IT Center for Clinical Research, University of Luebeck, Luebeck, Germany.
Institute of Medical Informatics, University of Luebeck, Luebeck, Germany.
JMIR Med Inform. 2024 Sep 17;12:e57853. doi: 10.2196/57853.
To ensure interoperability, both structural and semantic standards must be followed. For exchanging medical data between information systems, the structural standard FHIR (Fast Healthcare Interoperability Resources) has recently gained popularity. Regarding semantic interoperability, the reference terminology SNOMED Clinical Terms (SNOMED CT), as a semantic standard, allows for postcoordination, offering advantages over many other vocabularies. These postcoordinated expressions (PCEs) make SNOMED CT an expressive and flexible interlingua, allowing for precise coding of medical facts. However, this comes at the cost of increased complexity, as well as challenges in storage and processing. Additionally, the boundary between semantic (terminology) and structural (information model) standards becomes blurred, leading to what is known as the TermInfo problem. Although often viewed critically, the TermInfo overlap can also be explored for its potential benefits, such as enabling flexible transformation of parts of PCEs.
In this paper, an alternative solution for storing PCEs is presented, which involves combining them with the FHIR data model. Ultimately, all components of a PCE should be expressible solely through precoordinated concepts that are linked to the appropriate elements of the information model.
The approach involves storing PCEs decomposed into their components in alignment with FHIR resources. By utilizing the Web Ontology Language (OWL) to generate an OWL ClassExpression, and combining it with an external reasoner and semantic similarity measures, a precoordinated SNOMED CT concept that most accurately describes the PCE is identified as a Superconcept. In addition, the nonmatching attribute relationships between the Superconcept and the PCE are identified as the "Delta." Once SNOMED CT attributes are manually mapped to FHIR elements, FHIRPath expressions can be defined for both the Superconcept and the Delta, allowing the identified precoordinated codes to be stored within FHIR resources.
A web application called PCEtoFHIR was developed to implement this approach. In a validation process with 600 randomly selected precoordinated concepts, the formal correctness of the generated OWL ClassExpressions was verified. Additionally, 33 PCEs were used for two separate validation tests. Based on these validations, it was demonstrated that a previously proposed semantic similarity calculation is suitable for determining the Superconcept. Additionally, the 33 PCEs were used to confirm the correct functioning of the entire approach. Furthermore, the FHIR StructureMaps were reviewed and deemed meaningful by FHIR experts.
PCEtoFHIR offers services to decompose PCEs for storage within FHIR resources. When creating structure mappings for specific subdomains of SNOMED CT concepts (eg, allergies) to desired FHIR profiles, the use of SNOMED CT Expression Templates has proven highly effective. Domain experts can create templates with appropriate mappings, which can then be easily reused in a constrained manner by end users.
为了确保互操作性，必须遵循结构和语义标准。为了在信息系统之间交换医疗数据，最近流行的结构标准 FHIR（快速医疗互操作性资源）。关于语义互操作性，参考术语 SNOMED 临床术语（SNOMED CT）作为语义标准，允许后协调，相对于许多其他词汇具有优势。这些后协调的表达式（PCE）使 SNOMED CT 成为一种表达力强且灵活的中介语，允许对医疗事实进行精确编码。然而，这是以增加复杂性为代价的，并且在存储和处理方面也存在挑战。此外，语义（术语）和结构（信息模型）标准之间的界限变得模糊，导致出现所谓的 TermInfo 问题。尽管经常受到批评，但 TermInfo 重叠也可以探索其潜在优势，例如能够灵活转换 PCE 的部分内容。
本文提出了一种存储 PCE 的替代解决方案，即将它们与 FHIR 数据模型结合起来。最终，PCE 的所有组件都应该仅通过与信息模型的适当元素相关联的预协调概念来表示。
该方法涉及将 PCE 分解为与其对齐的 FHIR 资源的组件。通过利用 Web 本体语言（OWL）生成 OWL ClassExpression，并将其与外部推理器和语义相似性度量结合使用，确定最准确描述 PCE 的预协调 SNOMED CT 概念作为 Superconcept。此外，还确定了 Superconcept 和 PCE 之间不匹配的属性关系作为“Delta”。一旦手动将 SNOMED CT 属性映射到 FHIR 元素，就可以为 Superconcept 和 Delta 定义 FHIRPath 表达式，从而可以在 FHIR 资源中存储标识的预协调代码。
开发了一个名为 PCEtoFHIR 的 Web 应用程序来实现此方法。在对 600 个随机选择的预协调概念进行验证过程中，验证了生成的 OWL ClassExpressions 的形式正确性。此外，还使用了 33 个 PCE 进行了两次单独的验证测试。基于这些验证，证明了之前提出的语义相似性计算方法适用于确定 Superconcept。此外，使用 33 个 PCE 确认了整个方法的正确功能。此外，FHIR 专家审查并认为 FHIR StructureMaps 有意义。
PCEtoFHIR 提供服务来分解 PCE 以便存储在 FHIR 资源中。在为特定 SNOMED CT 概念（例如过敏）的子域创建到所需 FHIR 配置文件的结构映射时，使用 SNOMED CT Expression Templates 已被证明非常有效。领域专家可以创建具有适当映射的模板，然后最终用户可以以受限的方式轻松重复使用这些模板。
Zotero 插件
