Data Resource Profile: Health Insurance Review and Assessment Service Covid-19 Observational Medical Outcomes Partnership (HIRA Covid-19 OMOP) database in South Korea
International Journal of Epidemiology · 2024
DOI: 10.1093/ije/dyae062 ↗Kurz gefasst
Fachartikel in der Zeitschrift International Journal of Epidemiology, 2024. Verfasst von Chungsoo Kim und 5 weiteren Autor:innen. Beteiligte Einrichtung: Department of Biomedical Sciences, Ajou University Graduate School of Medicine, Suwon, Republic of Korea. Hauptthemen laut MEDLINE: COVID-19, Databases, Factual. Die Arbeit wurde bisher 5-mal zitiert (Crossref) und nennt 30 Literaturangaben.
Abstract (Originalsprache)
The Health Insurance Review and Assessment Service Covid-19 Observational Medical Outcomes Partnership (HIRA Covid-19 OMOP) database is a data source converted as a form of OMOP Common Data Model for rapid Covid-19 research based on national claims data. This database obtained representativeness by sampling approximately 20% of the entire Republic of Korea population (9 822 577 patients) and includes data from January 2018 to April 2022. The database includes both infected and uninfected individuals and is linked to the Covid-19 confirmation list of the Korea Centers for Disease Control and Prevention and the death registry of Statistics Korea. HIRA has completed multiple rounds of data open projects using this database for external researchers. The health insurance system in South Korea is known as the National Health Insurance (NHI) system. It is a mandatory insurance system that covers approximately 97% of Korean citizens and residents.1 The NHI system covers most medical expenses, including prescription drugs, diagnostic tests and medical procedures. The Health Insurance Review and Assessment Service (HIRA) in South Korea plays a crucial role in the health care system by monitoring and assessing the quality and appropriateness of medical services provided to patients under the NHI system.2 Its primary mission is to ensure that the medical services provided to patients are of high quality and cost-effective, by reviewing and assessing the appropriateness of medical services, evaluating the quality of care and monitoring healthcare use and costs. Not only their primary mission, but HIRA also has promoted a data open policy of their national claims data for public purposes such as health care and scientific research.3–5 The traditional policy of HIRA when providing healthcare data to researchers was to provide anonymized source data in a remote analytical server or secure research space. However, there was still a risk of data leakage, and it was also difficult due to limited resources. Therefore, HIRA established an internal infrastructure to enable distributed network research,6–8 following the common data model (CDM) approach. This enables distributed data analysis using the same query for health care data from multiple institutions.9–11 For example, during the pandemic of the Covid-19 in 2019, HIRA temporarily disclosed their claims containing information on patients with Covid-19, first in the world. At that time, the disclosed data format was observational medical outcomes partnership (OMOP) CDM maintained by the Observational Health Data Sciences and Informatics (OHDSI).12 Recently, HIRA facilitated distributed research analysis by granting public access to its HIRA Covid-19 OMOP database, including approximately 20% of the Korean population, and made it available for Covid-19 research purposes. This database operates in a distributed research manner, unlike typical data provision methods. Rather than providing data to researchers directly, analytical codes for study are designed based on CDM schemas and code mapping dictionary, and submitted to HIRA. Researchers in HIRA perform the analysis on the researcher's behalf, based on the submitted analytical codes. Afterward, only summary statistics excluding patient-level data are returned to the researcher. This distributed manner minimizes the risk of personal information leakage to data holders. HIRA plans to open data on a regular basis through evaluation after the temporal phases of the data open project. The HIRA Covid-19 OMOP database contains comprehensive information on Covid-19 infection from a representatively sampled Korean population. This includes information on the confirmed Covid-19 patient list, health insurance claims for infected and non-infected patients, and mortality data. Health insurance claims include demographics (sex, age, insurance type), visit type (outpatient or hospitalization), medical history (details about diagnosis, procedure and treatment), examination history and prescription history (detailed information on prescriptions including number of days prescribed and list of medications). All medical cost information is also included. The key and detailed information are presented in Table 1 and Supplementary Table S1 (available as Supplementary data at IJE online). Key information included in the HIRA COVID-19 OMOP database COVID-19, coronavirus disease 2019; HIRA, Health Insurance Review and Assessment; OMOP, Observational Medical Outcomes Partnership; RT-PCR, reverse transcription polymerase chain reaction. A schematic of the source data of HIRA claims is shown in Figure 1. The source data were operated by HIRA claims data warehouse, which includes national insurance claims that have been completed through a review process from 1 January 2018 to 30 April 2022. This covers a period of 52 months. The source data comprise the following five data tables: the claim specification (table named ‘200 T’), treatment/care specification (‘300 T’), diagnosis specification (‘400 T’), prescription details (‘530 T’) and health care institution (‘Ykiho’) tables. The 200 T table contains basic information about each claim, such as the claim key, patient surrogate key, health care institution key, primary and secondary diagnosis codes as Korean Standard Classification of Disease 7th revision (KCD-7), start and end dates of treatment, treatment duration, insurer's liability and details on the patient’s visit to the medical institution. The 300 T table contains data for treatment or care performed; electronic data interchange (EDI) codes for procedures/treatment materials/laboratory order and anatomical therapeutic chemical (ATC) codes for in-hospital drug prescription; and does and frequency of administration. The 400 T table contains information on the diagnosis as KCD-7 and the corresponding medical departments for each claim. The 530 T table comprises the ATC codes, dose, frequency of administration and duration of treatment for out-of-hospital prescribed medications. The Ykiho table encompasses the institution key, regional information and institution type information. Configuration of source data to develop the HIRA Covid-19 OMOP database. In the HIRA Covid-19 OMOP database, HIRA claims data were linked with the national death registries and Covid-19 confirmation lists. HIRA, Health Insurance Review and Assessment Service; Covid-19, coronavirus infectious disease 2019; OMOP, Observational Medical Outcomes Partnership The list of Covid-19 confirmed cases managed by the Korea Disease Control and Prevention Agency was linked to increase the validity of patients with Covid-19. Until March 2022, all Covid-19 cases were confirmed only by reverse transcription polymerase chain reaction (RT-PCR) testing. From April 2022, a rapid antigen test for Covid-19 was allowed in Republic of Korea. Furthermore, the mortality data were also linked with the national death registry of Statistics Korea. All data linkage was conducted before transforming to OMOP CDM. The target patients were selected by age/sex-stratified extraction of 20% of the total patients who had eligibility for NHI in 2021. Based on this population, claims were extracted from 1 January 2018 to 30 April 2022. The extracted data were converted into the OMOP CDM version 5.3.1, which allowed significant information to be organized in a standardized way.13 This conversion involved the creation of standardized clinical data tables, including Person, Condition_occurrence, Device, Drug_exposure, Procedure_occurrence and Visit_occurrence tables, which are commonly used in observational research. The Covid-19 confirmation list from the Korea Disease Control and Prevention Agency was included in the Observation table. Death records from Statistics Korea were linked by HIRA and presented in the Death table. The unique identification number of insured patients was anonymized with alternatives with matching digits when transformed into the OMOP CDM. The coding system in each table used a unique concept_id of the OMOP standardized vocabulary. Following OHDSI’s standard terminology guidelines,14 different vocabularies were used for each domain. A full list of mapped codes was disclosed in the online webpage so that researchers could check the standardized codes.15 For example, information on the Covid-19 diagnosis was included in the Condition_occurrence table as OMOP concept_id 37311061 (Covid-19), 439676 (Coronavirus infection), 4100065 (Disease due to Coronaviridae), and Covid-19 confirmation cases were included in the Observation table as OMOP concept_id 704996 (Patient meets Covid-19 laboratory diagnostic criteria). All conversion rules were followed by OHDSI recommendations and can be confirmed in detail in previous studies.16 Quality check of the converted HIRA Covid-19 OMOP database was also conducted according to OHDSI recommendations; the results of the Data Quality Dashboard17 are presented in Supplementary Figure S1 (available as Supplementary data at IJE online). Table 2 shows the number of individuals and records included in 15 data tables except standardized vocabulary tables. Information on 9 822 577 insured individuals was included in the HIRA Covid-19 database. The Condition_occurrence table including all diagnoses had a total of 2 308 775 200 records, and the drug_exposure table had 2 065 103 898 records of prescribed medications. The Measurement table including laboratory test orders (e.g. blood tests and urinalysis), but not value of the result, showed the highest number of cases at 3 billion. Number of people and records in the HIRA COVID-19 OMOP database COVID-19, coronavirus disease 2019; HIRA, Health Insurance Review and Assessment Service; NA, not available; OMOP, observational medical outcomes partnership. The basic characteristics of the total population and patients with Covid-19 are presented in Table 3. The proportion of women in the total population was 51.1%, but the proportion of women inpatients with Covid-19 was 53.7%. The age with the highest ratio was 40–59 years (32.0%); however in patients the Covid-19, the 0–19 years group had the highest proportion (32.6%). Figure 2, and Supplementary Table S2 (available as Supplementary data at IJE online), show the distribution of patients according to sex and 5-year age group. The proportion of mortality in this database was 14.0% in the total population but 0.3% in the Covid-19 patient group. Distribution of the number of patients by sex and 5-year age group in the HIRA Covid-19 OMOP database. The left green bar indicates the number of females and the right blue bar indicates the number of males. Each bar denotes the age group by 5 years, and the age group was determined based on the mid-year age in 2021. HIRA, Health Insurance Review and Assessment Service; Covid-19, coronavirus infectious disease 2019; OMOP, Observational Medical Outcomes Partnership Basic characteristics of patients in the HIRA COVID-19 OMOP database COVID-19, coronavirus disease 2019; HIRA, health insurance review and assessment service; OMOP, observational medical outcomes partnership. The total number of patients in the database, based on insured persons in 2021. The mid-year number of patients was calculated as of 31 July of the relevant year. Therefore, there is a difference between the total in the database and the mid-year number in 2021. Age group was calculated as of 31 July 2021. As of mid-2021, no patients aged ≥80 years were observed among the patients with COVID-19, because the HIRA COVID-19 OMOP database performed stratified sampling regardless of whether the patient was infected. Hospitalization refers to the number of patients with inpatient records. Death refers to the number of patients with death records. In the early period of the Covid-19 pandemic, the data of patients with Covid-19 were disclosed for the first time (pilot phase) by HIRA and presented in high-impact journals.18–25 HIRA intended to promote the regular opening of the HIRA Covid-19 OMOP database through two or more phases of temporal opening projects and evaluations. For the first temporal opening (from July 2022 to June 2023), more than 40 research projects have been conducted in a distributed manner. As a result, studies evaluating the drug effectiveness or the correlation between Covid-19 and specific diseases have been published.26,27 HIRA is currently initiating the second temporal opening phase of the HIRA Covid-19 OMOP database. The HIRA Covid-19 OMOP database has several advantages for researchers. First, the national claims database supports researchers to conduct tailored observational research on Covid-19. Research on the health care service, research on optimal treatments for patients with Covid-19, and research in accordance with Covid-19 quarantine policies is actively conducted. Moreover, it is possible to conduct cost-effectiveness analysis on Covid-19 because this database has all the cost information for reimbursement, making it ideal for cost-effectiveness evaluation studies on treatment. Second, the privacy protection of HIRA Covid-19 OMOP Database is excellent because only the summarized results from pre-specified analytics package are provided to researchers, in line with the distributed network research. To prevent leakage of personal information, results containing individual identifiers and statistics by individual/claims/health care institutions were prohibited from being exported. Last, HIRA Covid-19 OMOP database is encouraging collaborative study based on the OMOP structure. Multinational, multi-institutional research is essential for infectious disease, a disease of global common interest. Time-consuming preprocessing steps caused by heterogeneity of data structure between organizations can be solved in advance and used immediately for joint research. As many electronic medical records in the Republic of Korea are currently building OMOP infrastructure, and it is being actively adopted worldwide, we believe that the adoption of an OMOP structure is an appropriate choice for multicentre collaborative research. However, the current data resource has some limitations. To be a database that faithfully contributes to latest research, the following data may be additionally required: detailed variant information of Covid-19, vaccine data and non-reimbursement items including non-prescribing medications. Due to the nature of the claim data, the source database of HIRA has only limited laboratory test results. However, RT-PCR confirmed Covid-19 cases were linked using external data sources to provide clinical validity. HIRA source data depend on the NHI reimbursement list; therefore, non-reimbursement items and over-the-counter medications sold at pharmacies are not determined. This has the potential to serve as a confounder in health care research. For example, there is a limitation that the effect of over-the-counter medications cannot be considered in the risk comparison of drugs such as non-steroidal anti-inflammatory drugs (NSAIDs) in patients with Covid-19. However, this is a common limitation of claims data and is not unique to the HIRA Covid-19 OMOP database. Other limitations include the use of a distributed research network strategy in HIRA, so individual-level pooling analysis from multiple data sources is not possible. However various distributed research methodologies for OMOP CDM are being developed to compensate for this.24,28–30 Furthermore, there is a possibility of miscommunication or unintended results during the code submission process between researchers and analysts of HIRA. However, HIRA is trying to minimize misunderstandings and errors by simultaneously sharing not only results but logs from the statistical programme. Researchers who desire to use the database should obtain approvals from both ethical review boards and the HIRA research committee. The HIRA Covid-19 OMOP database is physically located inside HIRA and is maintained by the Big Data Department of the HIRA. HIRA conducts an annual update for this database. Additional release may be conducted if deemed necessary for a particular research question. Official application for HIRA CDM studies can be made through the HIRA Bigdata website [https://opendata.hira.or.kr]. Any enquiries about its use should be directed to co-first author D.H.Yu. [[email protected]]. HIRA is a public organization that can process personal information under the relevant laws of the Republic of Korea. Since the construction of data infrastructure itself does not fall under human subject research, deliberation by the institutional bioethics committee was exempted. However, for studies by individual researchers (both internal and external researchers), a review by the institutional review board is mandatory. This study is also reviewed by respective scientific and ethics committees (Health Insurance Review and Assessment Institutional Review Board: 2023046–001). The HIRA data underlying this article will be shared on reasonable request to the Big Data Department of HIRA. Please see Data Resource Access above. Supplementary data are available at IJE online. C.K. and D.H.Y. wrote the manuscript, .J.C., S.C.Y. and D.H.Y. designed the conversion logic of the database; D.H.Y. and H.B. converted data; C.K. and D.H.Y. constructed the analytical environment; and C.K., S.C.Y. and R.W.P. provided supervision for all processes. All authors reviewed the manuscript and provided critical revisions. None declared. This work was supported by the Health Insurance Review and Assessment Service (HIRA). The views expressed are those of the authors and not necessarily those of the HIRA. R.W.P. reports grant from the Ministry of Health & Welfare (Korea). S.C.Y. reports being CTO of PHI Digital health care.
Schlagwörter
MeSH-Schlagwörter (MEDLINE)
Autor:innen und Institutionen
- Department of Biomedical Sciences, Ajou University Graduate School of Medicine, Suwon, Republic of Korea.
- Big Data Department, Health Insurance Assessment and Review Services, Wonju, Republic of Korea.
- Department of Research, Keimyung University Dongsan Medical Center, Daegu, Republic of Korea.
- Department of Biomedical Systems Informatics, Yonsei University College of Medicine, Seoul, Republic of Korea.
- Institute for Innovation in Digital Healthcare, Yonsei University, Seoul, Republic of Korea.
- Department of Biomedical Informatics, Ajou University School of Medicine, Suwon, Republic of Korea.
ORCID-Profile: Chungsoo Kim, Rae Woong Park
Zitieren
APA
Kim, C., Yu, D. H., Baek, H., Cho, J., You, S. C., & Park, R. W. (2024). Data Resource Profile: Health Insurance Review and Assessment Service Covid-19 Observational Medical Outcomes Partnership (HIRA Covid-19 OMOP) database in South Korea. International Journal of Epidemiology. https://doi.org/10.1093/ije/dyae062
Vancouver
Kim C, Yu DH, Baek H, Cho J, You SC, Park RW. Data Resource Profile: Health Insurance Review and Assessment Service Covid-19 Observational Medical Outcomes Partnership (HIRA Covid-19 OMOP) database in South Korea. International Journal of Epidemiology. 2024. doi:10.1093/ije/dyae062
BibTeX
@article{Kim2024,
author = {Kim, Chungsoo and Yu, Dong Han and Baek, Hyeran and Cho, Jaehyeong and You, Seng Chan and Park, Rae Woong},
title = {Data Resource Profile: Health Insurance Review and Assessment Service Covid-19 Observational Medical Outcomes Partnership (HIRA Covid-19 OMOP) database in South Korea},
journal = {International Journal of Epidemiology},
year = {2024},
doi = {10.1093/ije/dyae062}
}
EUVIMED