Data model harmonization for the All Of Us Research Program: Transforming i2b2 data into the OMOP common data model
Data model harmonization for the All Of Us Research Program: Transforming i2b2 data into the OMOP common data model
复制标题
DOI：
10.1371/journal.pone.0212463
复制
发表时间：
2019-02-19
期刊：
影响因子：
3.7
通讯作者：
Murphy, Shawn N.
中科院分区：
文献类型：
--
作者：
Klann, Jeffrey G.;Joss, Matthew A. H.;Murphy, Shawn N.
我们所有人研究计划（AOU）正在建立一个全国性的队列，其中包括100万名患者的EHR和基因组数据。数据互操作性对该计划的成功至关重要。AOU正在围绕观察性医学成果伙伴关系（OMOP）数据模型标准化其EHR数据。OMOP是目前用于国家级举措的几种标准数据模型之一。每个模型都是独一无二的，这使得互操作性变得非常困难。i2b2数据仓库和分析平台在全球200多个站点使用，它使用灵活的本体驱动方法进行数据存储。我们以前证明了这个本体系统可以驱动数据重新配置，将数据转换为新的格式，而无需特定于站点的编程。我们之前在我们的12个站点的可扩展的健康研究共享（EMOP）网络上实现了这一点，将i2b2转换为以患者为中心的成果研究网络模型。方法和结果在这里，我们利用我们在i2b2高性能转换方面的投资来支持AOU OMOP数据管道。由于本体论已经获得了广泛的国家利益（通过Accrual to Clinical Trials网络，其他PCORnet网络和Nebraska Lexicon），我们利用了网站对该标准本体论的现有投资。我们开发了一个由ARCH-OMOP本体和OMOP概念映射字典驱动的i2b2到OMOP转换。我们在AOU新英格兰HPO（NEHPO）中演示并验证了我们的方法。首先，我们将i2b2中的假患者数据集转换为OMOP，并通过AOU工具验证数据在结构上符合OMOP。然后，我们将Partners Healthcare数据仓库中的一部分数据转换为OMOP。我们开发了一个评估清单，以确保转换后的数据具有自我完整性（例如，分布具有预期的形状，并且填充了所需的字段），使用OMOP的可视化Achilles数据质量工具。这种i2b2到OMOP的转换用于将NEHPO生产数据发送到AOU。它是开源的，可供其他研究项目使用。
BackgroundThe All Of Us Research Program (AOU) is building a nationwide cohort of one million patients' EHR and genomic data. Data interoperability is paramount to the program's success. AOU is standardizing its EHR data around the Observational Medical Outcomes Partnership (OMOP) data model. OMOP is one of several standard data models presently used in national-scale initiatives. Each model is unique enough to make interoperability difficult. The i2b2 data warehousing and analytics platform is used at over 200 sites worldwide, which uses a flexible ontology-driven approach for data storage. We previously demonstrated this ontology system can drive data reconfiguration, to transform data into new formats without site-specific programming. We previously implemented this on our 12-site Accessible Research Commons for Health (ARCH) network to transform i2b2 into the Patient Centered Outcomes Research Network model.Methods and resultsHere, we leverage our investment in i2b2 high-performance transformations to support the AOU OMOP data pipeline. Because the ARCH ontology has gained widespread national interest (through the Accrual to Clinical Trials network, other PCORnet networks, and the Nebraska Lexicon), we leveraged sites' existing investments into this standard ontology. We developed an i2b2-to-OMOP transformation, driven by the ARCH-OMOP ontology and the OMOP concept mapping dictionary. We demonstrated and validated our approach in the AOU New England HPO (NEHPO). First, we transformed into OMOP a fake patient dataset in i2b2 and verified through AOU tools that the data was structurally compliant with OMOP. We then transformed a subset of data in the Partners Healthcare data warehouse into OMOP. We developed a checklist of assessments to ensure the transformed data had self-integrity (e.g., the distributions have an expected shape and required fields are populated), using OMOP's visual Achilles data quality tool. This i2b2-to-OMOP transformation is being used to send NEHPO production data to AOU. It is open-source and ready for use by other research projects.