发表机构
Berlin Institute of Health at Charité - Universitätsmedizin Berlin; Freie Universität Berlin; Heidelberg University Hospital and BioQuant; Peter L. Reichertz Institute for Medical Informatics of TU Braunschweig and Hannover Medical School; VeraTech for Health; Institute of Medical Informatics, Charité - Universitätsmedizin Berlin; Clinical Study Center, Berlin Institute of Health at Charité - Universitätsmedizin Berlin; Maastricht DataHub, Maastricht University; Ludwig Boltzmann Institute for Digital Health and Prevention; Institute of Health Policy, Management and Evaluation, Dalla Lana School of Public Health, University of Toronto（柏林夏里特大学医学附属柏林健康研究所; 柏林自由大学; 海德堡大学医院与生物定量中心; 布伦瑞克工业大学和汉诺威医学院彼得·L·雷希茨医疗信息学研究所; VeraTech for Health; 柏林夏里特大学医学附属柏林医学信息学研究所; 柏林夏里特大学医学附属柏林健康研究所临床研究数据中心; 马斯特里赫特大学马斯特里赫特数据枢纽; 路德维希·玻尔兹曼数字健康与预防研究所; 多伦多大学达拉·兰纳公共卫生学院卫生政策、管理与评估研究所）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
该研究针对openEHR与OMOP CDM的互操作性问题，实施新一代Eos与OMOCL扩展映射，评估发现部分术语无法匹配、临床概念在OMOP中碎片化，为两大生态系统融合提供依据。
AI 中文摘要
背景：临床与研究数据系统间的互操作性对电子健康记录（EHR）数据的二次利用至关重要。openEHR标准提供结构化、模型驱动的临床信息，而OMOP Common Data Model（CDM）支持大规模观察性分析。Eos引擎与OMOP转换语言（OMOCL）此前引入了基于标准的转换方法，但有限的值集支持、僵化的就诊生成及不完整的映射覆盖限制了其更广泛的适用性。方法：实施新一代Eos与OMOCL以提升语义完整性并解决早期局限性。新增功能支持通过conceptMaps映射内部openEHR值集、使用原型查询语言（AQL）生成就诊事件，并扩展了国际原型映射库。通过评估映射覆盖度、术语完整性和领域分布对该框架进行评估，同时使用代表性原型映射检查OMOP的结构约束。结果：映射了196个openEHR原型，覆盖国际临床知识管理器中所有具有OMOP等效表的稳定原型。8.65%的主要概念标识符无法链接到OMOP标准术语。大多数映射针对测量领域（50.5%）和观察领域（41.0%）。结构分析显示，连贯的临床概念常需跨多个松散关联的OMOP表进行拆分；仅问题/诊断原型就需要20多个关联记录。结论：新框架增强了openEHR-OMOP的互操作性并减少了信息丢失。然而，OMOP内部的结构与语义局限性会导致碎片化，可能影响下游分析，表明两大生态系统需要更强的融合。
英文摘要
Background: Interoperability between clinical and research data systems is essential for enabling secondary use of EHR data. The openEHR standard provides structured, model-driven clinical information, while the OMOP Common Data Model (CDM) supports large-scale observational analytics. The Eos engine and OMOP Conversion Language (OMOCL) previously introduced a standards-based transformation approach, but limited value set support, rigid visit generation, and incomplete mapping coverage restricted broader applicability. Methods: A new generation of Eos and OMOCL was implemented to improve semantic completeness and address earlier limitations. New functionality enables mapping of internal openEHR value sets via conceptMaps, supports visit occurrence generation using Archetype Query Language (AQL), and expands the international archetype mapping library. The framework was evaluated by assessing mapping coverage, terminology completeness, and domain distribution. Structural constraints of OMOP were examined using representative archetype mappings. Results: 196 openEHR archetypes were mapped, covering all stable archetypes in the international Clinical Knowledge Manager with OMOP-equivalent tables. 8.65% of primary concept identifiers could not be linked to OMOP standard terminologies. Most mappings targeted the Measurement (50.5%) and Observation (41.0%) domains. Structural analysis showed that coherent clinical concepts often required fragmentation across multiple loosely connected OMOP tables; the Problem/Diagnosis archetype alone required more than 20 linked records. Conclusions: The new framework strengthens openEHR-OMOP interoperability and reduces information loss. However, structural and semantic limitations within OMOP introduce fragmentation that may affect downstream analytics, suggesting a need for greater convergence between both ecosystems.
