Figure 1 depicts the data flow of the CNRS. The clinical sequencing report is stored as a part of the pathology report in EHRs. The stored clinical sequencing reports are transferred to the operational data store (ODS). In ODS, the data are structured and standardized, and the sequencing information is saved in the CRDW. Through a key management server, the clinical data in the CRDW and the NGS result data in pathology reports are mapped to patient alternative numbers and not patient numbers. In addition, NGS result data in the CRDW are extracted and stored in the NGS viewer database (DB). Researchers can access the necessary deidentified specific genetic variation cohort in the clinical research portal and inspect NGS result data in the NGS result viewer.

The detailed information of each component is presented in subsequent sections.

The Health Insurance Review & Assessment Service in Korea reimburses laboratory-developed panel sequencing tests that include about 100 genes with 14 mandatory genes. As an example, Table 1 shows the panel genes that are used in the National Cancer Center (NCC), Korea.

The current clinical sequencing report of the NCC is illustrated in Figure 2. Currently, clinical sequencing reports are stored in a single table with 28 attributes in the EHR DB. This table is copied to the ODS on a weekly basis.

Figure 3 demonstrates the data structuring process. All data are structured and standardized according to ISO/TS 20428 during the extract transform and load process [10]. The ISO/TS 20428 standard defines the required and optional fields for sequencing reports, along with the metadata for each field. The required fields include the following 10 categories: clinical sequencing orders, information on the subject of care, information on the legally authorized person ordering clinical sequencing, performing laboratory, associated diseases and phenotypes, biomaterial information, genetic variations, classification of variants, recommended treatment, and addendum. The optional fields include the following seven categories: medical history, family history, reference genome version, racial genomic information, genetic variation, detailed sequencing information, and references.

Figure 3 shows that the clinical sequencing data are structured and loaded from the replicated ODS into pathologic and laboratory information, variant summary, fusion variant, and copy number variation (CNV) variant tables. Pathologic and laboratory information has the following nine attributes: identifiers, test order date, quality control results, sample type, report generation date, report generator information, sequencer type, recommended treatment, and references. Variant summary has the following nine attributes: gene name, exon ID, DNA change, protein change, variant information, allele frequency, effects of variants, pathogeny, and clinical relevance. Fusion variant has the following 11 attributes: gene name, chromosome, cytoband, break, transcript part, locus, gene strand, span read, split read, total read, and distance. CNV has the following nine attributes: gene name, locus, P value, gain region, total region, region ratio, gene count, region count, and significant region count.

Figure 4 shows that the NCC has deidentified the clinical data warehouse as well. The ODS receives clinical data from the table with the primary key as the patient ID and NGS result data from the table with the primary key as pathology ID from EHRs. The key management system receives the pathology ID and patient ID from EHRs and generates an alternative ID. It then sends the ID to the ODS. The ODS deletes the patient ID and pathology ID and sends the data to the CRDW with the alternative ID. By using an alternative ID, which is a pseudonym for the patient ID, the necessary deidentified clinical data can be combined with the sequencing result data.

Users can access the necessary information using a clinical research portal and NGS result viewer. The clinical research portal is a user interface to query or extract clinical and genomic data by changing search options. The NGS result viewer provides functionality to create a structured or standardized clinical sequencing report by converting an original unstructured pathology report using ISO/TS 20428.

This study was approved by the institutional review board (IRB) of the NCC in Korea (NCC2019-0535).

The CNRS converts text-based clinical sequencing reports in EHRs into structured data using international standards. Through the CNRS, the content, as shown in Figure 5, can be easily organized and data can be managed according to international standards. It also provides standardized data through the clinical research search portal; furthermore, using its functions, researchers can set up cohorts with specific mutations and add clinical data columns as needed. Thus, it can be inferred that the CNRS supports researchers in performing translational research by allowing them to easily extract the desired clinical and genomic data from EHRs. For example, non–small cell lung cancer genotyping requires mutation pattern analysis of KRAS, EGFR, and BRAF [1,24]. Similarly, other research requires clinical data such as that on cancer stage, smoking history, and death date for survival analysis [25]. These types of translational studies can be easily performed using the developed CNRS, and this has already been proven by researchers at the NCC.

The CNRS is provided to research projects that have been approved by the IRB. Researchers send data extraction requests to health information managers, and it takes about one to two weeks to review, refine, and provide clinical data from EHRs. The CNRS could retrieve data through the clinical research portal after receiving IRB approval, and it takes about two or three days from review to delivery after a data extraction request is made. In addition, the NCC has built a genomic cohort linking the NGS DB with cancer registries such as those of lung cancer and ovarian cancer.

There are two unique aspects of the CNRS as developed. One is that the CNRS uses an ISO standard (ISO/TS 20428) to support multicenter research. If other hospitals or research institutes use international standards, the data can be easily integrated into the same format. We also achieved the same results as successfully converting the data and manually cleansing the data previously. The other is that the CNRS can help protect patients’ privacy by deidentifying protected health information. To use the patients’ data for research purposes, researchers must obtain written consent from patients or deidentify the identifiable data. In this era of big data, deidentification is usually used for a number of reasons. However, if we deidentify the data, it is difficult to link the separate DBs. To overcome this issue, the CNRS adopted a key management server to pseudonymize the patient ID. This means that the CNRS works as an honest broker [26]. To strengthen the protection level, only authorized developers can access this key management server and users can receive the randomly assigned ID after combing clinical data and sequencing data using the key management system. Therefore, users cannot retrieve real patient IDs in EHRs or pseudonymized IDs in the key management system.

The main contribution of this study is that, for the first time, the ISO/TS 20428 standard was applied to the CRDW to standardize clinical genomic test results. As a result, we also demonstrated that this approach could enable easy search and analysis with clinical data in EHRs. However, it has limitations. We did not verify this system in multiple centers. We hope our approach will help other hospitals or institutions build their own systems.

Our continuing objectives are to extend the categories of clinical and sequencing data in the CNRS and consider the standards proposed by the Global Alliance for Genomics and Health, which has developed diverse practical standard application programming interfaces for international genomic research.

The CNRS converts the text-based clinical sequencing reports of EHRs into structured data using international standards and provides standardized data. In addition, the CNRS allows researchers to set up cohorts with specific mutations and add clinical data columns as needed. Therefore, it can be inferred that the CNRS enables researchers to conduct translational research by allowing them to easily extract the required clinical and genomic data from EHRs.