使用DESIREE项目基于指南的决策支持系统（GL-DSS）发布的原子推荐。领域知识表示为适用于乳腺癌患者的所有护理计划列表。引入了场景以便在这些理论护理计划上定位患者。CPB已在99个已解决的临床病例样本上进行了评估，总体性能为89.8%。
